Opportunities for Price Manipulation by Aggregators in Electricity Markets

Size: px
Start display at page:

Download "Opportunities for Price Manipulation by Aggregators in Electricity Markets"

From this document you will learn the answers to the following questions:

  • What is the second stage of an aggregator?

  • What do aggregators have the ablty to curtal?

  • What does the term " LMP " mean?

Transcription

1 1 Opportuntes for Prce Manpulaton by Aggregators n Electrcty Markets Navd Azzan Ruh, Krshnamurthy Dvjotham, Nangjun Chen, and Adam Werman arxv: v1 [cs.gt] 21 Jun 216 Abstract Aggregators are playng an ncreasngly crucal role n the ntegraton of renewable generaton n power systems. However, the ntermttent nature of renewable generaton makes market nteractons of aggregators dffcult to montor and regulate, rasng concerns about potental market manpulaton by aggregators. In ths paper, we study ths ssue by quantfyng the proft an aggregator can obtan through strategc curtalment of generaton n an electrcty market. We show that, whle the problem of maxmzng the beneft from curtalment s hard n general, effcent algorthms exst when the topology of the network s radal (acyclc). Further, we hghlght that sgnfcant ncreases n proft are possble va strategc curtalment n practcal settngs. Index Terms Aggregators, renewables, optmal curtalment, market power, locatonal margnal prce (LMP). I. INTRODUCTION Increasng the penetraton of dstrbuted, renewable energy resources nto the electrcty grd s a crucal part of buldng a sustanable energy landscape. To date, the enttes that have been most successful at promotng and facltatng the adopton of renewable resources have been aggregators, e.g. as SolarCty, Tesla, Enphase, Sunnova, SunPower, ChargePont [1] [3]. These aggregators nstall and manage rooftop solar nstallatons as well as household energy storage devces and electrc vehcle chargng systems. Some have fleets wth upwards of 8 MW dstrbuted energy resources [4], [5], and the market s expected to trple n sze by 22 [6], [7]. Aggregators play a varety of mportant roles n the constructon of a sustanable grd. Frst, and foremost, they are on the front lnes of the battle to promote nstallaton of rooftop solar and household energy storage, pushng for wde-spread adopton of dstrbuted energy resources by households and busnesses. Second, and just as mportantly, they provde a sngle nterface pont where utltes and Independent System Operators (ISOs) can nteract wth a fleet of dstrbuted energy resources across the network n order to obtan a varety of servces, from renewable generaton capacty to demand response. Ths servce s crucal for enablng system operators to manage the challenges that result from unpredctable, ntermttent renewable generaton, e.g., wnd and solar. However, n addton to the benefts they provde, aggregators also create new challenges both from the perspectve of the aggregator and the perspectve of the system operator. On the sde of the aggregator, the management of a geographcally dverse fleet of dstrbuted energy resources s a dffcult algorthmc challenge. On the sde of the operator, the partcpaton of aggregators n electrcty markets presents unque The authors are wth the Department of Computng and Mathematcal Scences, Calforna Insttute of Technology, Pasadena, CA, USA e- mal: {azzan,dvj,ncchen,adamw}@caltech.edu Manuscrpt receved Month DD, YYYY; revsed Month DD, YYYY. challenges n terms of montorng and mtgatng the potental of exercsng market power. In partcular, unlke tradtonal generaton resources, the ISO cannot verfy the avalablty of the generaton resources of aggregators. Whle the repar schedule of a conventonal generator can be made publc, the downtme of a solar generaton plant and the tmes when solar generaton s not avalable cannot be scheduled or verfed after the fact. Thus, aggregators have the ablty to strategcally curtal generaton resources wthout the knowledge of the ISO, and ths potentally creates sgnfcant opportuntes for them to manpulate prces. These ssues are partcularly salent gven current proposals for dstrbuton systems. Dstrbuton systems (whch are typcally radal networks) are heavly mpacted by the ntroducton of dstrbuted energy resources. As a result, there are a varety of current proposals to start dstrbuton-level power markets (see, for example [8] [9]), operated by Dstrbuton System Operators (DSOs). A future grd may even nvolve a herarchy of system operators dealng wth progressvely larger areas, net load and net generaton. In such a scenaro, aggregators could end up havng a sgnfcant proporton of the market share, and such markets may be partcularly vulnerable to strategc bddng practces of the aggregators. Thus, understandng the potental for these aggregators to exercse market power s of great mportance, so that regulatory authortes can take approprate steps to mtgate t as needed. A. Summary of Contrbutons Ths paper addresses both the algorthmc challenge of managng an aggregator and the economc challenge of measurng the potental for an aggregator to manpulate prces. Specfcally, ths work provdes a new algorthmc framework for managng the partcpaton of an aggregator n electrcty markets, and uses ths framework to evaluate the potental for aggregators to exercse market power. To those ends, the paper makes three man contrbutons. Frst, we ntroduce a new model for studyng the management of an aggregator. Concretely, we ntroduce a model of an aggregator n a two-stage market, where the frst stage s the ex-ante (or day-ahead) market that decdes the generaton schedule based on the soluton to a securty-constraned economc dspatch problem and the second stage s the ex-post (or real-tme) market to perform fne adjustment va Locatonal Margnal Prces (LMPs) based on updated nformaton. Second, we provde a novel algorthm for managng the partcpaton of an aggregator n the two stage market. The problem s NP-hard n general and s a blevel quadratc program, whch are notorously dffcult n practce. However, we develop an effcent algorthm that can be used by the

2 2 aggregators n radal networks to approxmate the optmal curtalment strategy and maxmze ther proft (Secton IV). Note that the algorthm s not just relevant for aggregators; t can also be used by the operator to asses the potental for strategc curtalment. The key nsght n the algorthm s that the optmzaton problem can be decomposed nto local peces and be solved approxmately usng a dynamc programmng over the graph. Thrd, we quantfy opportuntes for prce manpulaton (va strategc curtalment) by aggregators n the two stage market (Secton III). Our results hghlght that, n practcal scenaros, strategc curtalment can have a sgnfcant mpact on prces, and yeld much hgher profts for the aggregators. In partcular, the prces can be mpacted up to a few tens of $/MWh n some cases, and there s often more than 25% hgher proft, even wth curtalments lmted to 1%. Further, our results expose a connecton between the proft achevable va curtalment and a new market power measure ntroduced n [1]. B. Related Work Ths paper connects to, and bulds on, work n four related areas: () Quantfyng and mtgatng market power, () Cyberattacks n the grd, () Algorthms for managng dstrbuted energy resources, and (v) Algorthms for blevel programs. Quantfyng market power n electrcty markets: There s a large volume of lterature that focuses on dentfyng and measurng market power for generators n an electrcty market, see [11] for a recent survey. Early works on market power analyss emerged from mcroeconomc theory suggest measures that gnore transmsson constrants. For example, [12] ntroduced the pvotal suppler ndex (PSI), whch s a bnary value ndcatng whether the capacty of a generator s larger than the supply surplus, [13] later refned PSI by proposng resdental supply ndex (RSI). RSI s used by the Calforna ISO to assure prce compettveness [14]. The electrcty relablty councl of Texas uses the element compettveness ndex (ECI) [15], whch s based on the Herfndahl-Hrschmann ndex (HHI) [16]. Market power measures consderng transmsson constrants have emerged more recently. Some examples nclude, e.g., [17] [21], and [22]. Interested readers can refer to [23], whch proposes a functonal measure that unfes the structural ndces measurng market power on a transmsson constraned network n the prevous work. In contrast to the large lterature dscussed above, the lterature focused on market power of renewable generaton producers s lmted. Exstng works such as [1] and [24] study market power of wnd power producers gnorng transmsson constrants. The key dfferentator of the work n ths paper s that the use of the Locatonal Margnal Prce (LMP) framework, whch s standard practce n the electrcty market [25], [26], allows ths work to offer nsght about market power of aggregators when transmsson capacty s lmted. Cyber-attacks n the grd: The model and analyss n ths paper s also strongly connected to the cyber securty research communty, whch has studed how and when a malcous party can manpulate the spot prce n electrcty markets by compromsng the state measurement of the power grd va false data njecton, e.g., see [27] [31]. In partcular, [29], [3] shows that f a malcous party can corrupt sensor data, then t can create an arbtrage opportunty. Further, [27] shows that such attacks can mpact both the real tme spot prce and future prces by causng lne congestons. In ths paper, we do not allow aggregators to corrupt the state measurements of the power system, rather we consder a perfectly legal approach for prce manpulaton: strategc curtalment. However, strategc curtalment n the ex-post market can gan extra proft to the detrment of the power system, whch s a smlar mechansm to those hghlghted n cyber attack lterature. Techncally, the work n ths paper makes sgnfcant algorthmc contrbutons to the cyber-attack lterature. In partcular, the papers mentoned above focus on algorthmc heurstcs and do not provde formal guarantees. In contrast, our work presents a polynomal tme algorthm that provably maxmzes the proft of the aggregator. Algorthms for managng dstrbuted energy resources: There has been much work studyng optmal strateges for managng demand response and dstrbuted generaton resources to offer regulaton servces to the power grd. Ths work covers a varety of contexts. For example, researchers have studed frequency regulaton [32] [33] and voltage regulaton (or volt-var control) [34] [35]. A separate lne of work has been work on desgnng ncentves to encourage dstrbuted resources to provde servces to the power grd [36] [37]. However, the current paper s dstnct from all the work above n that we study strategc behavor by an aggregator of dstrbuted resources. Pror work does not model the strategc manpulaton of prces by the aggregator. Algorthms for blevel programs: The optmzaton problem that the strategc aggregator solves s a blevel program, snce the objectve (aggregator s proft) depends on the locatonal prces (LMPs). The LMPs are constraned to be equal to optmal dual varables arsng from economc dspatch-based market clearng procedure. These types of problems have been extensvely studed n the lterature, and fall under the class of Mathematcal Programs wth Equlbrum Constrants (MPECs) [38]. Even f the optmzaton problems at the two levels s lnear, the problem s known to be NP-hard [39]. Global optmzaton algorthms [4] can be used to solve ths problems to arbtrary accuracy (compute a lower bound on the objectve wthn a specfed tolerance of the global optmum). However, these algorthms use a spatal branch and bound strategy, and can take exponental tme n general. In contrast, solvers lke PATH [41], whle practcally effcent for many problems, are only guaranteed to fnd a local optmum. In ths paper, we show that for tree-structured networks (dstrbuton networks), an ɛ-approxmaton of the global optmum can be computed n tme lnear n the sze of the network and polynomal n 1 ɛ. II. SYSTEM MODEL In ths secton we defne the power system model that serves as the bass for the paper and descrbe how we model the way the Independent System Operator (ISO) computes the Locatonal Margnal Prces (LMPs). Locatonal margnal prcng s

3 3 adopted by the majorty of power markets n the Untes States [26], and our model s meant to mmc the operaton of twostage markets lke ISO New England, PJM Interconnecton, and Mdcontnent ISO, that use ex-post prcng strategy for correctng the ex-ante prces [25], [26]. We consder a power system wth n nodes (buses) and t transmsson lnes. The generaton and load at node are denoted by p and d respectvely, wth p = [ ] T p 1,..., p n and d = [ ] T d 1,..., d n. We use [n] to denote the set of buses {1,..., n}. The focus of ths paper s on the behavor of an aggregator, whch owns generaton capacty, possbly at multple nodes. We assume that the aggregator has the ablty to curtal generaton wthout penalty, e.g., by curtalng the amount of wnd/solar generaton or by not callng on demand response opportuntes. Let N a [n] be the nodes where the aggregator has generaton and denote ts share of generaton at node N a by p a (out of p ). The curtalment of generaton at ths node s denoted by α, where α p a. We defne our model for the decson makng process of the aggregator wth respect to curtalment n Secton III. Together, the net generaton delvered to the grd s represented by p α, where α j = j N a. The flow of lnes s denoted by f = [ ] T f 1,..., f t, where fl represents the flow of lne l: f = G(p α d), where G R t n s the matrx of generaton shft factors [42]. We also defne B R n t as the lnk-to-node ncdence matrx that transforms lne flows back to the net njectons as p α d = Bf. A. Real-Tme Market Prce At the end of each dspatch nterval, n real tme, the ISO obtans the current values of generaton, demands and flows from the state estmator. Based on ths nformaton, t solves a constraned optmzaton problem for market clearng. The objectve of the optmzaton s to mnmze the total cost of the network, based on the current state of the system. The ex-post LMPs are announced as a functon of the optmal Lagrange multplers of ths optmzaton. Mathematcally, the followng program has to be solved. mnmze f c T Bf (1a) λ, λ + : p Bf p + α + d p (1b) µ, µ + : f f f (1c) ν : f range(g) (1d) In the above, c s the offer prce for the generator. f l s the desred flow of lne l, and Bf = p + p α d, where p s the desred amount of change n the generaton of node. Constrant (1b) enforces the upper and lower lmts on the change of generatons, and constrant (1c) keeps the flows wthn the lne lmts. In practce, the generaton change lmts are often set to be constant values of p = 2 and p =.1, for all, [43]. The last constrant ensures that f l are vald flows,.e. f = G p for some generaton p. Varables λ, λ + R n +, µ, µ + R t + and ν R t rank(g) denote the Lagrange multplers (dual varables) correspondng to constrants (1b), (1c) and (1d). Note that the ISO does not physcally redspatch the generatons, and the optmal values of the above program are just the desred values. In fact, by announcng the (ex-post) LMPs, the ISO provdes ncentves for the generators to adjust ther generaton accordng to the goals of the ISO [26]. Defnton 1. The ex-post locatonal margnal prce (LMP) of node at curtalment level of α, denoted by λ (α), s λ (α) = c + λ + (α) λ (α). (2) We assume that there s a unque optmal prmal-dual par, and therefore the LMPs are unque. In general, there are several ways that ISOs ensure such condton, for nstance by addng a small quadratc term to the objectve. III. THE MARKET BEHAVIOR OF THE AGGREGATOR The key feature of our model s the behavor of the aggregator. As mentoned before, aggregators have generaton resources at multple locatons n the network and can often curtal generaton resources wthout the knowledge of the ISO. Of course, such curtalment may not be n the best nterest of the aggregator, snce t means offerng less generaton to the market. But, f through curtalment, prces can be mpacted, then the aggregator may be able to receve hgher prces for the generaton offered or make money through arbtrage of the prce dfferental. In ths secton, we develop a model for a strategc aggregator, whch we then use n order to understand n what stuatons curtalment may be benefcal. A. Prelmnares To quantfy the proft that the aggregator makes due to the curtalment, let us take a look at the total revenue n dfferent producton levels. Defnton 2. We defne the curtalment proft (CP) as the change n proft of the aggregator as a result of curtalment: γ(α) = N a (λ (α) (p a α ) λ () p a ) (3) Note that the curtalment proft can be postve or negatve n general. We say a curtalment level α > s proftable f γ(α) s strctly postve. The curtalment proft s mportant for understandng when t s benefcal for the aggregators to curtal. Note that we are not concerned about the cost of generaton here, as renewables have zero margnal cost. However, f there s a cost for generaton, then that results n an addtonal proft durng curtalment, whch makes strategc curtalment more lkely. B. A Proft-Maxmzng Aggregator A natural model for a strategc aggregator s one that maxmzes curtalment proft LMPs and curtalment constrants. Snce LMPs are the soluton to an optmzaton problem themselves, the aggregator s problem s a blevel

4 4 optmzaton problem. In order to be able to express ths optmzaton n an explct form, let us frst wrte the KKT condtons of the program (1). Prmal feasblty: Dual feasblty: Complementary slackness: p Bf p + α + d p f f f Hf = λ, λ +, µ, µ + (4a) (4b) (4c) (4d) λ + ((Bf) p + α + d p ) =, = 1,..., n (4e) λ ( p (Bf) + p α d ) =, = 1,..., n (4f) Statonarty: µ + l (f l f l ) =, l = 1,..., t (4g) µ l (f l f l) =, l = 1,..., t (4h) B T (c + λ + λ ) + µ + µ + H T ν = (4) Here H R (t rank(g)) t, and the range of G s the nullspace of H. Usng the KKT condtons derved above, the aggregator s problem can be formulated as follows. γ = maxmze α,f,λ,λ +,µ,µ +,ν γ(α) (5a) α p a, N a (5b) α j =, j N a (5c) (4) (5d) The objectve (5a) s the curtalment proft defned n (3). Constrants (5b) and (5c) ndcate that the aggregator can only curtals generaton at ts own nodes, and the amount of curtalment cannot exceed the amount of avalable generaton to t. Constrant (5d), whch s the KKT condtons, enforces the locatonal margnal prcng adopted by the ISO. Note that f one assumes that curtalments of a hgher amount than a lmt τ can be detected by the ISO, then we can smply replace p (a) n (5b) by τ. An mportant note about ths problem s that we have assumed the aggregator has complete knowledge of the network topology (G), and state estmates (p and d). Ths s, perhaps, optmstc; however one would hope that the market desgn s such that aggregators do not have proftable manpulatons even wth such knowledge. The results n ths paper ndcate that ths s not the case. C. Connectons between Curtalment Proft and Market Power Whle our setup so far seems dvorced from the noton of market power, t turns out that there s a fundamental relatonshp between market power and the noton of curtalment proft ntroduced above. As mentoned earler, there has been sgnfcant work on market power n electrcty markets, but work s only begnnng to emerge on the market power of renewable generaton producers. One mportant work from ths lterature s [1], and the followng s the proposed noton of market power from that work. Defnton 3. For α aggregator s defned as ( λ (α ) λ () η = λ (), the market power (ablty) of the ) ( ) α / p a In ths defnton the value of η captures the ablty of the generator/aggregator to exercse market power. Intutvely, n a market wth hgh value of η, the aggregator can sgnfcantly ncrease the prce by curtalng a small amount of generaton. Interestngly, the optmal curtalment proft s closely related to ths noton of market power. We summarze the relatonshp n the followng proposton, whch s proven n Appendx C. Proposton 1. If the curtalment proft γ s postve then the market power η > 1. Furthermore, the larger the curtalment proft s, the hgher s the market power. Ths proposton hghlghts that the noton of market power n [1] s consstent wth an aggregator seekng to maxmze ther curtalment proft, and hgher curtalment proft corresponds to more market power. IV. OPTIMIZING CURTAILMENT PROFIT The aggregator s proft maxmzaton problem s challengng to analyze, as one would expect gven ts blevel form. In fact, blevel lnear programmng s NP-hard to approxmate up to any constant multplcatve factor n general [44]. Furthermore, the objectve of the program (5) s quadratc (blnear) n the varables, rather than lnear. Ths combnaton of dffcultes means that we cannot hope to provde a complete analytc characterzaton of the behavor of a proft maxmzng aggregator. In ths secton, we begn wth the case of a sngle-bus aggregator and buld to the case of general mult-bus aggregators n acyclc networks. For the sngle-bus aggregator, the optmal curtalment can be found exactly, n polynomal tme. For the general case, we cannot provde an exact algorthm, but we do provde a practcal approxmaton algorthm for general mult-bus aggregators n acyclc networks (e.g. dstrbuton networks). A. An Exact Algorthm for a Sngle-Node Aggregator Even n the smplest case, when the aggregator has only a sngle node,.e. ts entre generaton s located n a sngle bus, t s not trval how to solve the aggregator s proft maxmzaton problem. (6)

5 5 B. An Approxmaton Algorthm for Mult-Bus Aggregators n Radal Networks Fg. 1. The LMP at bus as a functon of curtaled generaton at that bus. Shaded areas ndcate the aggregator s revenue at the normal condton and at the aggreaton. The frst step toward solvng the problem s already dffcult. In partcular, n order to understand the effect of curtalment on the proft, we frst need to understand how does curtalment mpact the prces an mpact whch s not monotonc n general. However, although LMPs are not monotonc n general, t turns out that n sngle-bus curtalment, the LMP s ndeed monotonc wth respect to the curtalment. The proof of the followng lemma s n Appendx A. Lemma 2. The LMP of any bus s monotoncally ncreasng wth respect to the curtalment at that bus. That s λ (α ) λ (α) f α > α, and α j = α j for all j = [n]\{}. A consequence of the above lemma s that the prce λ s a monotoncally ncreasng starcase functon of α, for any bus, as depcted n Fg. 1. Further, the bndng constrants of (1) do not change n the ntervals, and thus the dual varables reman the same,.e., the LMPs reman constant. When a constrant becomes bndng/non-bndng, the LMP jumps to the next level. In Fg. 1, the blue and red shaded areas are proft at the normal condton and at the curtalment, respectvely. The dfference between the two areas the curtalment proft. In partcular, f the red area s larger than the blue one, the aggregator s able to earn a postve curtalment proft on bus. The optmal curtalment α also happens where the red area s maxmzed. It should be clear that the optmal curtalment always happens at the verge of a prce change, not n the mddle of a constant nterval (otherwse t can be ncreased by curtalng less). Gven the knowledge of the network and state estmates, t s possble to fnd the jump ponts (.e. where the bndng constrants change) and evaluate them for proftablty. Therefore, f there are not too many jumps, an exhaustve search over the jump ponts can yeld the optmal curtalment. Based on ths observaton, we have the followng theorem, whch s proven n Appendx B. Theorem 3. The exact optmal curtalment for an aggregator wth a sngle bus, n a network wth t lnes, can be found by an algorthm wth runnng tme O(t ). Clearly, ths approach does not extend to large mult-bus aggregators. The followng sectons use a dfferent, more sophstcated algorthmc approach for that settng. In ths secton, we show that the aggregator proft maxmzaton problem, whle hard n general, can be solved n an approxmate sense to determne an approxmately-feasble approxmately-optmal curtalment strategy n polynomal tme usng an approach based on dynamc programmng. In partcular, we show that an ɛ-approxmaton of the optmal curtalment proft can be obtaned usng an algorthm wth runnng tme that s lnear n the sze of the network and polynomal n 1 ɛ. Before we state the man result of ths secton, we ntroduce the noton of an approxmate soluton to (5) n the followng defnton. Defnton 4. A soluton (α, f, λ, λ +, µ, µ +, ν) to (5) s an ɛ-accurate soluton f the constrants are volated by at most ɛ and γ (α) γ ɛ. Note that, f one s smply nterested n approxmatng γ (as a market regulator would be), the ɛ-constrant volaton s of no consequence, and an ɛ-accurate soluton of (5) suffces to compute an ɛ-approxmaton to γ. Gven the above noton of approxmaton, our man theorem s the followng (proof s n Appendx D): Theorem 4. An ɛ-accurate soluton to the optmal aggregator curtalment problem (5) for an n-bus radal network can be found by an algorthm wth runnng tme cn ( ) 1 9 ɛ where c s a constant that depends on the parameters p a, B, d, p, f, f. On a lnear (feeder lne) network, the runnng tme reduces to cn ( ) 1 6. ɛ We now gve an nformal descrpton of the approxmaton algorthm. Consder a radal dstrbuton network wth nodes labeled [n], (where 1 denotes the substaton bus, where the radal network connects to the transmsson grd). Radal dstrbuton networks have a tree topology (they do not have cycles). We denote bus 1 as the root of the tree, and buses wth only one neghbor as leaves. Every node (except the root) has a unque parent, defned as the frst node on the unque path connectng t to the root node. The set of nodes k that have a gve node as ts parent are sad to be ts chldren. It can be shown that the strategc curtalment problem on any radal dstrbuton network can be expressed as an equvalent problem on a network where each node has maxmum degree 3 (known as a bnary tree, see Appendx D). Thus, we can lmt our attenton to networks of ths type, where every node has a unque parent and at most 2 chldren. For a node, let c 1 (), c 2 () denote ts chldren (where c 1 =, c 2 = s allowed snce a node can have fewer than two chldren). We use the shorthand p net () = f c1() + f c2() f (p α d ) (4a) reduces to p p net () p where f 1 = and f =. The matrx H n (4c) s an empty matrx (the nullspace of the matrx B s of dmenson ), so ths constrant can be

6 6 Fg. 2. The representaton of a bnary tree. For any node, and ts chldren denoted c 1 (), c 2 (). dropped. Usng ths addtonal structure, the problem (5) can be rewrtten (after some algebra) as: maxmze λ,f,α n λ (p a α ) =1 (7a) α p a, [n] (7b) p p net () p, [n] (7c) f f f, [n] \ {1} (7d) c, f p net () = p λ = c, f p < p net () < p, [n] (7e) c, f p net () = p, f f = f λ cj() λ =, f f < f < f, [n], j = 1, 2 (7f), f f = f where λ s the LMP at bus. Note that we assumed that there s some aggregator generaton and potental curtalment at every bus (however ths s not restrctve, snce we can smply set p a = at buses where the aggregator owns no assets). Defne x = (λ, f, α ). (7) s of the form: n max g (x ) for some functons g (.) x:h (x,x c1 (),x c2 ()), [n] =1 and h (.). Ths form s amenable to dynamc programmng, snce f we fx the value of x, the optmzaton problem for the subtree under s decoupled from the rest of the network. Set κ n (x) =, defne κ for < n recursvely as: 2 j=1 g c j() ( xcj() ) + κ (x) = max x c1 (),x c2 () h ( (x,x c1 (),x c2 ()) κ cj() xcj()) Then, the optmal value can be computed as = max x κ 1 (x) + g 1 (x). However, the above recurson requres an nfnte-dmensonal computaton at every step, snce the value of κ needs to be calculated for every value of x. To get around ths, we note that the varables λ, f, α are bounded, and hence x can be dscretzed to le n a certan set X such that every feasble x s at most δ(ɛ ) away (n nfnty-norm sense) from some pont n X (Lemma 5). The dscretzaton error can be quantfed, and ths error bound can be used to relax the constrant to h (x, x +1 ) ɛ guaranteeng that any soluton to (5) s feasble for the relaxed constrant. Ths allows us to defne a dynamc program (algorthm 1). γ Algorthm 1 Dynamc programmng on bnary tree S { : c 1 () =, c 2 () = } κ (x) x X, S whle S n do S { S : c 1 (), c 2 () S} S, x X : κ (x) max x 1 X c 1 (),x 2 X c 2 () h (x,x 1,x 2) ɛ S S S end whle γ max x X1 κ 1 (x) + g 1 (x) j=1,2 g cj() ( x j ) + κcj() (x ) The algorthm essentally starts at the leaves of the tree and proceeds towards the root, at each stage updatng κ for nodes whose chldren have already been updated (stoppng at root). Along wth the dscretzaton error analyss n Appendx D, ths essentally concludes Theorem 4. It s worth notng that prevous work on dstrbuton level markets have used AC power flow models (at least n some approxmate form) due to the mportance of voltage constrants and reactve power n a dstrbuton system [45]. Our approach extends n a straghtforward way to ths settng as well, as the dynamc programmng structure remans preserved (the KKT condtons wll smply be replaced by the correspondng condtons for the AC based market clearng mechansm). V. THE IMPACT OF STRATEGIC CURTAILMENT Usng the algorthms developed n the prevous sectons, we can now move to characterzng the potental mpact of strategc curtalment. We do ths by frst provdng an llustratve example of how curtalment leads to a larger proft for a smple sngle-bus aggregator n a small, 6-bus network. Then, we consder curtalment n larger networks and show the global effect of strategc behavor. We use IEEE 14-, 3-, and 57-bus test cases, and ther enhanced versons from NICTA Energy System Test case Archve [46]. Bus 4 35 $/MWh 2-3 MW LMP4=35 $/MWh Bus 1 2 $/MWh 2-4 MW LMP1=2 $/MWh 85 MW 25 MW Bus 5 23 $/MWh 2-25 MW LMP5=28.7 $/MWh 375 MW 1/1 MW -65.2/9 MW 3 MW Generator 15 MW 25.2/1 MW 25 MW 1/1 MW -9/9 MW 73 MW 3 MW -31/9 MW Bus 2 25 $/MWh 2-2 MW LMP2=25 $/MWh 397 MW Load -96/1 MW -76.4/9 MW Bus 6 Bus 3 22 $/MWh 2-3 MW LMP3=25 $/MWh 3 MW 28 MW 24 $/MWh 2-4 MW LMP6=24 $/MWh s 2$/MWh, generaton output range s 2 4 MW. In the gven operatng condton, the actual load at bus 1 s 15 MW and the actual generaton output s 375 MW. As a result, 1 MW power flows from bus 1 to 2, reachng the lne capacty 1 MW. A negatve power flow value means the power flows n the opposte drecton of the defned lne orentaton, e.g. 96 MW power flows from bus 3 to 2 n lne l 23, whose capacty s 1 MW. code for the add-then-remove algorthm s Algorthm 1, where the lnes 2 to 5 correspond to addng loosenng constrants whle lnes 6 to 14 for removng tghtenng constrants. C. A case study and nsghts To brng out ntutons, we mplement the proposed algorthm n an example 6 bus system n Fg. 1, where G 1 s B Globecom Communcaton and Informaton System Securty Symp Fg. 3. The 6-bus example network from [27], used to llustrate the effect of curtalment. A. An Illustratve Example We use a 6-bus example network from [27], n whch the amounts of generaton are 375.2, 73., 299.6, 84.8, C hypothetcally 2 reducng t n the ex-post LMP problem LMP at the tagged bus s re IV. ATTACK I In ths secton, we realze usng LR attack, where df pendng on the attackers s c estngly, we also fnd that has mpact to future electrc A. LR attack to real-tme L Suppose that a false data data a =(a 1,a 2,..,a m ) n Then, the receved measurem z = H From (3), the ML estmate and the estmate of lne flo wth the actual state estmat x ˆx = Pa. Meanwhle, th r = z H x =(I Load redstrbuton (LR) The attack vector of a LR a a p = and e T a d =al generaton and demand see attacks do not compromse because generaton measure protected and can be verfe between ISOs and utlty c and load measurements are w

7 7 LMP ($/MWh) Normal Curtalment Bus No. Fg. 4. The locatonal margnal prces for the 6-bus example before and after the curtalment. 25., MW (See Fg. 3). The loads and the orgnal offer prces for the generators can be found n the fgure. In ths case, the lnes l 12, l 14 and l 56 are carryng ther maxmum flow, and the real-tme LMPs are 2., 25., 25., 35., 28.7, 24. $/M W h, respectvely. Assume that the aggregator owns node 1 and ams to ncrease ts proft by curtalng the generaton at ths node. It can be seen that by decreasng the generaton at node 1 by.15, from MW to MW, the bndng and non-bndng constrants n problem (1) change, and as a result the ISO determnes the new LMPs as 25.8, 25., 25., 35., 3.6, 24. $/M W h. Fg. 4 shows the LMPs, before and after the curtalment. The curtalment proft s γ = = 2172 $ / h,whch means that the aggregator has been able to ncrease ts proft by 2172 $/h. For dspatch ntervals of length 15 mnutes, ths amount s around $543 per dspatch nterval. B. Case Studes %age error n optmal value Tme (s) Fg. 5. The dfference from the optmal soluton as a functon of the runnng tme of the algorthm, on a 9-bus network wth 1% curtalment allowance. We smulate the behavor of aggregators wth dfferent szes,.e. dfferent number of buses, n a number of dfferent networks. We use the IEEE 14-, 3-, and 57-bus test cases. Snce studyng market manpulaton makes sense only when there s congeston n the network, we scale the demand (or equvalently the lne flows) untl there s some congeston n the network. In order to examne the proft and market power of aggregator as a functon of ts sze, we assume that the way aggregator grows s by sequentally addng random buses to ts set (more or less lke the way e.g. a solar frm grows). Then at any fxed set of buses, t can choose dfferent curtalment strateges to maxmze ts proft. In other words, for each of ts nodes t should decde whether to curtal or not (assumng that the amount of curtalment has been fxed to a small porton). We assume that the total generaton of the aggregator n each bus s 1 MW and t s able to curtal 1% of t (.1 MW ). For each of the three networks, Fg. 6 shows the proft for a random sequence of nodes. Comparng the no-curtalment proft wth the strategc-curtalment proft, reveals an nterestng phenomenon. As the sze of the aggregator (number of ts buses) grows, not only does the proft ncrease (whch s expected), but also the dfference between the two curves ncrease, whch s the curtalment proft. More specfcally, the latter does not need to happen n theory. However n practce, t s observed most of the tme, and t ponts out that larger aggregators have hgher ncentve to behave strategcally, and they can ndeed gan more from curtalment. Fnally, to demonstrate the performance of our approxmaton algorthm on acyclc networks, we plot the suboptmalty gap versus runnng tme of the algorthm (Fg. 5) on a radal verson of the IEEE 9-bus network (taken from [47] ). It can be seen that wth the optmal soluton can be computed wth error n 1s. VI. CONCLUDING REMARKS Understandng the potental for market manpulaton by aggregators s crucal for electrcty market effcency n the new era of renewable energy. In ths paper, we characterzed the proft an aggregator can make by strategcally curtalng generaton n the ex-post market as the outcome of a blevel optmzaton problem. Ths model captures the realstc prce clearng mechansm n the electrcty market. Wth ths formulaton when the aggregator s located n a sngle bus, we have shown that the locatonal margnal prce s monotoncally ncreasng wth the curtalment; the proft drectly correlates wth market power defned n the lterature, and we have an exact polynomal-tme algorthm to solve the aggregators proft maxmzaton problem. The aggregator s strategc curtalment problem n a general settng s a dffcult b-level optmzaton problem, whch s ntractable. However, we show that for radal dstrbuton networks (where aggregators are lkely located) there s an effcent algorthm to approxmate the soluton up to arbtrary precson. Fnally, we show va smulatons on realstc test cases that, frst, there s potentally large proft for aggregators by manpulatng the LMPs n the electrcty market, and second, our algorthm can effcently fnd the (approxmately) optmal curtalment strategy. We vew ths paper as a frst step n understandng market power of aggregators, and more generally, towards market desgn for ntegratng renewable energy and demand response from geographcally dstrbuted sources. Wth the result of our paper, t s nterestng to ask what can the operator do to address ths problem. In partcular, how to desgn market rules for aggregators to maxmze the contrbuton of renewable energy yet mtgate the exercse of market power. Also, extendng the analyss to the case of multple aggregators n the market s another nterestng drecton for future research. REFERENCES [1] K. Bulls, Why SolarCty s succeedng n a dffcult solar ndustry, 212.

8 x Strategc Curtalment Normal 2 Strategc Curtalment Normal 2.5 Strategc Curtalment Normal 6 2 Proft ($/h) 5 4 Proft ($/h) 15 1 Proft ($/h) Aggregator Sze (MW) Aggregator Sze (MW) Aggregator Sze (MW) Fg. 6. The proft under the normal (no-curtalment) condton and under (optmal) strategc curtalment, as a functon of sze of the aggregator n IEEE test case networks: a) IEEE 14-Bus Case, b) IEEE 3-Bus Case, and c) IEEE 57-Bus Case. The dfference between the blue and red curves, whch s the curtalment proft, s non-decreasng. [2] J. John, SolarCty and Tesla: a utltys worst nghtmare? Energy-Storage-Fleet, 214. [3] E. Wesoff, Earnngs from SunPower, Tesla, Enphase, plus new fundng for Sunnova, SolarCty, 1366, Sva, Nexeon, Enphase-and-New-Fundng-for-Sunnova-SolarC, 216. [4] Solar Energy Industres Assocaton (SEIA), state-solar-polcy, accessed: [5] 215 top 5 north amercan solar contractors, solarpowerworldonlne.com. [6] N. Ltvak, U.S. commercal solar landscape , greentechmeda.com/research/report/us-commercal-solar-landscape , Report. [7] C. Honeyman, U.S. resdental solar economc outlook , us-resdental-solar-economc-outlook , Report. [8] [9] [1] Y. Yu, B. Zhang, and R. Rajagopal, Do wnd power producers have market power and exercse t? n PES General Meetng Conference & Exposton, 214 IEEE. IEEE, 214, pp [11] P. Twomey, R. Green, K. Neuhoff, and D. Newbery, A revew of the montorng of market power: The possble roles of TSOs n montorng for market power ssues n congested transmsson systems, 215, Tech. Report. [12] J. Bushnell, C. Day, M. Duckworth, R. Green, A. Halseth, E. G. Read, J. S. Rogers, A. Rudkevch, T. Scott, Y. Smeers et al., An nternatonal comparson of models for measurng market power n electrcty, n Energy Modelng Forum Stanford Unversty, [13] A. Sheffrn and J. Chen, Predctng market power n wholesale electrcty markets, n Proc. of the Western Conference of the Advances n Regulaton and Competton, South Lake Tahoe, 22. [14] Calforna Independent System Operator, Market power and compettveness, [15] Electrc Relablty Councl of Texas, ERCOT protocols, [16] R. Schmalensee and B. W. Golub, Estmatng effectve concentraton n deregulated wholesale electrcty markets, The RAND Journal of Economcs, pp , [17] D. T. Scheffman and P. T. Spller, Geographc market defnton under the us department of justce merger gudelnes, Journal of Law and Economcs, pp , [18] S. Borensten, J. Bushnell, E. Kahn, and S. Stoft, Market power n calforna electrcty markets, Utltes Polcy, vol. 5, no. 3, pp , [19] S. S. Oren, Economc neffcency of passve transmsson rghts n congested electrcty systems wth compettve generaton, The Energy Journal, pp , [2] J. B. Cardell, C. C. Htt, and W. W. Hogan, Market power and strategc nteracton n electrcty networks, Resource and energy economcs, vol. 19, no. 1, pp , [21] L. Xu and R. Baldck, Transmsson-constraned resdual demand dervatve n electrcty markets, Power Systems, IEEE Transactons on, vol. 22, no. 4, pp , 27. [22] L. Xu and Y. Yu, Transmsson constraned lnear supply functon equlbrum n power markets: method and example, n Power System Technology, 22. Proceedngs. PowerCon 22. Internatonal Conference on, vol. 3. IEEE, 22, pp [23] S. Bose, C. Wu, Y. Xu, A. Werman, and H. Mohsenan-Rad, A unfyng market power measure for deregulated transmsson-constraned electrcty markets, Power Systems, IEEE Transactons on, 215. [24] P. Twomey and K. Neuhoff, Wnd power and market power n compettve markets, Energy Polcy, vol. 38, no. 7, pp , 21. [25] A. L. Ott, Experence wth PJM market operaton, system desgn, and mplementaton, Power Systems, IEEE Transactons on, vol. 18, no. 2, pp , 23. [26] T. Zheng and E. Ltvnov, Ex post prcng n the co-optmzed energy and reserve market, IEEE Trans. on Power Sys., vol. 21, no. 4, pp , 26. [27] S. B and Y. J. Zhang, False-data njecton attack to control real-tme prce n electrcty market, n Global Communcatons Conference (GLOBECOM), 213 IEEE. IEEE, 213, pp [28], Usng covert topologcal nformaton for defense aganst malcous attacks on dc state estmaton, Selected Areas n Communcatons, IEEE Journal on, vol. 32, no. 7, pp , July 214. [29] L. Xe, Y. Mo, and B. Snopol, False data njecton attacks n electrcty markets, n Smart Grd Communcatons (SmartGrdComm), 21 Frst IEEE Internatonal Conference on. IEEE, 21, pp [3], Integrty data attacks n power market operatons, Smart Grd, IEEE Transactons on, vol. 2, no. 4, pp , 211. [31] Y. Lu, P. Nng, and M. K. Reter, False data njecton attacks aganst state estmaton n electrc power grds, ACM Transactons on Informaton and System Securty (TISSEC), vol. 14, no. 1, p. 13, 211. [32] Y. Ln, P. Barooah, S. Meyn, and T. Mddelkoop, Expermental evaluaton of frequency regulaton from commercal buldng hvac systems, Smart Grd, IEEE Transactons on, vol. 6, no. 2, pp , 215. [33] H. Hao, A. Kowl, Y. Ln, P. Barooah, and S. Meyn, Ancllary servce for the grd va control of commercal buldng hvac systems, n Amercan Control Conference (ACC), 213. IEEE, 213, pp [34] B. Zhang, A. Lam, A. D. Domínguez-García, and D. Tse, An optmal and dstrbuted method for voltage regulaton n power dstrbuton systems, Power Systems, IEEE Transactons on, vol. 3, no. 4, pp , 215. [35] D. B. Arnold, M. Negrete-Pncetc, M. D. Sankur, D. M. Auslander, and D. S. Callaway, Model-free optmal control of var resources n dstrbuton systems: An extremum seekng approach, IEEE Transactons on Power Systems, vol. PP, no. 99, pp. 1 11, 215. [36] M. Negrete-Pncetc and S. Meyn, Markets for dfferentated electrc power products n a smart grd envronment, n Power and Energy Socety General Meetng, 212 IEEE. IEEE, 212, pp [37] Y. Chen, A. Busc, and S. Meyn, Indvdual rsk n mean feld control wth applcaton to automated demand response, n Decson and Control (CDC), 214 IEEE 53rd Annual Conference on. IEEE, 214, pp [38] M. C. Ferrs, S. P. Drkse, and A. Meeraus, Mathematcal programs wth equlbrum constrants: Automatc reformulaton and soluton va constraned optmzaton, 22. [39] O. Ben-Ayed and C. E. Blar, Computatonal dffcultes of blevel lnear programmng, Operatons Research, vol. 38, no. 3, pp , 199. [4] Z. H. Gümüş and C. A. Floudas, Global optmzaton of nonlnear blevel programmng problems, Journal of Global Optmzaton, vol. 2, no. 1, pp. 1 31, 21. [41] S. P. Drkse and M. C. Ferrs, The path solver: a nommonotone stablzaton scheme for mxed complementarty problems, Optmzaton Methods and Software, vol. 5, no. 2, pp , [42] Shft factor: methodology and example, pdf, 24. [43] D. B. Patton, P. LeeVanSchack, and J. Chen, 213 assessment of the electrcty markets n new england, Potomac Economcs, 214. [44] S. Dempe, V. Kalashnkov, G. A. Pérez-Valdés, and N. Kalashnkova, Blevel Programmng Problems: Theory, Algorthms and Applcatons to Energy Networks. Sprnger, 215. [45] E. Ntakou and M. Caramans, Dstrbuton network electrcty market clearng: Parallelzed pmp algorthms wth mnmal coordnaton, n Decson and Control (CDC), 214 IEEE 53rd Annual Conference on. IEEE, 214, pp [46] C. Coffrn, D. Gordon, and P. Scott, NESTA, the NICTA energy system test case archve, arxv: , 214. [47] B. Kocuk, S. S. Dey, and X. A. Sun, Inexactness of sdp relaxaton and vald nequaltes for optmal power flow, IEEE Transactons on Power Systems, vol. 31, no. 1, pp , Jan 216.

9 9 A. Proof of Lemma 2 APPENDIX Let us take a look at the ISO s optmzaton problem (1), whch s a lnear program. It s not hard to see that the dual of ths problem s as follows. maxmze λ,λ +,µ,µ +,ν ( p + p α d) T λ + ( p + α + d p) T λ + + f T µ f T µ + B T (c + λ + λ ) µ + µ + + H T ν = λ, λ +, µ, µ + (8a) (8b) (8c) If one focuses on the terms nvolvng α for a certan, the objectve of the above optmzaton problem s n the form: ( p +p α d )λ +( p +α +d p )λ + plus a lnear functon of the rest of the varables (.e. the rest of λ, λ +, as well as µ, µ +, ν). There s no α n the constrants, and the frst two terms of ths objectve are the only parts where α appears (and wth opposte sgns). We need to show that f α s changed to α + δ for some δ >, then c + λ +new λ new c + λ + λ, where λ +new, λ new are the optmal solutons of the new problem. We prove ths n a general settng. Consder the followng two optmzaton problems. f = sup x 1,x 2 R x 3 R m f new = sup x 1,x 2 R x 3 R m T a 1 x 1 + a 2 x 2 + a3 x 3 (9a) s.t. (x 1, x 2, x 3 ) S (9b) T (a 1 δ)x 1 + (a 2 + δ)x 2 + a3 x 3 (1a) s.t. (x 1, x 2, x 3 ) S (1b) Assume that the optmal values of the problems are attaned at (x 1, x 2, x 3) and (x new 1, x new 2, x new 3 ), respectvely. We clam that x new 2 x new 1 x 2 x 1 (Ths precsely mples the LMP condton n our case,.e. λ +new λ new λ + λ ). Suppose by way of contradcton that x new 2 x new 1 < x 2 x 1. We know that a 1 x 1 + a 2 x 2 + a T 3 x 3 a 1 x 1 + a 2 x 2 + a T 3 x 3, (x 1, x 2, x 3 ) S. Therefore we have (a 1 δ)x new 1 + (a 2 + δ)x new 2 + a T 3 x new 3 = a 1 x new 1 + a 2 x new 2 + a T 3 x new 3 δx new 1 + δx new 2 a 1 x 1 + a 2 x 2 + a T 3 x 3 + δ(x new 2 x new 1 ) < a 1 x 1 + a 2 x 2 + a T 3 x 3 + δ(x 2 x 1) = (a 1 δ)x 1 + (a 2 + δ)x 2 + a T 3 x 3. The frst nequalty above follows from the fact that (x new 1, x new 2, x new 3 ) S. Now the above mples that (x new 1, x new 2, x new 3 ) s not the optmal soluton of (1), and t s a contradcton. As a result, x new 2 x new 1 x 2 x 1. B. Proof of Theorem 3 Snce we are n the sngle-bus curtalment regme, α has only one nonzero component. For the sake of convenence, we denote that element tself by a scalar α throughout ths proof (no α s vector n ths proof). The proof conssts of the followng two peces: 1) From each jump pont, the pont where the next jump happens can be computed n polynomal tme, 2) There are at most polynomally (n ths case even lnearly) many jumps. Assumng that the soluton to the program (1) s unque, for any fxed value of α, exactly t of the constrants (1b,1c,1d) are bndng (actve). We can express these bndng constrants as Af = b(α), where A R t t s an nvertble matrx, and b(α) R t s a vector that depends on α. As long as the bndng constrants do not change, the matrx A s fxed and the optmal soluton s lnear n α (.e. f = A 1 b(α)). Then, for smplcty, we can express the soluton as f(α) = f + αa, for some t-vectors f and a. Now f we look at the non-bndng (nactve) constrants of (1), they can also be expressed as Ãf < b, for some matrx à and vector b of approprate dmensons. Insertng f nto ths set of nequaltes yelds Ãf + αãa < b, or equvalently α(ãa) < b (Ãf ), for all = 1, 2,..., (2n + 2t rank(g)). Now we need to fgure out that, wth ncreasng α, whch of the non-bndng constrants becomes bndng frst and wth exactly how much ncrease n α. If for some we have (Ãa), then t s clear that ncreasng α cannot make constrant bndng. If (Ãa) > then the constrant can be wrtten as α < b (Ãf ) (Ãa). Computng the rght-hand sde for all, and takng ther mnmum, tells us exactly whch constrant wll become bndng next and how much change n the current value of α results n that. The complexty of ths procedure s O(t ) for computng f and a, plus O(t(2n + 2t)) = O(nt + t 2 ) for computng the lowest bound among all the constrants. Hence the complexty s O(t ). The above procedure descrbes how the next jump pont can be computed effcently from the current pont. The exact same procedure can be repeated for reachng the subsequent jump ponts. All remans to show s the second pece of the proof, whch s that the number of jump ponts are bounded polynomally. To show the last part note that by ncreasng α, f a bndng constrant becomes non-bndng, t wll not become bndng agan. As a result, each constrant can change at most twce, and therefore the number of jumps s at most twce the number of constrants. Thus, the number of jumps

10 1 s O(n + t), and the overall complexty of the algorthm s O((n + t)t ) = O(t ). C. Proof of Proposton 1 From the defnton of γ(α ) = λ (α )(p a α ) λ ()p a t follows that γ(α ) λ ()(p a α ) = λ (α ) λ () pa p a α = 1 + λ (α ) λ () (1 α λ () p a ) λ (α ) λ () λ () = λ (α ) λ () λ () (1 + α p a ) α p a. (11) Therefore we have p a ( λ ()(p a γ(α λ (α ) ( ) ) λ () α ) = / α )α λ () p a 1 = η 1. Snce the left-hand sde parameters are all postve, f γ(α ) >, we can conclude that η > 1. Moreover, t s clear that the larger the value of γ(α ) s, the hgher the value of η s. Note that we used the approxmaton (1 α p ) 1 1+ α a p, snce the a curtalment s small wth respect to the generaton; however, the rght-hand sde expresson (11) s an upper bound on the left-hand sde anyway, and the result holds exactly. flow from ts parent s just splt n some way between ts chldren. Ths n fact enforces the flow conservaton constrant at b. Smlar constructon can be appled to any node of the tree wth more than two chldren, untl no such node exsts. It can be seen that the number of nodes n the new graph s stll lnear n n. So any tree can be transformed to a bnary one by the above procedure. For the rest of the analyss we focus on the ɛ- approxmaton of the dynamc program on the resultng bnary tree. The optmzaton problem (5) on a bnary tree, can be wrtten after some algebra as the followng. max λ,f,α n λ (p α ) =1 (12a) α p a, = 1,..., n (12b) p f c1() + f c2() f p + α + d p, = 1,..., n (12c) f f f, = 2,..., n (12d) { (λ c )(f c1() + f c2() f p + α + d p ) (λ c )(f c1() + f c2() f p + α + d p ) = 1,..., n (12e), D. Proof of Theorem 4 Lemma 5 (δ-dscretzaton). Gven a set C [L 1, L 1 ] [L k, L k ], there exsts a fnte set X such that z C z X, max 1 k z z δ and X contans at most V/δ k ponts, where V = k =1 (L L ) s a constant (the volume of the box). X s sad to be an δ-dscretzaton of C and wrtten as X (δ). Lemma 6 (Reducton to Bnary Tree). Any tree wth arbtrary degrees can be reduced to a bnary tree by ntroducng addtonal dummy nodes to the network. Proof. Take any node b n the tree wth some parent a and k > 2 chldren c 1,..., c k. There exsts m > such that 2 m < k 2 m+1 for some m. We wll show that ths subgraph can be made a bnary tree by ntroducng O(k) dummy nodes (n m levels) between b and ts chldren. The addtonal nodes and edges are defned as follows: up to m levels: b b, b b 1, b b, b b 1, b 1 b 1, b 1 b 11, b b, b b 1,..., b 11 b 111, b... c 1, b... c 2, b...1 c 3... Ths s transparently a bnary tree wth O(k) nodes. Each of the new nodes has zero njecton, and effectvely the ncomng { (λ λ cj())(f cj() f c j()) (λ λ cj())(f cj() f cj()), = 1,..., n, j = 1, 2 (12f) The constrants α p a and f f f, along wth a pror bound on lambda λ λ λ can be used to defne the box where x = (λ, f, α ) lves. Then an ɛ-accurate soluton s a soluton to the followng problem. n λ (p α ) (13a) max λ,f,α =1 p ɛ f c1() + f c2() f p + α + d p + ɛ, = 1,..., n (13b) { (λ c )(f c1() + f c2() f p + α + d p ) ɛ (λ c )(f c1() + f c2() f p + α + d p ) ɛ { (λ λ cj())(f cj() f c j()) ɛ (λ λ cj())(f cj() f cj()) ɛ = 1,..., n, (13c) = 1,..., n, j = 1, 2 (13d) Assumng a δ-dscretzaton of the constrant set, each of the constrants (as well as ɛ-accuracy of the objectve) mposes,

Institute of Informatics, Faculty of Business and Management, Brno University of Technology,Czech Republic

Institute of Informatics, Faculty of Business and Management, Brno University of Technology,Czech Republic Lagrange Multplers as Quanttatve Indcators n Economcs Ivan Mezník Insttute of Informatcs, Faculty of Busness and Management, Brno Unversty of TechnologCzech Republc Abstract The quanttatve role of Lagrange

More information

Support Vector Machines

Support Vector Machines Support Vector Machnes Max Wellng Department of Computer Scence Unversty of Toronto 10 Kng s College Road Toronto, M5S 3G5 Canada wellng@cs.toronto.edu Abstract Ths s a note to explan support vector machnes.

More information

UTILIZING MATPOWER IN OPTIMAL POWER FLOW

UTILIZING MATPOWER IN OPTIMAL POWER FLOW UTILIZING MATPOWER IN OPTIMAL POWER FLOW Tarje Krstansen Department of Electrcal Power Engneerng Norwegan Unversty of Scence and Technology Trondhem, Norway Tarje.Krstansen@elkraft.ntnu.no Abstract Ths

More information

J. Parallel Distrib. Comput.

J. Parallel Distrib. Comput. J. Parallel Dstrb. Comput. 71 (2011) 62 76 Contents lsts avalable at ScenceDrect J. Parallel Dstrb. Comput. journal homepage: www.elsever.com/locate/jpdc Optmzng server placement n dstrbuted systems n

More information

Feature selection for intrusion detection. Slobodan Petrović NISlab, Gjøvik University College

Feature selection for intrusion detection. Slobodan Petrović NISlab, Gjøvik University College Feature selecton for ntruson detecton Slobodan Petrovć NISlab, Gjøvk Unversty College Contents The feature selecton problem Intruson detecton Traffc features relevant for IDS The CFS measure The mrmr measure

More information

1 Example 1: Axis-aligned rectangles

1 Example 1: Axis-aligned rectangles COS 511: Theoretcal Machne Learnng Lecturer: Rob Schapre Lecture # 6 Scrbe: Aaron Schld February 21, 2013 Last class, we dscussed an analogue for Occam s Razor for nfnte hypothess spaces that, n conjuncton

More information

Project Networks With Mixed-Time Constraints

Project Networks With Mixed-Time Constraints Project Networs Wth Mxed-Tme Constrants L Caccetta and B Wattananon Western Australan Centre of Excellence n Industral Optmsaton (WACEIO) Curtn Unversty of Technology GPO Box U1987 Perth Western Australa

More information

The Greedy Method. Introduction. 0/1 Knapsack Problem

The Greedy Method. Introduction. 0/1 Knapsack Problem The Greedy Method Introducton We have completed data structures. We now are gong to look at algorthm desgn methods. Often we are lookng at optmzaton problems whose performance s exponental. For an optmzaton

More information

BERNSTEIN POLYNOMIALS

BERNSTEIN POLYNOMIALS On-Lne Geometrc Modelng Notes BERNSTEIN POLYNOMIALS Kenneth I. Joy Vsualzaton and Graphcs Research Group Department of Computer Scence Unversty of Calforna, Davs Overvew Polynomals are ncredbly useful

More information

On the Optimal Control of a Cascade of Hydro-Electric Power Stations

On the Optimal Control of a Cascade of Hydro-Electric Power Stations On the Optmal Control of a Cascade of Hydro-Electrc Power Statons M.C.M. Guedes a, A.F. Rbero a, G.V. Smrnov b and S. Vlela c a Department of Mathematcs, School of Scences, Unversty of Porto, Portugal;

More information

8.5 UNITARY AND HERMITIAN MATRICES. The conjugate transpose of a complex matrix A, denoted by A*, is given by

8.5 UNITARY AND HERMITIAN MATRICES. The conjugate transpose of a complex matrix A, denoted by A*, is given by 6 CHAPTER 8 COMPLEX VECTOR SPACES 5. Fnd the kernel of the lnear transformaton gven n Exercse 5. In Exercses 55 and 56, fnd the mage of v, for the ndcated composton, where and are gven by the followng

More information

Recurrence. 1 Definitions and main statements

Recurrence. 1 Definitions and main statements Recurrence 1 Defntons and man statements Let X n, n = 0, 1, 2,... be a MC wth the state space S = (1, 2,...), transton probabltes p j = P {X n+1 = j X n = }, and the transton matrx P = (p j ),j S def.

More information

Luby s Alg. for Maximal Independent Sets using Pairwise Independence

Luby s Alg. for Maximal Independent Sets using Pairwise Independence Lecture Notes for Randomzed Algorthms Luby s Alg. for Maxmal Independent Sets usng Parwse Independence Last Updated by Erc Vgoda on February, 006 8. Maxmal Independent Sets For a graph G = (V, E), an ndependent

More information

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ).

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ). REVIEW OF RISK MANAGEMENT CONCEPTS LOSS DISTRIBUTIONS AND INSURANCE Loss and nsurance: When someone s subject to the rsk of ncurrng a fnancal loss, the loss s generally modeled usng a random varable or

More information

Abteilung für Stadt- und Regionalentwicklung Department of Urban and Regional Development

Abteilung für Stadt- und Regionalentwicklung Department of Urban and Regional Development Abtelung für Stadt- und Regonalentwcklung Department of Urban and Regonal Development Gunther Maer, Alexander Kaufmann The Development of Computer Networks Frst Results from a Mcroeconomc Model SRE-Dscusson

More information

Answer: A). There is a flatter IS curve in the high MPC economy. Original LM LM after increase in M. IS curve for low MPC economy

Answer: A). There is a flatter IS curve in the high MPC economy. Original LM LM after increase in M. IS curve for low MPC economy 4.02 Quz Solutons Fall 2004 Multple-Choce Questons (30/00 ponts) Please, crcle the correct answer for each of the followng 0 multple-choce questons. For each queston, only one of the answers s correct.

More information

Power-of-Two Policies for Single- Warehouse Multi-Retailer Inventory Systems with Order Frequency Discounts

Power-of-Two Policies for Single- Warehouse Multi-Retailer Inventory Systems with Order Frequency Discounts Power-of-wo Polces for Sngle- Warehouse Mult-Retaler Inventory Systems wth Order Frequency Dscounts José A. Ventura Pennsylvana State Unversty (USA) Yale. Herer echnon Israel Insttute of echnology (Israel)

More information

Module 2 LOSSLESS IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 2 LOSSLESS IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module LOSSLESS IMAGE COMPRESSION SYSTEMS Lesson 3 Lossless Compresson: Huffman Codng Instructonal Objectves At the end of ths lesson, the students should be able to:. Defne and measure source entropy..

More information

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis The Development of Web Log Mnng Based on Improve-K-Means Clusterng Analyss TngZhong Wang * College of Informaton Technology, Luoyang Normal Unversty, Luoyang, 471022, Chna wangtngzhong2@sna.cn Abstract.

More information

Chapter 7: Answers to Questions and Problems

Chapter 7: Answers to Questions and Problems 19. Based on the nformaton contaned n Table 7-3 of the text, the food and apparel ndustres are most compettve and therefore probably represent the best match for the expertse of these managers. Chapter

More information

Can Auto Liability Insurance Purchases Signal Risk Attitude?

Can Auto Liability Insurance Purchases Signal Risk Attitude? Internatonal Journal of Busness and Economcs, 2011, Vol. 10, No. 2, 159-164 Can Auto Lablty Insurance Purchases Sgnal Rsk Atttude? Chu-Shu L Department of Internatonal Busness, Asa Unversty, Tawan Sheng-Chang

More information

Fault tolerance in cloud technologies presented as a service

Fault tolerance in cloud technologies presented as a service Internatonal Scentfc Conference Computer Scence 2015 Pavel Dzhunev, PhD student Fault tolerance n cloud technologes presented as a servce INTRODUCTION Improvements n technques for vrtualzaton and performance

More information

When Network Effect Meets Congestion Effect: Leveraging Social Services for Wireless Services

When Network Effect Meets Congestion Effect: Leveraging Social Services for Wireless Services When Network Effect Meets Congeston Effect: Leveragng Socal Servces for Wreless Servces aowen Gong School of Electrcal, Computer and Energy Engeerng Arzona State Unversty Tempe, AZ 8587, USA xgong9@asuedu

More information

How To Improve Power Demand Response Of A Data Center Wth A Real Time Power Demand Control Program

How To Improve Power Demand Response Of A Data Center Wth A Real Time Power Demand Control Program Demand Response of Data Centers: A Real-tme Prcng Game between Utltes n Smart Grd Nguyen H. Tran, Shaole Ren, Zhu Han, Sung Man Jang, Seung Il Moon and Choong Seon Hong Department of Computer Engneerng,

More information

SPEE Recommended Evaluation Practice #6 Definition of Decline Curve Parameters Background:

SPEE Recommended Evaluation Practice #6 Definition of Decline Curve Parameters Background: SPEE Recommended Evaluaton Practce #6 efnton of eclne Curve Parameters Background: The producton hstores of ol and gas wells can be analyzed to estmate reserves and future ol and gas producton rates and

More information

Efficient Project Portfolio as a tool for Enterprise Risk Management

Efficient Project Portfolio as a tool for Enterprise Risk Management Effcent Proect Portfolo as a tool for Enterprse Rsk Management Valentn O. Nkonov Ural State Techncal Unversty Growth Traectory Consultng Company January 5, 27 Effcent Proect Portfolo as a tool for Enterprse

More information

Multiple-Period Attribution: Residuals and Compounding

Multiple-Period Attribution: Residuals and Compounding Multple-Perod Attrbuton: Resduals and Compoundng Our revewer gave these authors full marks for dealng wth an ssue that performance measurers and vendors often regard as propretary nformaton. In 1994, Dens

More information

Pricing Data Center Demand Response

Pricing Data Center Demand Response Prcng Data Center Demand Response Zhenhua Lu, Irs Lu, Steven Low, Adam Werman Calforna Insttute of Technology Pasadena, CA, USA {zlu2,lu,slow,adamw}@caltech.edu ABSTRACT Demand response s crucal for the

More information

Feasibility of Using Discriminate Pricing Schemes for Energy Trading in Smart Grid

Feasibility of Using Discriminate Pricing Schemes for Energy Trading in Smart Grid Feasblty of Usng Dscrmnate Prcng Schemes for Energy Tradng n Smart Grd Wayes Tushar, Chau Yuen, Bo Cha, Davd B. Smth, and H. Vncent Poor Sngapore Unversty of Technology and Desgn, Sngapore 138682. Emal:

More information

THE DISTRIBUTION OF LOAN PORTFOLIO VALUE * Oldrich Alfons Vasicek

THE DISTRIBUTION OF LOAN PORTFOLIO VALUE * Oldrich Alfons Vasicek HE DISRIBUION OF LOAN PORFOLIO VALUE * Oldrch Alfons Vascek he amount of captal necessary to support a portfolo of debt securtes depends on the probablty dstrbuton of the portfolo loss. Consder a portfolo

More information

An Interest-Oriented Network Evolution Mechanism for Online Communities

An Interest-Oriented Network Evolution Mechanism for Online Communities An Interest-Orented Network Evoluton Mechansm for Onlne Communtes Cahong Sun and Xaopng Yang School of Informaton, Renmn Unversty of Chna, Bejng 100872, P.R. Chna {chsun,yang}@ruc.edu.cn Abstract. Onlne

More information

Forecasting the Direction and Strength of Stock Market Movement

Forecasting the Direction and Strength of Stock Market Movement Forecastng the Drecton and Strength of Stock Market Movement Jngwe Chen Mng Chen Nan Ye cjngwe@stanford.edu mchen5@stanford.edu nanye@stanford.edu Abstract - Stock market s one of the most complcated systems

More information

The OC Curve of Attribute Acceptance Plans

The OC Curve of Attribute Acceptance Plans The OC Curve of Attrbute Acceptance Plans The Operatng Characterstc (OC) curve descrbes the probablty of acceptng a lot as a functon of the lot s qualty. Fgure 1 shows a typcal OC Curve. 10 8 6 4 1 3 4

More information

Production. 2. Y is closed A set is closed if it contains its boundary. We need this for the solution existence in the profit maximization problem.

Production. 2. Y is closed A set is closed if it contains its boundary. We need this for the solution existence in the profit maximization problem. Producer Theory Producton ASSUMPTION 2.1 Propertes of the Producton Set The producton set Y satsfes the followng propertes 1. Y s non-empty If Y s empty, we have nothng to talk about 2. Y s closed A set

More information

A hybrid global optimization algorithm based on parallel chaos optimization and outlook algorithm

A hybrid global optimization algorithm based on parallel chaos optimization and outlook algorithm Avalable onlne www.ocpr.com Journal of Chemcal and Pharmaceutcal Research, 2014, 6(7):1884-1889 Research Artcle ISSN : 0975-7384 CODEN(USA) : JCPRC5 A hybrd global optmzaton algorthm based on parallel

More information

ANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING

ANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING ANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING Matthew J. Lberatore, Department of Management and Operatons, Vllanova Unversty, Vllanova, PA 19085, 610-519-4390,

More information

A Novel Methodology of Working Capital Management for Large. Public Constructions by Using Fuzzy S-curve Regression

A Novel Methodology of Working Capital Management for Large. Public Constructions by Using Fuzzy S-curve Regression Novel Methodology of Workng Captal Management for Large Publc Constructons by Usng Fuzzy S-curve Regresson Cheng-Wu Chen, Morrs H. L. Wang and Tng-Ya Hseh Department of Cvl Engneerng, Natonal Central Unversty,

More information

ESTABLISHING TRADE-OFFS BETWEEN SUSTAINED AND MOMENTARY RELIABILITY INDICES IN ELECTRIC DISTRIBUTION PROTECTION DESIGN: A GOAL PROGRAMMING APPROACH

ESTABLISHING TRADE-OFFS BETWEEN SUSTAINED AND MOMENTARY RELIABILITY INDICES IN ELECTRIC DISTRIBUTION PROTECTION DESIGN: A GOAL PROGRAMMING APPROACH ESTABLISHIG TRADE-OFFS BETWEE SUSTAIED AD MOMETARY RELIABILITY IDICES I ELECTRIC DISTRIBUTIO PROTECTIO DESIG: A GOAL PROGRAMMIG APPROACH Gustavo D. Ferrera, Arturo S. Bretas, Maro O. Olvera Federal Unversty

More information

Optimal Bidding Strategies for Generation Companies in a Day-Ahead Electricity Market with Risk Management Taken into Account

Optimal Bidding Strategies for Generation Companies in a Day-Ahead Electricity Market with Risk Management Taken into Account Amercan J. of Engneerng and Appled Scences (): 8-6, 009 ISSN 94-700 009 Scence Publcatons Optmal Bddng Strateges for Generaton Companes n a Day-Ahead Electrcty Market wth Rsk Management Taken nto Account

More information

Problem Set 3. a) We are asked how people will react, if the interest rate i on bonds is negative.

Problem Set 3. a) We are asked how people will react, if the interest rate i on bonds is negative. Queston roblem Set 3 a) We are asked how people wll react, f the nterest rate on bonds s negatve. When

More information

Dynamic Pricing for Smart Grid with Reinforcement Learning

Dynamic Pricing for Smart Grid with Reinforcement Learning Dynamc Prcng for Smart Grd wth Renforcement Learnng Byung-Gook Km, Yu Zhang, Mhaela van der Schaar, and Jang-Won Lee Samsung Electroncs, Suwon, Korea Department of Electrcal Engneerng, UCLA, Los Angeles,

More information

A Secure Password-Authenticated Key Agreement Using Smart Cards

A Secure Password-Authenticated Key Agreement Using Smart Cards A Secure Password-Authentcated Key Agreement Usng Smart Cards Ka Chan 1, Wen-Chung Kuo 2 and Jn-Chou Cheng 3 1 Department of Computer and Informaton Scence, R.O.C. Mltary Academy, Kaohsung 83059, Tawan,

More information

v a 1 b 1 i, a 2 b 2 i,..., a n b n i.

v a 1 b 1 i, a 2 b 2 i,..., a n b n i. SECTION 8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS 455 8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS All the vector spaces we have studed thus far n the text are real vector spaces snce the scalars are

More information

Chapter 4 ECONOMIC DISPATCH AND UNIT COMMITMENT

Chapter 4 ECONOMIC DISPATCH AND UNIT COMMITMENT Chapter 4 ECOOMIC DISATCH AD UIT COMMITMET ITRODUCTIO A power system has several power plants. Each power plant has several generatng unts. At any pont of tme, the total load n the system s met by the

More information

How To Calculate The Accountng Perod Of Nequalty

How To Calculate The Accountng Perod Of Nequalty Inequalty and The Accountng Perod Quentn Wodon and Shlomo Ytzha World Ban and Hebrew Unversty September Abstract Income nequalty typcally declnes wth the length of tme taen nto account for measurement.

More information

"Research Note" APPLICATION OF CHARGE SIMULATION METHOD TO ELECTRIC FIELD CALCULATION IN THE POWER CABLES *

Research Note APPLICATION OF CHARGE SIMULATION METHOD TO ELECTRIC FIELD CALCULATION IN THE POWER CABLES * Iranan Journal of Scence & Technology, Transacton B, Engneerng, ol. 30, No. B6, 789-794 rnted n The Islamc Republc of Iran, 006 Shraz Unversty "Research Note" ALICATION OF CHARGE SIMULATION METHOD TO ELECTRIC

More information

Application of Multi-Agents for Fault Detection and Reconfiguration of Power Distribution Systems

Application of Multi-Agents for Fault Detection and Reconfiguration of Power Distribution Systems 1 Applcaton of Mult-Agents for Fault Detecton and Reconfguraton of Power Dstrbuton Systems K. Nareshkumar, Member, IEEE, M. A. Choudhry, Senor Member, IEEE, J. La, A. Felach, Senor Member, IEEE Abstract--The

More information

Minimal Coding Network With Combinatorial Structure For Instantaneous Recovery From Edge Failures

Minimal Coding Network With Combinatorial Structure For Instantaneous Recovery From Edge Failures Mnmal Codng Network Wth Combnatoral Structure For Instantaneous Recovery From Edge Falures Ashly Joseph 1, Mr.M.Sadsh Sendl 2, Dr.S.Karthk 3 1 Fnal Year ME CSE Student Department of Computer Scence Engneerng

More information

where the coordinates are related to those in the old frame as follows.

where the coordinates are related to those in the old frame as follows. Chapter 2 - Cartesan Vectors and Tensors: Ther Algebra Defnton of a vector Examples of vectors Scalar multplcaton Addton of vectors coplanar vectors Unt vectors A bass of non-coplanar vectors Scalar product

More information

Survey on Virtual Machine Placement Techniques in Cloud Computing Environment

Survey on Virtual Machine Placement Techniques in Cloud Computing Environment Survey on Vrtual Machne Placement Technques n Cloud Computng Envronment Rajeev Kumar Gupta and R. K. Paterya Department of Computer Scence & Engneerng, MANIT, Bhopal, Inda ABSTRACT In tradtonal data center

More information

2008/8. An integrated model for warehouse and inventory planning. Géraldine Strack and Yves Pochet

2008/8. An integrated model for warehouse and inventory planning. Géraldine Strack and Yves Pochet 2008/8 An ntegrated model for warehouse and nventory plannng Géraldne Strack and Yves Pochet CORE Voe du Roman Pays 34 B-1348 Louvan-la-Neuve, Belgum. Tel (32 10) 47 43 04 Fax (32 10) 47 43 01 E-mal: corestat-lbrary@uclouvan.be

More information

We are now ready to answer the question: What are the possible cardinalities for finite fields?

We are now ready to answer the question: What are the possible cardinalities for finite fields? Chapter 3 Fnte felds We have seen, n the prevous chapters, some examples of fnte felds. For example, the resdue class rng Z/pZ (when p s a prme) forms a feld wth p elements whch may be dentfed wth the

More information

Number of Levels Cumulative Annual operating Income per year construction costs costs ($) ($) ($) 1 600,000 35,000 100,000 2 2,200,000 60,000 350,000

Number of Levels Cumulative Annual operating Income per year construction costs costs ($) ($) ($) 1 600,000 35,000 100,000 2 2,200,000 60,000 350,000 Problem Set 5 Solutons 1 MIT s consderng buldng a new car park near Kendall Square. o unversty funds are avalable (overhead rates are under pressure and the new faclty would have to pay for tself from

More information

CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK. Sample Stability Protocol

CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK. Sample Stability Protocol CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK Sample Stablty Protocol Background The Cholesterol Reference Method Laboratory Network (CRMLN) developed certfcaton protocols for total cholesterol, HDL

More information

Joint Scheduling of Processing and Shuffle Phases in MapReduce Systems

Joint Scheduling of Processing and Shuffle Phases in MapReduce Systems Jont Schedulng of Processng and Shuffle Phases n MapReduce Systems Fangfe Chen, Mural Kodalam, T. V. Lakshman Department of Computer Scence and Engneerng, The Penn State Unversty Bell Laboratores, Alcatel-Lucent

More information

Robust Design of Public Storage Warehouses. Yeming (Yale) Gong EMLYON Business School

Robust Design of Public Storage Warehouses. Yeming (Yale) Gong EMLYON Business School Robust Desgn of Publc Storage Warehouses Yemng (Yale) Gong EMLYON Busness School Rene de Koster Rotterdam school of management, Erasmus Unversty Abstract We apply robust optmzaton and revenue management

More information

Dynamic Fleet Management for Cybercars

Dynamic Fleet Management for Cybercars Proceedngs of the IEEE ITSC 2006 2006 IEEE Intellgent Transportaton Systems Conference Toronto, Canada, September 17-20, 2006 TC7.5 Dynamc Fleet Management for Cybercars Fenghu. Wang, Mng. Yang, Ruqng.

More information

Causal, Explanatory Forecasting. Analysis. Regression Analysis. Simple Linear Regression. Which is Independent? Forecasting

Causal, Explanatory Forecasting. Analysis. Regression Analysis. Simple Linear Regression. Which is Independent? Forecasting Causal, Explanatory Forecastng Assumes cause-and-effect relatonshp between system nputs and ts output Forecastng wth Regresson Analyss Rchard S. Barr Inputs System Cause + Effect Relatonshp The job of

More information

On the Interaction between Load Balancing and Speed Scaling

On the Interaction between Load Balancing and Speed Scaling On the Interacton between Load Balancng and Speed Scalng Ljun Chen, Na L and Steven H. Low Engneerng & Appled Scence Dvson, Calforna Insttute of Technology, USA Abstract Speed scalng has been wdely adopted

More information

On the Interaction between Load Balancing and Speed Scaling

On the Interaction between Load Balancing and Speed Scaling On the Interacton between Load Balancng and Speed Scalng Ljun Chen and Na L Abstract Speed scalng has been wdely adopted n computer and communcaton systems, n partcular, to reduce energy consumpton. An

More information

An Evaluation of the Extended Logistic, Simple Logistic, and Gompertz Models for Forecasting Short Lifecycle Products and Services

An Evaluation of the Extended Logistic, Simple Logistic, and Gompertz Models for Forecasting Short Lifecycle Products and Services An Evaluaton of the Extended Logstc, Smple Logstc, and Gompertz Models for Forecastng Short Lfecycle Products and Servces Charles V. Trappey a,1, Hsn-yng Wu b a Professor (Management Scence), Natonal Chao

More information

A DYNAMIC CRASHING METHOD FOR PROJECT MANAGEMENT USING SIMULATION-BASED OPTIMIZATION. Michael E. Kuhl Radhamés A. Tolentino-Peña

A DYNAMIC CRASHING METHOD FOR PROJECT MANAGEMENT USING SIMULATION-BASED OPTIMIZATION. Michael E. Kuhl Radhamés A. Tolentino-Peña Proceedngs of the 2008 Wnter Smulaton Conference S. J. Mason, R. R. Hll, L. Mönch, O. Rose, T. Jefferson, J. W. Fowler eds. A DYNAMIC CRASHING METHOD FOR PROJECT MANAGEMENT USING SIMULATION-BASED OPTIMIZATION

More information

8 Algorithm for Binary Searching in Trees

8 Algorithm for Binary Searching in Trees 8 Algorthm for Bnary Searchng n Trees In ths secton we present our algorthm for bnary searchng n trees. A crucal observaton employed by the algorthm s that ths problem can be effcently solved when the

More information

What is Candidate Sampling

What is Candidate Sampling What s Canddate Samplng Say we have a multclass or mult label problem where each tranng example ( x, T ) conssts of a context x a small (mult)set of target classes T out of a large unverse L of possble

More information

How To Solve An Onlne Control Polcy On A Vrtualzed Data Center

How To Solve An Onlne Control Polcy On A Vrtualzed Data Center Dynamc Resource Allocaton and Power Management n Vrtualzed Data Centers Rahul Urgaonkar, Ulas C. Kozat, Ken Igarash, Mchael J. Neely urgaonka@usc.edu, {kozat, garash}@docomolabs-usa.com, mjneely@usc.edu

More information

Enabling P2P One-view Multi-party Video Conferencing

Enabling P2P One-view Multi-party Video Conferencing Enablng P2P One-vew Mult-party Vdeo Conferencng Yongxang Zhao, Yong Lu, Changja Chen, and JanYn Zhang Abstract Mult-Party Vdeo Conferencng (MPVC) facltates realtme group nteracton between users. Whle P2P

More information

An Analysis of Central Processor Scheduling in Multiprogrammed Computer Systems

An Analysis of Central Processor Scheduling in Multiprogrammed Computer Systems STAN-CS-73-355 I SU-SE-73-013 An Analyss of Central Processor Schedulng n Multprogrammed Computer Systems (Dgest Edton) by Thomas G. Prce October 1972 Techncal Report No. 57 Reproducton n whole or n part

More information

Joe Pimbley, unpublished, 2005. Yield Curve Calculations

Joe Pimbley, unpublished, 2005. Yield Curve Calculations Joe Pmbley, unpublshed, 005. Yeld Curve Calculatons Background: Everythng s dscount factors Yeld curve calculatons nclude valuaton of forward rate agreements (FRAs), swaps, nterest rate optons, and forward

More information

An Alternative Way to Measure Private Equity Performance

An Alternative Way to Measure Private Equity Performance An Alternatve Way to Measure Prvate Equty Performance Peter Todd Parlux Investment Technology LLC Summary Internal Rate of Return (IRR) s probably the most common way to measure the performance of prvate

More information

Credit Limit Optimization (CLO) for Credit Cards

Credit Limit Optimization (CLO) for Credit Cards Credt Lmt Optmzaton (CLO) for Credt Cards Vay S. Desa CSCC IX, Ednburgh September 8, 2005 Copyrght 2003, SAS Insttute Inc. All rghts reserved. SAS Propretary Agenda Background Tradtonal approaches to credt

More information

Research Article Enhanced Two-Step Method via Relaxed Order of α-satisfactory Degrees for Fuzzy Multiobjective Optimization

Research Article Enhanced Two-Step Method via Relaxed Order of α-satisfactory Degrees for Fuzzy Multiobjective Optimization Hndaw Publshng Corporaton Mathematcal Problems n Engneerng Artcle ID 867836 pages http://dxdoorg/055/204/867836 Research Artcle Enhanced Two-Step Method va Relaxed Order of α-satsfactory Degrees for Fuzzy

More information

Economic Models for Cloud Service Markets

Economic Models for Cloud Service Markets Economc Models for Cloud Servce Markets Ranjan Pal and Pan Hu 2 Unversty of Southern Calforna, USA, rpal@usc.edu 2 Deutsch Telekom Laboratores, Berln, Germany, pan.hu@telekom.de Abstract. Cloud computng

More information

DEFINING %COMPLETE IN MICROSOFT PROJECT

DEFINING %COMPLETE IN MICROSOFT PROJECT CelersSystems DEFINING %COMPLETE IN MICROSOFT PROJECT PREPARED BY James E Aksel, PMP, PMI-SP, MVP For Addtonal Informaton about Earned Value Management Systems and reportng, please contact: CelersSystems,

More information

Study on Model of Risks Assessment of Standard Operation in Rural Power Network

Study on Model of Risks Assessment of Standard Operation in Rural Power Network Study on Model of Rsks Assessment of Standard Operaton n Rural Power Network Qngj L 1, Tao Yang 2 1 Qngj L, College of Informaton and Electrcal Engneerng, Shenyang Agrculture Unversty, Shenyang 110866,

More information

Solving Factored MDPs with Continuous and Discrete Variables

Solving Factored MDPs with Continuous and Discrete Variables Solvng Factored MPs wth Contnuous and screte Varables Carlos Guestrn Berkeley Research Center Intel Corporaton Mlos Hauskrecht epartment of Computer Scence Unversty of Pttsburgh Branslav Kveton Intellgent

More information

Optimization under uncertainty. Antonio J. Conejo The Ohio State University 2014

Optimization under uncertainty. Antonio J. Conejo The Ohio State University 2014 Optmzaton under uncertant Antono J. Conejo The Oho State Unverst 2014 Contents Stochastc programmng (SP) Robust optmzaton (RO) Power sstem applcatons A. J. Conejo The Oho State Unverst 2 Stochastc Programmng

More information

Data Broadcast on a Multi-System Heterogeneous Overlayed Wireless Network *

Data Broadcast on a Multi-System Heterogeneous Overlayed Wireless Network * JOURNAL OF INFORMATION SCIENCE AND ENGINEERING 24, 819-840 (2008) Data Broadcast on a Mult-System Heterogeneous Overlayed Wreless Network * Department of Computer Scence Natonal Chao Tung Unversty Hsnchu,

More information

Energy Efficient Routing in Ad Hoc Disaster Recovery Networks

Energy Efficient Routing in Ad Hoc Disaster Recovery Networks Energy Effcent Routng n Ad Hoc Dsaster Recovery Networks Gl Zussman and Adran Segall Department of Electrcal Engneerng Technon Israel Insttute of Technology Hafa 32000, Israel {glz@tx, segall@ee}.technon.ac.l

More information

Linear Circuits Analysis. Superposition, Thevenin /Norton Equivalent circuits

Linear Circuits Analysis. Superposition, Thevenin /Norton Equivalent circuits Lnear Crcuts Analyss. Superposton, Theenn /Norton Equalent crcuts So far we hae explored tmendependent (resste) elements that are also lnear. A tmendependent elements s one for whch we can plot an / cure.

More information

Fisher Markets and Convex Programs

Fisher Markets and Convex Programs Fsher Markets and Convex Programs Nkhl R. Devanur 1 Introducton Convex programmng dualty s usually stated n ts most general form, wth convex objectve functons and convex constrants. (The book by Boyd and

More information

Medium and long term. Equilibrium models approach

Medium and long term. Equilibrium models approach Medum and long term electrcty prces forecastng Equlbrum models approach J. Vllar, A. Campos, C. íaz, Insttuto de Investgacón Tecnológca, Escuela Técnca Superor de Ingenería-ICAI Unversdad ontfca Comllas

More information

Calculation of Sampling Weights

Calculation of Sampling Weights Perre Foy Statstcs Canada 4 Calculaton of Samplng Weghts 4.1 OVERVIEW The basc sample desgn used n TIMSS Populatons 1 and 2 was a two-stage stratfed cluster desgn. 1 The frst stage conssted of a sample

More information

Staff Paper. Farm Savings Accounts: Examining Income Variability, Eligibility, and Benefits. Brent Gloy, Eddy LaDue, and Charles Cuykendall

Staff Paper. Farm Savings Accounts: Examining Income Variability, Eligibility, and Benefits. Brent Gloy, Eddy LaDue, and Charles Cuykendall SP 2005-02 August 2005 Staff Paper Department of Appled Economcs and Management Cornell Unversty, Ithaca, New York 14853-7801 USA Farm Savngs Accounts: Examnng Income Varablty, Elgblty, and Benefts Brent

More information

Course outline. Financial Time Series Analysis. Overview. Data analysis. Predictive signal. Trading strategy

Course outline. Financial Time Series Analysis. Overview. Data analysis. Predictive signal. Trading strategy Fnancal Tme Seres Analyss Patrck McSharry patrck@mcsharry.net www.mcsharry.net Trnty Term 2014 Mathematcal Insttute Unversty of Oxford Course outlne 1. Data analyss, probablty, correlatons, vsualsaton

More information

Pricing Model of Cloud Computing Service with Partial Multihoming

Pricing Model of Cloud Computing Service with Partial Multihoming Prcng Model of Cloud Computng Servce wth Partal Multhomng Zhang Ru 1 Tang Bng-yong 1 1.Glorous Sun School of Busness and Managment Donghua Unversty Shangha 251 Chna E-mal:ru528369@mal.dhu.edu.cn Abstract

More information

An MILP model for planning of batch plants operating in a campaign-mode

An MILP model for planning of batch plants operating in a campaign-mode An MILP model for plannng of batch plants operatng n a campagn-mode Yanna Fumero Insttuto de Desarrollo y Dseño CONICET UTN yfumero@santafe-concet.gov.ar Gabrela Corsano Insttuto de Desarrollo y Dseño

More information

Addendum to: Importing Skill-Biased Technology

Addendum to: Importing Skill-Biased Technology Addendum to: Importng Skll-Based Technology Arel Bursten UCLA and NBER Javer Cravno UCLA August 202 Jonathan Vogel Columba and NBER Abstract Ths Addendum derves the results dscussed n secton 3.3 of our

More information

Availability-Based Path Selection and Network Vulnerability Assessment

Availability-Based Path Selection and Network Vulnerability Assessment Avalablty-Based Path Selecton and Network Vulnerablty Assessment Song Yang, Stojan Trajanovsk and Fernando A. Kupers Delft Unversty of Technology, The Netherlands {S.Yang, S.Trajanovsk, F.A.Kupers}@tudelft.nl

More information

Calculating the high frequency transmission line parameters of power cables

Calculating the high frequency transmission line parameters of power cables < ' Calculatng the hgh frequency transmsson lne parameters of power cables Authors: Dr. John Dcknson, Laboratory Servces Manager, N 0 RW E B Communcatons Mr. Peter J. Ncholson, Project Assgnment Manager,

More information

IMPACT ANALYSIS OF A CELLULAR PHONE

IMPACT ANALYSIS OF A CELLULAR PHONE 4 th ASA & μeta Internatonal Conference IMPACT AALYSIS OF A CELLULAR PHOE We Lu, 2 Hongy L Bejng FEAonlne Engneerng Co.,Ltd. Bejng, Chna ABSTRACT Drop test smulaton plays an mportant role n nvestgatng

More information

Descriptive Models. Cluster Analysis. Example. General Applications of Clustering. Examples of Clustering Applications

Descriptive Models. Cluster Analysis. Example. General Applications of Clustering. Examples of Clustering Applications CMSC828G Prncples of Data Mnng Lecture #9 Today s Readng: HMS, chapter 9 Today s Lecture: Descrptve Modelng Clusterng Algorthms Descrptve Models model presents the man features of the data, a global summary

More information

Efficient Bandwidth Management in Broadband Wireless Access Systems Using CAC-based Dynamic Pricing

Efficient Bandwidth Management in Broadband Wireless Access Systems Using CAC-based Dynamic Pricing Effcent Bandwdth Management n Broadband Wreless Access Systems Usng CAC-based Dynamc Prcng Bader Al-Manthar, Ndal Nasser 2, Najah Abu Al 3, Hossam Hassanen Telecommuncatons Research Laboratory School of

More information

A Lyapunov Optimization Approach to Repeated Stochastic Games

A Lyapunov Optimization Approach to Repeated Stochastic Games PROC. ALLERTON CONFERENCE ON COMMUNICATION, CONTROL, AND COMPUTING, OCT. 2013 1 A Lyapunov Optmzaton Approach to Repeated Stochastc Games Mchael J. Neely Unversty of Southern Calforna http://www-bcf.usc.edu/

More information

Period and Deadline Selection for Schedulability in Real-Time Systems

Period and Deadline Selection for Schedulability in Real-Time Systems Perod and Deadlne Selecton for Schedulablty n Real-Tme Systems Thdapat Chantem, Xaofeng Wang, M.D. Lemmon, and X. Sharon Hu Department of Computer Scence and Engneerng, Department of Electrcal Engneerng

More information

PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 12

PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 12 14 The Ch-squared dstrbuton PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 1 If a normal varable X, havng mean µ and varance σ, s standardsed, the new varable Z has a mean 0 and varance 1. When ths standardsed

More information

LIFETIME INCOME OPTIONS

LIFETIME INCOME OPTIONS LIFETIME INCOME OPTIONS May 2011 by: Marca S. Wagner, Esq. The Wagner Law Group A Professonal Corporaton 99 Summer Street, 13 th Floor Boston, MA 02110 Tel: (617) 357-5200 Fax: (617) 357-5250 www.ersa-lawyers.com

More information

Cost Minimization using Renewable Cooling and Thermal Energy Storage in CDNs

Cost Minimization using Renewable Cooling and Thermal Energy Storage in CDNs Cost Mnmzaton usng Renewable Coolng and Thermal Energy Storage n CDNs Stephen Lee College of Informaton and Computer Scences UMass, Amherst stephenlee@cs.umass.edu Rahul Urgaonkar IBM Research rurgaon@us.bm.com

More information

Multi-Period Resource Allocation for Estimating Project Costs in Competitive Bidding

Multi-Period Resource Allocation for Estimating Project Costs in Competitive Bidding Department of Industral Engneerng and Management Techncall Report No. 2014-6 Mult-Perod Resource Allocaton for Estmatng Project Costs n Compettve dng Yuch Takano, Nobuak Ish, and Masaak Murak September,

More information

APPLICATION OF PROBE DATA COLLECTED VIA INFRARED BEACONS TO TRAFFIC MANEGEMENT

APPLICATION OF PROBE DATA COLLECTED VIA INFRARED BEACONS TO TRAFFIC MANEGEMENT APPLICATION OF PROBE DATA COLLECTED VIA INFRARED BEACONS TO TRAFFIC MANEGEMENT Toshhko Oda (1), Kochro Iwaoka (2) (1), (2) Infrastructure Systems Busness Unt, Panasonc System Networks Co., Ltd. Saedo-cho

More information

POLYSA: A Polynomial Algorithm for Non-binary Constraint Satisfaction Problems with and

POLYSA: A Polynomial Algorithm for Non-binary Constraint Satisfaction Problems with and POLYSA: A Polynomal Algorthm for Non-bnary Constrant Satsfacton Problems wth and Mguel A. Saldo, Federco Barber Dpto. Sstemas Informátcos y Computacón Unversdad Poltécnca de Valenca, Camno de Vera s/n

More information