1 FIA: Information as an economic good Enabling role of the INFORMATION BROKER (an example from Energy domain)
2 We cannot change what we don t know Alexander Bell, Thomas Edison, Nicola Tesla, John V. Atanasoff, Clifford E. Berry But we can apply ICT capacity to process information distributedly & elicit knowledge to energy (smart grid)
3 Information and anticipatory knowledge Have you ever heard any Commercial Company anticipating information about expected production drops ( looses for shareholders)? Is it normal for Insurance Companies to share the risks in order to reduce the exposition/ potential loses? The renewable (solar, wind) energy is depending on the weather conditions, so the production drops happen normally. It affect all the blind producers because of the SLAs. What if the first plant experiencing any drop broadcasts this anticipatory knowledge to the neighbours. Is it reducing unpredictability or shares SLA s penalties, if any?
4 Carbon Economy E j (t) E i (t) Energy is NOT free Imbalance E j (t) - E i (t) Min Eccesso di consumo Imbalance Imbalance Eccesso di produzione Mankind consumes different kinds of energy for living, industry and other activities because of the goals. Causing increased emissions is natural, but to respect the Nature the carbon footprint should be contained. A low carbon economy depends on the way we use the energy, transport, and mobility. How the right information can help to achieve a virtually carbon-neutral impact with a current level of safety? Past behaviour (collection of load shapes) + soft computing = more precise load forecast
5 Information is an economic good Phenomenon Information (load shapes) Capability to do something Value Billing data Billing procedure Revenue Consumption data Analysis procedures Savings Load data Load Forecast procedures Business operations measurement Cost of data cause but information processing? Events (real-time) Lamp on-off Lack of capacity to process real-time events (is not free)
6 Information is an economic good Right information = product on internet market (smart grids) Value(information, f(enabled Capability)) = F( CostOf(data) ) (load shape), at the right time (real-time), in the right place (role) enables the new business. Load shapes + real-time -> independent PV producers to make them cooperating. Load profiles + real-time -> Information broker in energy makes savings real. Energy dynamics + real-time -> Dynamic load balancing = lower imbalance = savings Real-time M2M information brokerage is the service enabling the new business relationships Value(information, f(enabled Capability)) = F( CostOf(data) ) Value(load shape, Billing) = CostOf(ManualMetering) CostOf(AMM) Value(load shape, Load balancing) = C(ε ξ) - CostOf(AMM 15 -AMM realtime ) Value(load shape, PV forecast) = C(ε ξ) - CostOf(AMM 15 -AMM realtime ) Value(real-time load shape, Broker) = 0,5*MarketSize*Savings - CostOf(AMM rt )
7 WhatIf vis-à-vis HowTo Information containing some anticipatory knowledge GAMUT is an implementation coming from HPPC/SEA EU project. In backoffice we create the model In front office we apply it to the datasets coming in real-time. Data Set Designer Repository Data Set Generator (Client) Data Set Generator (Server) Data Analyzer (Client) Data Analyzer (Server) Genetic Engine (Client) Genetic Engine (Server) Model Manager Data Set Designer Dataset populated Data Analyzer Dataset filtered Genetic Engine Model Manager % Rating Decision Data Set Generator Dataset generation Analisys and filtering Load shape = real-time sdata Model creation Prevision Model Transformation maps Data customer Application of the model
8 Information processing is a cost or added value? Phenomenon measurement cause Cost of data Information processing says about one energy production drop happened between 2 and 3 PM. More knowledge about the event requires real-time data + data transmission/elaboration expenses. New information should bring more value than the processing costs, or should enable something new. but is it feasible? and why to do this?
9 Low Carbon Economy: individually or collectively? 347,37 TWh (2009, I): Each E j (t) -> Max, imbalance -> Min Fault, looses Looses, fault Fault, looses Looses, fault Energy plant E 1 (t) Control center Residential consumption (LV) 40% ca. Hydroelectric plant E i (t) Transport network Distribution network E k (t) Substation (distribution) Fault, looses Genration from Renewable sources (less predictable) Looses, fault E j (t) Substation (transport) Fault, loses E l (t) Looses, fault Industrial consumption (HV) 50% is likely inefficient Power grid is balanced dynamically. Some imbalance in power grid is normal. But peak energy is expensive. Balance = exact knowledge about the demand and the offer. Unfortunately we don t have exact numbers. The right information it the source for optimisation. Imbalance = money = resources lost = CO 2. Imbalance can be lowered. A low carbon economy = new way to use the information + how we use the energy, transport, and mobility
10 New technology Legacy challenges Utilities Buy/Sell load shapes, behaviour N independent consumers C i and producers P i each running its own probabilistic forecast Producers Customers 1 second sampling is not needed to operate same way like in the past. Same business, but +costs. Personalisation of the services = +costs. Information processing = +costs.
11 New technology new F.I. challenges Wish to compete: more info = value Utilities Old stakeholder broker Buy/Sell load shapes, behaviour Producers Customers Real-time = +knowledge = +market chance = +personalisation = better quality of service. Personalisation of the services = +competitiveness. Information processing = +opportunities. How to transform +costs (for me) in +revenues (for me), and +savings (for my users)? but Pattern 2 appears similar to Pattern 1, even if some delayed some compressed likely because of X (new knowledge)
12 UC1. Information broker: acting independently or collectively? Use Case 1 (liberalized market) C 1 and C 2, P 1 and P 2 P 1 =20*0,13+30*0,11+40*0,12+10*0,15=12,2 P 2 =30*0,11+20*0,12+10*0,15+20*0,12=9,6 P =P 1 +P 2 = 21,8 Use Case 2 (liberalized market) Broker, Knowledge about load profiles, C 1 and C 2, P 1 and P 2 broker P 1 =20*0,09+30*0,09+40*0,09+10*0,11=9,2 P 2 =30*0,09+20*0,09+10*0,09+20*0,11=7,6 P =P 1 +P 2 = 16,8 ΔP = 5 ΔP/P=29% ΔP 1 =2,6 (21%), ΔP 2 =1,6 (16%) Wish you save money on the energy bill? Will you pay broker revenue = 50% of extra savings? The added value comes from the anticipatory knowledge about pending needs. So 50% on -10% out of 30% (residential) But cost of ICT infrastructure and info processing
13 UC2. Cooperation: advantages and challenges Decision maker TCP-IP Information flow fieldbus Intelligent server N 1 N 2 N j N k LV i Information bus (FI) LV j Future Internet interconnects LV segments and make them interoperable. Events can be shared now. FI makes possible sharing information between neighbours belonging to different LV segments. Common information space = common knowledge, e.g. anticipatory knowledge being unbounded. E prev i(t k+1 ) - E real i(t k+1 ) = ε i N independent nodes: error = n*ε i E k (t k ) = k * E i (t k +Δt) anticipation Multi-agent system: ε i + (n-1) * ξ I, ε i > ξ I because of anticip. knowledge
14 Anticipatory knowledge and its value M. Simonov et all, Information Processing in Smart Grids and Consumption Dynamics, in A. Soro et all, Information Retrieval and mining in Distributed Environments, SCI, Vol. 324/2011, pp , Springer, Production drop Duration Lowering imbalance = dynamic load balancing based on = enabled by the knowledge about real-time events (causes of the energy dynamics) Expected consumption (known now) Δt Value added: Save = Σ(E prev E next ) = Σδ E
15 Multi-agent cooperation to reduce imbalance M. Simonov e all, Modello di gestione dell energia solare in tempo reale, in: FOCUS AEIT, Vol. 1, 2011, pp. 1-7, AEIT, 2011 (in press) F. Grimaccia, M. Simonov et all, Modelli predittivi per la produzione da impianti fotovoltaici mediante tecniche di soft computing, in: FOCUS AEIT, V.1, pp.1-7, AEIT (in press) E i (t 0 ),, E i (t l ) E j (t 0 ),, E j (t l ) Rainstorm approaching E 1 (t 0 ),, E 1 (t l ) <DE: Cloud, -25%, min> <DE: <DE: Cloud, Cloud, -25%, -25%, 3 3 min> min> Anticipatory knowledge FI E k (t 0 ),, E k (t l ) E m (t 0 ),, E m (t l )
16 Anticipatory knowledge and its value Phenomenon measurement E i (t 0 ),, E i (t l ) E prev i (t k+1 ) - Ereal i (t k+1) = ε i error of probabilistic forecast cause Cost of data < Cost of energy Cost of data < Value of data Value of data <> Cost of energy? a) Incentivated prices b) Market prices E j (t 0 ),, E j (t l ) E k (t 0 ),, E i (t l ) E k (t k ) = k * E i (t k +Δt) anticipation because sharing knowledge broker M.Simonov et all, Real time energy management in smart cities by Future Internet, in G. Tselentis et all, Towards the Future Internet, pp , IOS Press, 2010.
17 UC3. Mastering smart grid with desired characteristics In DG, the inversion of the active power s flow might happen. The flow from LV to MV and HV might make damages. We have proposed a method for dynamic partitioning to build one micro-grid showing desired behavioural characteristics. No inversion of the flows = total consumption locally M. Simonov et all, Digital Energy: Clustering Micro Grids for Social Networking, Int.J. of Virtual Communities and Social Networking, Vol.1 (3), pp , IGI Global ~ ~ ~ ~ ~ ~ Intermediate steps How to 1) Use the information (load shapes) for similarity clustering 2) Use soft computing (GA and ANN) 3) Optimize clusters placing users into clusters (social groups) 3) Materialize new micro-grid consuming in full (all the time) 1 Meteo (service) Power (service) 2 Industry 3 Business 4 Residential 5
18 Impact: the value of the above information Distributed energy generation brings advantages, but bi-directional flows can invert flows (from LV to HV), whose can damage energy actors. SLA should be respected (penalty): firm commitment = +precision of the forecast. Generation from RES is less predictable because of the weather (wind = +Eolic, clouds = -PV). Less imbalance = savings = software to manage the distributed topology of the intelligent nodes in smart grid. Sharing information = non probabilistic forecast = better precision = less imbalance = additional revenues N independent energy generation plants = N independent forecasts = N possible errors. One cooperative multi-agent system = multi-parametric optimisation and better observability. Dedicated software can be distributed to all nodes to make them intelligent. 1% of PV+Eolic imbalance = 6,6 Meuro/year in 2009 in I only... MS Energy ,90 A low carbon economy = new way to use the information + how we apply it to energy.
19 Questions? Availability of real time information in energy domain = tangible advantages. Cost of data < Avoiding penalty because of the inverted flows (in DG from RES). Cost of data < Savings and business opportunities for the Broker Cost of data < Reduction of imbalance = savings Cost of data < Cooperation between RES plants Cost of data < new possible share on Internet market (SW + real-time counter + broadband) Cost of data < Intraday M2M trading Cost of data < Green certificate trading (analytical accounting of kinds of energy) MS Energy ,90 Cost of data < Value gained in CO 2 terms.
20 More renewable, more predictable, less imbalance M. Simonov e all, Modello di gestione dell energia solare in tempo reale, in: AEIT, Vol. 1, 2011, pp. 1-7, AEIT, 2011 (in press) F. Grimaccia, M. Simonov et all, Modelli predittivi per la produzione da impianti fotovoltaici mediante tecniche di soft computing, in: AEIT, V.1, pp.1-7, AEIT (in press) The anticipatory knowledge about the events to happen will permit to increase the percentage of PVs, keeping the power grid stable, because solving unpredictability. Weather forecast E i (t 0 ),, E i (t l ) PV panels E m (t 0 ),, E m (t l ) PV panels Energy Web Anticipatory knowledge FI DE: Cloud, -25%, 3 min Smart City E k (t 0 ),, E k (t l ) Broadcasting the Cloud event www? E j (t 0 ),, E j (t l ) Smart City (2) Cloud