Process data validation in CCGT and nuclear power plants

Process data validation in CCGT and nuclear power plants Jansky J., Langenstein M. BTB-Jansky GmbH, Leonberg, Germany ABSTRACT Process data validation is employed for monitoring and optimising processes in nuclear power plants, combined-cycle gas turbine power plants, coal-fired power plants, refuse incineration plants and gas power pools, as well as in the chemical and petrochemical industries. It is a method for determining a consistent process image, taking closed mass, energy and materials balances into account. It takes the place of: Cost-intensive acceptance and inspection measurements, Part of the instrumentation within the processes. If it is installed online, it supplies a continuous stream of results, allowing impermissible developments during operation or in the plant to be quantitatively estimated. Data validation also yields information about operating states in non-instrumented or insufficiently instrumented parts of the plant. If data validation is integrated in the installation early on during the project planning phase, it is possible to reduce the number of measuring points compared with that considered necessary in the past. Since the saving per measuring point is approximately 15,000 EURO, this can represent a substantial reduction in the cost for investments in the plant. Depending on the type and size of the plant, as much as 125,000 EURO can be saved additionally every year because fewer key measuring points need to be calibrated. This work describes experience with online process data validation in CCGT and nuclear power plants (boiling water and pressurised water reactors). INTRODUCTION Process data validation is a method for determining a consistent process image of data records containing mean measured values, taking closed mass, energy and materials balances into account. The method is based on a procedure for identifying the most probable process state in the mathematical-statistical sense. It is described in VDI 2048 [1]. When it comes to mapping a real process, this means that all available measured quantities, in other words including all redundant measured quantities, are combined in a single plant diagram. The overdetermined system of equations that results is solved using the Gaussian calculus of observations. The "true" values calculated with a probability greater than 95 % reflect the most likely physical state of the process. This method is particularly effective in connection with process monitoring and optimisation [2], [3], [4]. This work describes experience with online process data validation in: CCGT power plants, and Nuclear power plants (boiling water and pressurised water reactors) The improvements in operating management efficiency and the potential calibration economies which are facilitated by online validation are quantified. 1

THEORETICAL BASIS The theoretical basis of data validation is explained with the aid of an example. Figure 1 shows a SPLITTER into which one mass flow enters and from which two mass flows exit. A mass flow rate of 500 t/h is measured for stream 1, while the value for stream 2 is 245 t/h and that for stream 3 is 250 t/h. The following condition must be fulfilled: m stream 1 = mstream 2 + mstream 3 (1) Fig. 1: Example Each of the measured values is assigned a permissible standard deviation equivalent to 5 % (of the measured value), resulting in the Gaussian normal distribution shown in Figure 1 being allocated to each measurement. The simply overdetermined system is then solved by means of a calculus of observations taking a minimisation criteria into account. The minimisation criteria (PENALTY function) is as follows: 2 measurement reconciled value = minimum (2) standard deviation A PENALTY-value of 0.0256 is calculated for this example. The results of the calculus of observations are also presented in Figure 1. The validated (reconciled) values now satisfy Eq. 1 (mass balance closed). USE IN NUCLEAR POWER PLANTS Process data validation with VALI III [5[ is used in nuclear power plants in order to: Perform acceptance measurements Trace start-up activities in the plants Determine the mean coolant temperature more accurately Determine the thermal reactor output more accurately Use the validation results as a calibration standard Reduce the cost of calibrating measuring points 2

Fig. 2: Primary circuit of a nuclear power plant Figure 2 shows the primary circuit of a nuclear power plant on the VALI user interface. Figures 3 to 5 show the results of a validation in a nuclear power plant. Temperature in [ C] 311 310,5 310 309,5 309 308,5 308 307,5 JEC10CT711 Reconciled JEC20CT721 Reconciled JEC30CT731 Reconciled JEC40CT741 Reconciled JEC10CT711 Measurement JEC20CT721 Measurement JEC30CT731 Measurement JEC40CT741 Measurement 307 306,5 306 7.9.98 7.9.98 7.9.98 7.9.98 7.9.98 7.9.98 7.9.98 7.9.98 7.9.98 Fig. 3: Graph of the mean coolant temperature Figure 3 reveals that the measured mean coolant temperatures are up to 2 K higher than the reconciled values. Since the mean coolant temperature is one of the most important controlled variables in the process, it is vital that the true temperatures should be determined precisely. This may determine whether the nuclear power plant is able to reach its full 100 % output at all. 3

2160 2150 sum of measurements sum of reconciled values 2140 Feed water flow in [kg/s] 2130 2120 2110 2100 Mean value deviation: 25 kg/s (1% of the total feed water flow) 2090 09.01.01 10.01.01 11.01.01 12.01.01 13.01.01 14.01.01 15.01.01 16.01.01 17.01.01 18.01.01 19.01.01 Date Fig. 4: Total feed water mass flow of the steam generator 68 300 67 66 250 65 64 200 Pressure in [bar] 63 62 61 60 59 150 100 PENALTY value 58 57 LAB60CP001 reconciled value LAB60CP001 measurement PENALTY Reconciled 50 56 0 23.7.00 00:00:00 02.8.00 00:00:00 12.8.00 00:00:00 22.8.00 00:00:00 01.9.00 00:00:00 11.9.00 00:00:00 21.9.00 00:00:00 Fig. 5: Drifting measuring point 4

Figure 4 compares the feed water mass flow of the four steam generators with the reconciled values. The deviation between the measured feed water mass flow and the reconciled mass flow is 25 kg/s, equivalent to 1 % of the total mass flow (accuracy range of the measuring devices). The aim is to use the reconciled values as a calibration standard and to calibrate the measuring devices such that the measured values correspond to the reconciled values. Since the feed water mass flow measurement is relevant to the safety of the plant, the approval of the independent expert is required. This procedure has already been completed, i.e. the results of the validation can be used as a calibration standard. In the example described here, the electrical output is increased by 13 MW el. Figure 5 confirms how the feed water pressure measurement drifts upstream of one steam generator. The same graph also shows the PENALTY-function over the drift period. The PENALTY-function is the sum of all deviations of the model as a whole according to Eq. 2. During operation under 100 % load, the PENALTY-value for this nuclear power plant is 70. The measuring point drift causes the PENALTY-value to rise to 300. This is indicated by the system directly by means of a reference to the faulty measuring point that caused this increase. Data validation is thus an extremely effective instrument for monitoring processes, since it enables state-oriented calibration of measuring points. State-oriented calibration allows a nuclear power plant to economise 125,000 EURO per year. The investment costs for a data validation system are hence recuperated within just twelve months. USE IN CCGT POWER PLANTS The aim of data validation in combined-cycle gas turbine power plants is to: Perform acceptance tests much more precisely than is possible using conventional acceptance test methods Obtain closed mass and energy balances (important when supplying district heat) Monitor the process more efficiently Obtain more detailed information about the gas turbine Identify potential problems in the combustion chambers at an early stage Monitor wholesale purchases of natural gas and the heating energy billed by the supplier Reduce the cost of calibrating measuring points Figure 6 shows a gas turbine on the VALI user interface. Compressor HP turbine Generator Fig. 6: Model of a gas turbine LP turbine 5

A more detailed comparison of acceptance tests with and without data validation can be found in [6]. Table 1 shows the considerable differences that exist between a conventional acceptance test and an acceptance test with data validation. Table 1: Comparison of the results of a conventional acceptance test and an acceptance test with data validation Description Conventional acceptance test Acceptance test with data validation Value Tolerance Value Tolerance Guaranteed net electrical output 366.7 MW 366.7 MW Measured net electrical output 367 MW +/- 0.929 MW 366.98 MW +/- 0.573 MW Guaranteed net heat consumption 6211.7 kj/kwh 6211.7 kj/kwh Measured net heat consumption 6195.4 kj/kwh +/- 146.5 kj/kwh 6196.4 kj/kwh +/- 16.6 kj/kwh If only those measured quantities that are directly needed to calculate the end result are taken into account, the prospective additional consumption of heat amounts to 146.5 KJ/kWh. This represents an average annual financial loss to the customer of 1,648,250 EURO. If a larger number of measured quantities over and above those directly needed for the calculation are used in conjunction with data validation, the number of uncertainties in the end results is significantly reduced. The prospective additional consumption of heat amounts instead to 16.6 kj/kwh, equivalent to an annual financial loss to the customer of 186,700 EURO. The maximum value of 200,000 EURO laid down in the contract is thus adhered to. 96 95 94 GAS_MFC1 Reconciled Messwert Mittel 11 10,8 10,6 Methananteil in [Vol.- %] 93 92 91 90 89 88 87 86 03.02.01 00:00 08.02.01 00:00 13.02.01 00:00 18.02.01 00:00 23.02.01 00:00 28.02.01 00:00 05.03.01 00:00 unterer Heizwert in [kwh/nm³] 10,4 10,2 10 9,8 9,6 9,4 LOWER HEAT VALUE Messwertmittel 9,2 LOWER HEAT VALUE Reconciled 9 03.02.01 00:00 08.02.01 00:00 13.02.01 00:00 18.02.01 00:00 23.02.01 00:00 28.02.01 00:00 05.03.01 00:00 Fig. 7: Methane component analysed and reconciled Fig. 8: Lower heating value analysed and reconciled Figure 7 shows the analysed and reconciled methane components of the gas. Figure 8 compares the lower heating value based on analysed values with the reconciled lower heating value. These values cannot be traced ONLINE without a validation because the results of the analysis are never available simultaneously. Figure 9 shows the measured and reconciled main steam temperatures. The maximum measured temperature is up to 5 K higher than the reconciled temperature. This means that the plant is not being operated with its maximum efficiency. Reconciled data is required in order to establish this, however. 6

545 540 Main steam temperature in [ C] 535 530 525 40LBA11CT001_XQ50 Measurement 40LBA12CT001_XQ50 Measurement 40LBA20CT001_XQ50 Measurement 40LBA20CT002_XQ50 Measurement 40LBA20CT006_XQ50 Measurement 40LBA12CT001_XQ50 Reconciled 520 01.10.00 00:00 21.10.00 00:00 10.11.00 00:00 30.11.00 00:00 20.12.00 00:00 09.01.01 00:00 29.01.01 00:00 18.02.01 00:00 10.03.01 00:00 Fig. 9: Graph of the measured and reconciled main steam temperatures 1170 1165 Temperature at inlet of EV combustion chamber in [ C] 1160 1155 1150 1145 1140 1135 1130 41MBA30CT912_XJ11 Measurement 41MBA30CT912_XJ11 Reconciled 03.02.01 00:00 08.02.01 00:00 13.02.01 00:00 18.02.01 00:00 23.02.01 00:00 28.02.01 00:00 05.03.01 00:00 Fig. 10: Measured and reconciled EV combustion chamber temperatures 7

Figure 10 shows the measured and reconciled EV combustion chamber temperatures. The measured value is approximately 20 K lower than the reconciled value, corresponding to a 2 % deviation. CONCLUSIONS Process optimisation is only possible at all if detailed information about the complete set of process values is available. Particularly in the case of large power plants (nuclear: 1300 MW, CCGT: 500 MW), measured values inside the permissible ranges may result in considerable financial losses. Failure to exploit the full thermal reactor output over a period of several decades in the order of 40 MW th can mean a loss of several hundred million euros in today's terms. Similar losses are likely to be incurred by a CCGT power plant if the operating regime is not optimised. For this reason, all plant owners are urgently advised to implement continuous data validation in their process monitoring concept. REFERENCES [1] VDI 2048, Uncertainties of measurement during acceptance tests on energy-conversion and power plants - fundamentals, October 2000 [2] U. Brockmeier; Validierung von Prozeßdaten in Kraftwerken VGB-Kraftwerkstechnik Issue 9/99 Pages 61-66 [3] E. Grauf, J. Jansky, M. Langenstein; Investigation of the real process data on basis of closed mass and energy balances in nuclear power plants (NPP); SERA-Vol. 9, Safety Engineering and Risk Analysis - 1999, Pages 23-40; edited by J.L. Boccio; ASME 1999 [4] Grauf, E., Jansky, J., Langenstein, M.: Validation of process data in nuclear power plants (NPPs), 8th International Conference on Nuclear Engineering (ICONE) April 2-8, 2000 Baltimore, MD USA [5] Vali III USER GUIDE, BELSIM S.A., Liege, Belgium, October 2000 [6] Dr. Streit, Wienstrom; "Arbeitsentwurf zu VDI 2048" dated 13.10.2000 8