ACCURATE MODELING FOR LOSSES REDUCTION USING A REAL-TIME POWER FLOW AT ENELVEN

Transcription

1 ACCURATE MODELING FOR LOSSES REDUCTION USING A REAL-TIME POWER FLOW AT ENELVEN Renato CESPEDES I. Roytelman, A.Ilo P.Parra,L.Rodriguez,H.Socorro,T. Romero KEMA Consulting Colombia SIEMENS USA,Austria ENELVEN - Venezuela rcespedes@kemaconsulting.com iroytelm@siemens-emis.com proyecto.dms@enelven.com.ve INTRODUCTION ENELVEN is the utility responsible for generation, transmission and distribution of electric energy in the Zulia State,Venezuela. ENELVEN has more than 600,000 users and distributes approximately 1,700 MW of power. Due to the quality of service requirements and the need to improve the operation efficiency (power and blackout losses were approximately 24% in the year 2001), ENELVEN started an ambitious plan to manage and automate their extensive distribution power system. ENELVEN s distribution system includes eight types of distribution transformer connections, some of which are not very common. Accurate simulation of the loading for each banked transformer leg, in said connections, was required as a part of the correct power flow solution. The main tool used in the network analysis portion of the Distribution Management System applications is the real-time multiphase Distribution Systems Power Flow (DSPF).The distribution power flow solution is used for two main purposes: a) Check loading and voltage constraints - quality service requirements. b) Power loss calculation and identification of the lines and transformers where losses are the highest - operation efficiency. This paper presents an overview of the adopted solution approach with emphasis on the extensive transformers and loads modeling developed as a part of DSPF for ENELVEN s distribution system. It also shows some uncommon DSPF results from the ENELVEN distribution feeders which prove a real need for such extensive modeling. I. PROJECT OBJECTIVES To supply the electricity service in its area, ENELVEN and its sister company ENELCO, part of the same holding, have distribution networks as described hereinafter: Table 1.1a ENELVEN Distribution System Type Volts (kv) Feeders Length (km) Urban ,684.2 Urban Rural NE Rural ,854.3 NW Rural ,671.2 SW Total Table 1.1b ENELCO Distribution System Voltage (kv) Feeders Length (km) , Total Starting in 1997 when an overall strategic plan was defined by ENELVEN with the assistance of KEMA Consulting for improving the Distribution system operation, several steps were taken with company goals which received the total support of ENELVEN management. Since project inception, ENELVEN has defined general and specific project objectives. Some of them are of qualitative nature, among which the most important are: General Objectives: Improve quality of service. More efficient management of the distribution system under normal operating conditions. Respond more adequately to service outages. Maintain technical losses at a minimum level. KEM_Cespedes_A1 Session 3 Paper No

2 Improve distribution system operation security and reliability. Efficient substation management. Personalize customer service from trouble call to service restoration. Detect and follow-up the evolution of non-technical losses. Specific Objectives: Reduce the time dedicated to outage repairs. Reduce the service restoration times. Reduce the number of users affected by outages and service suspensions. Reduce the investment costs by more efficient use of the system equipment. Reduce operation and maintenance costs. Reduce non-technical losses. Increase customer satisfaction. Improve distribution system operation planning and analysis. Increase the overall company benefit to cost ratio related with distribution operation. These objectives have short-, medium- and long-term goals and have been attained in accordance with the phased project development. II. DISTRIBUTION MANAGEMENT SYSTEM (DMS) The information system for Distribution Management includes a hardware and software infrastructure for the distribution network operation and control (based on existing SCADA systems) and information exchange equipment between the DMS and other ENELVEN and ENELCO systems. The system comprises the following: 1. A system located at a control center with SCADA and DMS functions sharing the same user interface. 2. A complete model of the distribution network (Distribution System Operation Model, DSOM). 3. A Trouble Call System (TCS) with corresponding interfaces to the DMS and other systems. 4. Interfaces with other ENELVEN/ENELCO systems as indicated in Figure 1. As shown in Figure 1, some applications are part of the DMS including: Outage Management, Switching Order Procedure, Crew Management, Distribution Power Flow, etc. All are supported by a user interface and other support functions together with a Historical Management System. The remainder of this paper is dedicated to the Distribution System Power Flow (DSPF) and its particular characteristics as required for the ENELVEN solution. Distribution Management System Trouble Call System Crew Management Outage Management System Switching Order Management Fault Location, Isolation and Service Restoration Distribution Power Flow A P I s Support Functions User Interface Historical Information System D S O M (Tools + Methods) Database External Systems E R P AM FM / GIS C I S S C A D A Figure 1: DMS Main components and Interfaces As shown in Figure 1, some applications are part of the DMS including: Outage Management, Switching Order Procedure, Crew Management, Distribution Power Flow, etc. All are supported by a user interface and other support functions together with a Historical Management System. The remainder of this paper is dedicated to the Distribution System Power Flow (DSPF) and its particular characteristics as required for the ENELVEN solution. III. ACCURATE DISTRIBUTION POWER FLOW, WHY? a) General Consideration RTUs IEDs Distribution Automation KEM_Cespedes_A1 Session 3 Paper No

3 DSPF is used as a routine for monitoring the distribution feeder loading and voltages. It is also a part of the application known as Volt-Var Control, which task is to enhance the use of the reactive power. This last application is normally among the top of those DMS functions that produce direct benefits to the company showing high benefit to cost ratio. For verifying operation constraints, (currents and voltages) it was not required until recently to have an accurate transformer modeling. The main reason for this is an absence of information on exact values of the loads connected to these transformers. American type distribution systems, to which ENELVEN s is similar, have a large number (hundreds) of relatively small rating single-phase and three-phase transformers (mainly in the kva range) connected to the same feeder. Small distribution transformers normally do not have any measurements. DSPF used mainly for planning, not for real-time, had almost no sources to get additional load information. In addition, currents were checked for overload conditions on the main feeders only and voltages were verified at the primary voltage level side. It is also obvious that an accurate three-phase DSPF requires a large amount of information on the network topology and phase connections, and electrical parameters of the distribution lines, transformers, loads and capacitors. b) ENELVEN Specific Distribution System Characteristics The ENELVEN Distribution System widely uses standard for American networks transformers connected in wye/wye and phase to neutral wire. In addition, ENELVEN also utilizes a variety of different transformer connections including wye grounded/delta and open wye grounded/open delta banked transformers with different capacity sizes in the bank legs, transformers connected in delta/delta, open delta/open delta and delta/wye grounded configurations, as well as single-phase transformers connected phase-to-phase (Table 2). For all these connections, it is absolutely impossible to predict primary transformer side phase loading based on the knowledge of secondary side load, as is possible for the aforementioned phase-to-ground and wye/wye transformer connections [1]. Table 2: ENELVEN/ENELCO transformer Connection types Total Connection Type transformer banks Total installed power 1 Single-ph, % % ph-ground 2 Single-ph, % % ph-phase 3 Wye wye 4.69 % % 4 Wye delta 6.07 % % 5 Open wye 3.05 % 4.50 % open delta 6 Delta Wye 4.95 % 6.90 % 7 Delta 0.96 % 1.42 % Delta 8 Open delta 0.41 % 0.68 % Open delta Total % % The latter statement is illustrated by the example, taken from ENELVEN s distribution feeder, and shown in Tables 3. As presented below, the same transformer bank with the same load connected in wye/delta and wye/wye produces different primary side phase loadings. It also shows the difference in total primary side active and reactive powers, which means a difference in transformer power losses. Table 3. Example of Primary side power flow for 100/167.5/100 kva transformer bank connected in wye/delta and wye/wye with the same secondary side load Ph P kw wye/delt a P kw wye/wy e Q kvar wye/d elta Q kvar wye/w ye Loading % wye/delt a Loading % wye/wy e A (ab) B (bc) C (ca) Total Therefore, the ENELVEN distribution network variety of different transformer connections combined with an aforementioned strategic requirement of maintaining power loss at a minimum level (which is only a few percent of the total load), creates the requirement for accurate DSPF. There is also no choice to base accurate DSPF on a full-scale input data model. Implementation of the Distribution Management System opens the possibility to improve the quality of KEM_Cespedes_A1 Session 3 Paper No

4 DSPF by moving to a Distribution Real-Time Power Flow (DRTPF). DRTPF uses real-time information about network topology and analog measurements at some distribution network key points. In combination with using billing information on customer consumption, it gradually removes the main obstacle on approximate nature of the DSPF. The latter progress in the creation of a corporate GIS database, which can be used for distribution system data preparation, makes the task of input data preparation less laborious through automation. Even under these circumstances however, the decision to use an accurate three-phases unbalanced DSPF for network with a few hundred thousand transformers and close to one million nodes was carefully justified. IV. DSPF INPUT DATA PREPARATION The large volume of DSPF input data makes data import automation absolutely necessary. The input data is coming from two different sources: definition of the network elements (lines, transformers, switches, capacitors) and their connectivity is imported from the corporate GIS database. Electrical parameters (impedances, phases, loads values, and etc.) are imported from a different source. It must be stressed that GIS stores the distribution network by geographical areas, not by electrical connectivity. One geographical area may include a few distribution feeders connected to the different supply sources. This information is fed into the SCADA database where network electrical connectivity is built and checked by tracing functions. SCADA also connects the distribution network to the Supply Substations, which already exist in the ENELVEN Energy Management System (EMS). As soon as all electrical elements (lines, transformers, capacitors, switches) of the same distribution feeder are connected and the feeder topology is established in SCADA, electrical parameters of these elements are imported through a different database. A set of filters is established in order to test the feasibility of each imported value (impedance, admittance, rating and etc.) and to identify possible errors. These two mentioned data flows (topology and electrical parameters) are merged in the DSOM as shown in Figure 2. Real-time measurements, statuses of the switches and tap positions are transmitted to SCADA from the field RTU and are updated in the DSOM. The DSOM serves as a real-time database located in the computer shared memory. DSPF is running continuously as a UNIX process. In addition to all standard power flow features, DSPF provides two additional functions: final checking of the network topology including phase connectivity and serving as a Distribution State Estimator. It checks consistency of the measurements and scales loads according to these measurements. Load values, calculated by DSPF, are written back to DSOM to be used by other DMS network application programs. V. LOAD MODELING As previously mentioned, lack of information on load values is a general problem for distribution systems. Active and reactive power measurements are available for very few ENELVEN loads. For the majority of the distribution transformers, the connected load value can be only estimated based on the expected (designed) transformer peak loading. In several cases, an average load consumption (during 24 hours) can be calculated from the monthly billing data. This data is more reliable than peak load. In combination with typified 24 hours load profiles for the main types of customers KEM_Cespedes_A1 Session 3 Paper No

5 GIS Data Network elements geographical location SCADA Network description and topology Field equipment statuses, measurements Network Elements Electrical Parameters Line/transformer impedances, phases, taps, loads, capacitors Distribution System Operational Model (DSOM) 1) Load active powers are set according to the closest P measurement upstream (feeder head in most cases) as constant and non-dependent of voltage values. Power loss is assumed to a default percentage of the measured P. Load reactive powers are calculated from the active power through individual load power factor. 2) Both load P and Q are scaled iteratively by the power flow until the calculated P and Q are equal to the measured values. Load Modeling D R T P F Distribution Network Applications Figure 2: Structure of Distribution System Operational Model Interaction (residential, commercial, industrial etc.), this information serves as the basis for load modeling. ENELVEN is currently in the process of researching and developing a mathematical tool to calculate transformer loads based on the invoiced (billed) energy. The main source of load data improvement is real-time measurements taken from the supply transformers, feeder heads and occasionally along the feeder. These measurements are used by DSPF for load scaling in such a way that calculated active and reactive powers at the points where measurements are taken are equal to the measured values. The procedure of load adjustment, according to the measured values, is called Load Scaling and serves as a sort of distribution systems state estimator without the complexity of a real estimation function. The ENELVEN distribution system has both active P and reactive Q measurements at almost all feeder heads and supply transformers. The Load Scaling procedure is done in two consecutive steps: once before power flow (step 1) and iteratively inside power flow (step 2); As a result of step 2, initial load power factors are changed. In the case of significant power factor change, a special warning is generated. Very often, wrong shunt capacitors connections (connected through non-telemetered switches) are the reason for the warning, which helps to identify blown capacitor switches. Additional load modeling problems arise from the loads connected to the secondary transformer s side in delta with uneven legs. In spite of the fact that only one three-phase load is normally described in input data, physically there are two loads: three-phase balanced and singlephase, connected to the largest transformer bank. Inside the DRPF, these loads are split into threephase load and single-phase load by using the following equations: TX G = K_1 * LOAD 1 + K_2 * LOAD 3 TX P = K_3 * LOAD 1 + K_4 * LOAD 3 where TX G, TX P are the largest and the smallest transformer ratings in the bank. Coefficients K1-K4, currently used by ENELVEN, are shown in Table 4. They are considered configurable parameters and can be changed in the future. They are different for delta and for open delta connections. Table 4. Load Distribution Coefficients KEM_Cespedes_A1 Session 3 Paper No

6 Connection K_1 K2 K3 K4 Delta 1,000 0,333 0,000 0,333 Open Delta 1,000 0,333 0,000 0,667 VI. DRTPF ALGORITHMICAL SOLUTIONS The Distribution Power Flow model used in DRTPF is based on phase coordinates with single-line circuit presentation. Each circuit element (line, transformer, load, capacitors) is represented as an admittance matrix with size 3*3, 2*2 or a single value. The advantage of this approach is that a standard sparse matrix technique for ordering, factorization and forward/backward substitution can be applied. The Current Injection (CI) method is used for the Power Flow solution. The main idea of the method is described by the following two matrix equations: [V] = [Y-1] *[I] [I] = [S] / [V] where: [V], [I], [S] are vectors of nodal voltages, currents and powers, [Y-1] is a factorized nodal admittance matrix. The CI method is used in applications where the majority of the nodes may be represented as loads and the number of PV buses is limited. The distribution systems satisfy these conditions. Different transformer connections modeling is one of the most challenging requirements for ENELVEN DRTPF. According to the general approach of the circuit element simulation, each transformer type has its nodal admittance matrix [Y_t] included in the general circuit matrix [Y]. The basic steps for computing this matrix are as follows: 1. Build the branch admittance matrix [Y_b] for two or three transformer banks. 2. Build the branch to bus connection matrix [C] for given bank connection type. 3. Compute the nodal admittance matrix for the transformer bank as: [Y_transformer] = [Ct] [Y_b][C] Phase-coordinated power flow methods have a common problem with Wye/Delta transformers. These transformers create energized islands isolated from the ground (common bus) [2]. A special technique is used to factorize nodal admittance matrix bypassing a peculiar matrix problem. VII. TESTING THE DRTPF SOLUTION FEASIBILITY In spite of the fact that DRTPF is solved independently for each distribution subsystem, the size of one subsystem, which includes part of the network supplied from the same injection source (from one up to ten feeders), is very significant (from a few hundred up to a few thousand transformers). That is why the double checking of input data by analyzing the feasibility of the Power Flow results immediately after the data was prepared became extremely important. The following algorithm formalizes the main steps of the engineering analysis used for Power Flow results testing: 1. Solve subsystem with no load scaling and 1.1 Check substation transformer loading. If it is not feasible, check loads. 1.2 Check substation transformer low side voltage, voltage drop, tap increment. 1.3 Check power factors for the injection source and feeder heads. If these values are more than 10% different from the average load power factor (0.87 in most cases), return to the input data and check power factors for the individual loads. 1.4 Check the ratio of the total power loss to the total injected power for the whole subsystem and for each feeder. If any of these ratios is not feasible: Compare total transformer power loss with total line power loss. KEM_Cespedes_A1 Session 3 Paper No

7 1.4.2 If transformer power losses are not feasible, compare transformer no-load losses with transformer load losses. Check transformer impedances or admittances If line power losses are not feasible, check line impedances. 1.5 Select all distribution transformers with loading above and below average. In input data, compare these transformer ratings with connected load nominal powers. 1.6 Select all buses with voltage violations If violated bus is at the secondary transformer side, check transformer voltage drop and tap increment If violated bus is at the primary side, see voltages on the trace upstream from this bus to the injection source. 2. Solve subsystem with P scaling only and no 2.1 Compare injection source and feeder head P, Q flows and power factors to those with no-scaling. Calculate correspondent ratios. If the difference is significant: Check statuses for normally opened and sectionalized switches Check non-conforming load input data and measurements. 3. Solve subsystem with P scaling only but 3.1 Compare injection source and feeder head power factors with those from step Solve subsystem with both P and Q scaling and 4.1 Check individual load power factors. If they are significantly different from the initial factor, check current status of the capacitor switches. approach to implement a fully automated data check preprocessor. The results obtained with the DRTPF are promising both in terms of performance and results that are used in the real-time day-to-day operation at ENELVEN s control center. References 1. W.H. Kersting Distribution System Modeling and Analysis, CRC Press, 2001, 314 p. 2. D. Anderson, B.F. Wollenberg Solving for Three Phase Connectivity Isolated Busbar Voltages Using Phase Component Analysis, IEEE Trans. On Power Systems, Vol. 10, No.1, 1995, pp J. B. Patton, D. T. Rizy, J. S. Lawler, Applications software for modeling distribution automation operations on the Athens Utilities Board, IEEE Transactions on Power Delivery, Vol. 5, No. 2, April W. G. Scott, Automating the restoration of distribution services in major emergencies, IEEE transactions on Power Delivery, Vol. 5, Guidelines for Evaluating Distribution Automation, EPRI Report EL-3728, Nov D. L. Brown, J. W. Skeen, P. Daryani, F. A. Rahimi, Prospects for Distribution Automation at Pacific Gas & Electric Company, IEEE Transactions on Power Delivery, Vol. 6, No 4, October R. Céspedes, L. Mesa, C. Hoyos, Practical Experiences Of The Implementation Of Substation And Distribution Automation At Empresas Publicas De Medellin, Distributech, Miami, Fla., Acknowledgement: The authors wish to thank the ENELVEN management for their continuous support during the entire development of the project which has concluded in a full operation DMS located in Maracaibo, Zulia, Venezuela. VIII. CONCLUSIONS The implementation of an accurate three-phase unbalanced power flow is required by ENELVEN in order to satisfy quality of service requirements and improve operation efficiency. The main reason for this requirement is the extremely unbalanced nature of the distribution circuit due to a wide variety of unbalanced transformer connections and the need to estimate losses at feeder and transformer levels which are small compared with the feeder load. Data preparation for the ENELVEN volume distribution network is extremely laborious and requires a high level of automation. The various proposed steps are a first KEM_Cespedes_A1 Session 3 Paper No