10kV OTP, kl Node voltage (p.u.) Time in minutes since 5.

Transcription

1 Coordinated Control of Cascaded Tap Changers in a Radial Distribution Network Mats Larsson Industrial Electrical Engineering & Automation (IEA) Lund Institute of Technology Introduction The purpose of distribution network voltage control is to compensate for distribution variations and disturbances in the feeding transmission network. A number of on tap changers (OLTC) are available to regulate distribution network voltages. OLTC operations cause transients and wear on the OLTCs themselves. Therefore it is desirable to minimize the number of tap operations. Poor OLTC control can also be dangerous from a voltage stability point of view. Node voltage (p.u.) kv OTP, kl Time in minutes since 5. Figure. Field measurement of voltage in the kv substation at stra Tommarp ( TP) during pickup Presently, OLTC control is based entirely on local measurements, with no coordination between dierent voltage levels or branches of the network. Since there are a lot of cascaded tap changers this gives rise to unwanted eects in that tap operations in dierent levels or branches of the networks might counteract one another. Figure shows a eld measurement of the voltage in a kv substation which has three cascaded OLTCs in the network above. Tap operations made at TP are indicated by. We see a number of short spikes due to counteracting tap operations. A desired property of the OLTC control systems is selectivity, which means that only the correct OLTCs should react to given contingencies and variations. In [2] an optimal control based on tap optimization was derived, which provides perfect selectivity but suers from a number of drawbacks. Due to uncertainty in line impedances and dynamics the optimal control cannot be expected to work well in an actual implementation. Fuzzy control has been most successfully used for control problems where the control objectives are dicult to quantify, where there is large plant uncertainty or Mats.Larsson@iea.lth.se

2 TLA JSD ÖTP feeding network 32kV 2kV 5kV 2kV kv Figure 2. The network branch used in the eld test. where one has some heuristic knowledge which will improve control. Another useful benet is that MIMO systems can be handled in a straightforward way. To demonstrate a centralized control with improved selectivity the distribution system branch Tomelilla (TLA), J rrestad (JSD) and stra Tommarp ( TP) in Figure 2 is being km used in a eld test. Fuzzy Rule-Based Control Malmö Lund For the OLTC control problem voltage deviations and time have been considered as process variables. Their membership functions are given in Figure 3. From the gure we see that a voltage deviation of -.3 is considered low to the degree of about.9 and very low to the degree of about.5. By inspection of simulation results from [2] the following heuristics were formulated : Trelleborg S k å n e Ystad Tomelilla Ö. Tommarp. OLTC should control its' voltage as close as possible to the reference value(voltage following). 2. Do not make tap operation if any tap changer higher up in the network is about to make an operation in the same direction (selectivity). 3. Do not make tap operation upwards just before connection of capacitor banks (avoid excessive voltage overshoot). 4. Do not make tap operation downwards just before disconnection of capacitor banks. (avoid excessive voltage undershoot). 5. If voltage deviation is very large, act immediately. The membership functions and the defuzzication thresholds have been chosen, such that the voltage tolerances given by the rst heuristic rule are about the same as for the conventional control. The heuristics (2-4) species modications to the rst rule which softens the constraints given by the rst rule at time when it is useful. The heuristics can be formulated as a set of rules and a fuzzy inference can be used to determine the control surfaces as seen in gure 4 for the tap changers in TLA and JSD. Järrestad

3 .9.9 Grade of membership low high very low very high Grade of membership Soon Connect Soon Disconnect Voltage deviation from setpoint (p.u.) Time of day (hours) Figure 3. for time. Left - Membership functions for voltage deviations. Right - Membership functions.5.5 TLA Control JSD Control TLA Voltage Deviation (p.u.) JSD Voltage Deviation (p.u.).4.2 TLA Voltage Deviation (p.u.) JSD Voltage Deviation (p.u.).4 Figure 4. Control surfaces Simulation Results The fuzzy rule-based control has been tested and tuned in simulations with the patterns described in [2]. Note that the heuristics incorporate coordination with existing capacitor bank control, an aspect that was not considered in the optimal control in [2]. Therefore the fuzzy-rule based control in some cases is better than the optimal one in terms of fewer number of tap operations. Figure 5 shows simulation results. The left gure shows the voltage prole at kv level for one case. We can see that the number of spikes has been reduced with the fuzzy rule-based control, although the voltage deviation is sometimes slightly larger. At the time of the capacitor bank switchings (at and ) the voltage overshoot is smaller for the rule-based control. The right gure shows a comparison of the fuzzy-rule based, the local (conventional) and the optimal control. Implementation LonWorks [] is a tool for for solving distributed control problems. In a LonWorks network, no central control or master-slave architecture is used. Intelligent control devices, called

4 .3.2 Voltage (p.u.) Local, new tuning Fuzzy rule based Time of day Figure 5. Simulation results nodes, communicate with each other using a common protocol. Each node typically has a simple task, like measuring or controlling an actuator. The most important components are : Neuron Chip is the core component of any node. It is a chip that internally consists of three microprocessors, two of which handle the communication with other nodes. The third processor is used to execute the node application program, which operates at3, MHz. There is also an assortment of memory, ROM that stores the operating system and the protocol. EEPROM and RAM for application execution and storage. Transceivers are connecting circuitry for communication on a physical network. Several versions 25,are available for dierent media. No of Tap Operations Adapters for connection of external units llike a PC. 2, The nodes communicate using the LonTalk protocol, which implements the seven layers of the OSI model. The protocol is designed with control applications in mind and is therefore optimized for use with large quantities of small data packets. A number 5,of standard nodes are manufactured by various vendors, and custom nodes can be programmed in Neuron C, a C implementation with an additional real-time kernel. One of the unique features of LonWorks networks are the network variables which are variables that are dened globally. They are stored and changed by the producer node,, but can be connected to or polled by any other node in the system. The necessary communication is done automatically, transparent to the programmer. The neuron chip has a rather limited memory area for user applications. Therefore their tasks has to be relatively 5, simple. More complicated applications can be implemented in a PC and interfaced to the LonWorks network using for example a serial interface. The essential components of the tap changer control system are :, Host Application, a MATLAB application which is running in a PC in the TLA substation. The application Optimal receives ongoing updates Fuzzy of the rulebased measurements from the local network, generates control signals for the OLTCs and determines voltage reference values. Measurements (powertech) are logged every minute. Voltage control can be algorithm Local, new tuning

5 monitored by the application user interface. It communicates with the LonWorks network through a DDE server and a serial interface. DDE Server, handles communication between the LonWorks network and the host application. Communication Nodes, custom LonWorks nodes which package measurements and control signals for transmission by radio. They are used as gateways to a radiolink through a serial interface. These nodes also make per unit-normalization of the various measurements. OLTC Controller Nodes, custom LonWorks nodes for control of the two OLTCs, either by a local control algorithm for use in case of communication faults or remotely controlled from the centralized controller. Generate increase/decrease-pulses to OLTCs and monitor tap positions by counting receipt signals. The controllers are capable of controlling two parallel transformers in master-slave mode. Analog Input Nodes, standard nodes that perform AD conversion and ltering of current signals from measurement transducers (Weidm ller). Each substation has a local TP/78 bus network to which the nodes are connected. Figure 6 shows a schematic of the network in the TLA substation. The nodes communicate within the station using network variables and with other stations through the gateway to the radio-link. Acknowledgment tap +/- tap receipt radio link The author would like to thank our helpful colleagues, especially Curt Lindqvist, Head of System Operation at Sydkraft, and Daniel Karlsson, Senior Specialist at Sydkraft who initiated the project, Prof. Gustaf Olsson who supervises my work at IEA and sterlen Energi for letting me play around with their power system. The piece of work has been funded jointly by Sydkraft AB and ELFORSK through the ELEK- TRA programme. References Twisted pair bus P active power Figure 6. networks OLTC Control Q reactive power Gateway V voltage Outline of the local bus SLTA PC [] Echelon Coporation. Neuron C Programmers Guide Miranda Avenue, Palo Alto, California, [2] M. Larsson and D. Karlsson. Coordinated control of cascaded tap changers in a radial distribution network. In Proceedings of IEEE/KTH Stockholm PowerTech Conference, June 8-22, 995. Publication SPT PS