Power System Stabilizer Based on Adaptive Control Techniques

84 WM 018-8

Display and Storage of Data The handling of data for efficient display and storage in a dedicated turbine-generator supervisory system requires con¬ sideration of several factors. 1 ) Hundreds of turbine-generator parameters should be avail¬ able to the operator on demand without using an inordi¬ nate amount of space in the control room. 2) Measured and calculated data must be reliably and accu¬ rately transmitted from the turbine-generator and boiler feedpump turbine to the master controller. 3) The several hundred parameters to be monitored require that care be taken to minimize the amount of wiring

required. provide the

4) Reliable and readily accessible media means

required to for storing and recovering large are

August 1984, pp. 1983-1989

System Stabilizer Based on Adaptive Control Techniques Power

A. Ghosh and G. Led wich University of Queensland, Brisbane, Australia O. P. Malik and G. S. Hope University of Calgary

Calgary, Alta., Canada

amounts of historical data.

Adaptive control can be described as the changing of microcomputer system, as controller parameters based on the changes in system a method of meeting these operating point. The application of adaptive control tech¬ requirements. niques to the excitation control of a synchronous alternator A single CRT screen is used with a modern operator looks very attractive because both the configuration and console to present the various display pages. This reduces operating point of a power system change often. A brief the space required in the control room. The use of both overview of adaptive controllers suitable for excitation con¬ magnetic disk and tape for storage affords quick access to trol along with their merits and demerits is given in this paper. historical data via the disk and long-term storage on magnetic The paper is divided into three parts. In the first part the tape. Also, the magnetic medium affords the needed density adaptive controllers are presented. Two different categories for storing large amounts of data. of adaptive control algorithms, i.e. with implicit identification Finally, the concept of using several microcomputers with and with explicit identification, are discussed. Model refer¬ PROM based memory allows complete debugging and docu¬ ence adaptive controllers, which have been used previously mentation prior to shipping from the factory while maintain¬ for excitation control, fall under algorithms with implicit identification. In this type of controller a reference model is ing the ability to make changes in the field if required. The turbine-generator supervisory system provided by the driven by the same input as the actual model, and the purpose turbine-generator manufacturer who is familiar with power of the control is to minimize the difference between actual generating equipment operating requirements provides a model output and reference model output. The difficulty with modern, reliable, and efficient means of monitoring and this algorithm is the choice of reference model. In a nonrecording data. The overall availability of valuable power deterministic system like power system, a reference model cannot be chosen easily, and a bad choice of reference model generating equipment is greatly enhanced with its use. might cause undesirable system performance. In the algorithms with explicit identification, the system transfer function is first identified. To do this recursive least squares (RLS) identification technique might be used. As this identification technique is recursive, the system output is first sampled at chosen discrete intervals of time. Then based on the identified parameters, a controller is designed to give the BEARING DATA desired system performance. The first of such controllers JOURNAL BEARING VIBRATION ANO TEMP A PROM based distributed described in this paper, afford

VIBRATION IN MILS

^HXHxHKKh TEMPERATURE IN *F

BRG INLET 115 »F

THRUST BEARING PLATE TEMP *F

H R RI R TURBINE SIDE

Fig.

IEEE Power

1.

Bearing

GENERATOR SIDE

data

display page.

Engineering Review, August 1984

described in the paper is a minimum variance (M.V.) control¬ ler. This type of controller minimizes the variance of the output by first predicting the next measurement for zero control, and then choosing control value to bring output error to zero. The M.V. controller has two properties, (i) the closed loop system will be unstable if the dynamics of the sampled system are non-minimum phase, i.e., if the system has a zero on or outside the unit circle, and (ii) the saturation of controller will occur when the control input is band limited. These properties make this technique difficult to use for excitation control problem. The second technique described is a linear quadratic optimal (L.Q.) controller. This controller minimizes a perform¬ ance index of linear quadratic form, and has the advantage of always producing stable control provided the parameter estimates are exact. However, this controller will incur a heavy computational burden. The third technique described is pole-assigned (P.A.) controller, which places the closed-loop poles governing the closed-loop response of the system, at predescribed locations. This controller is similar to the model reference controller because of the lack of guidance in choosing the new locations for the closed-loop poles. In a non-deterministic system the choice may turn out to be a poor choice. A modified form of the pole-assigned controlled, and named the pole shifting (PS) controller, is described in the paper. In this algorithm the open-loop poles are radially shifted towards the origin to obtain the closed-loop poles. This controller has the desirable property of not having to choose two parameters to obtain the desired performance.

33

In the second part of the paper the excitation control scheme is described. In this scheme the adaptive controller is used as a stabilizer, and the control output is then used to supplement the output of an analog AVR. The voltage regulator maintains the control under steady-state condi¬ tions, while the adaptive stabilizer damps out the transients. Studies and results are presented in the last part. The two most promising adaptive controllers, namely linear quadratic and pole shifting, are used in separate studies as a stabilizer. Results of a comparative study of the adaptive stabilizers, using electrical power signal as input, and a fixed parameter conventional stabilizer using accelerating power as input, show the improvement obtained with the use of the adaptive controllers. Discusser: W. Watson

The discussion and case study demonstrate the need for flexible planning, i.e., a planning process which is robust (the disastrous economic and reliability impacts of long-term uncertainties are minimal). A new arrangement called options would add flexibility to the scheduling of large thermal resources. These resources require a long time from inception to completion. An option would allow the resource to move through the time consuming but relatively inexpensive siting and design stages and to be placed in a ready condition. In that condition, the project could be scheduled, i.e., placed on hold or constructed, depending on the demand for electricity. Such options would become, in effect, an insurance policy that would allow the region to plan for high growth rates without prematurely committing to build to those rates. A planning approach is presented which is flexible and ensures readiness of the region to meet any of the possible future load growths. A set of computer models, i.e., load forecasting models, an optimization-based screening model for selection of new resources, and a detailed simulation model; along with heuristics and judgements from the basis for flexible long-term generation planning. Discusser: N. Savage

84 WM 010-5

August 1984, pp. 1990-1996

Planning Approach Recognizing Long Term Load Growth Uncertainty A Flexible Generation

Zia A. Yamayee, Senior Member, IEEE Clarkson College of Technology, Potsdam, NY Hossein Hakimmashhadi, Member, IEEE Gonzaga University, Spokane, WA

Planning electric energy resources for the future is compli¬ cated by uncertainties in future load, hydro conditions (in the pre-dominantly hydro system of the Pacific Northwest) and uncertainties associated with existing and planned resources. In addition to numerous uncertainties, there are unquantifiable costs and benefits associated with different technologies, e.g., environmental impacts. Political considerations cannot be ignored in planning decision process. Long-term load growth rate uncertainty which is a function of economic and demographic conditions, and electricty prices of the future is extremely important. For example, the difference between 20-year low and high load growth rates is expected to be about 11000 avg. MW, in a 15000 avg. MW system of the Pacific Northwest. The actual load growth is expected to be between low and high. If we plan for low load growth and high load materializes the shortage of resources is evident. On the other hand, planning for high load could result in over-building. This paper addresses the issue of long-term load growth uncertainty in long-range generation planning, and how to deal with it. A historical review of the cost impacts of over-building and a case study which estimates the cost and reliability implications of under-building are presented. In the early seventies, the expected load forecasts indicated insufficient resources as early as July 1, 1983. Hence, Pacific Northwest Utilities sponsored construction of numerous large thermal power plants. July 1, 1983, with a number of plants terminated and/or delayed, the region is expected to have firm surplus until the early nineties. This notion of over-building has cost the region billions of dollars. To assess the adverse impacts of under-building a hypotheti¬ cal example is presented. It is assumed that the load growth in the next 20-year period is high (2.5 percent per year), but the region implements the plan developed for the medium-low (1.5 percent per year) load. The results show the region's expected cost would go up by almost six billion dollars and the system reliability would degrade substantially. 34

84 WM 195-4

August 1984, pp. 1997-2005

Extending Service Life of Installed

15-35 KV Extruded Dielectric Cables

G. S. Eager, Jr., Fellow, IEEE, C Katz, Senior Member, IEEE, and B. Fryszczyn Cable Technology Laboratories, Inc., New Brunswick, NJ F. E. Fischer, Member, IEEE, and E. Thalmann, Senior Member, IEEE Orange & Rockland Utilities, Inc., Pearl River, NY Substantial quantities of 1 5 to 35 kV underground power cable utilizing extruded dielectric insulation installed within the last 20 years have suffered excessive electrical service failures. The main cause for the failures has been found to be the formation of electrochemical trees in the insulation. It is now generally accepted that the permeation of conducting liquids, principally water, in combination with voltage stress causes electrochemical trees to develop. The effect of these trees is to reduce the voltage breakdown strength of the cable to a point where occasional switching and lightning surges, and in some cases, normal service voltage cannot be sustained, thereby reducing the life of the cable. Treatment and subsequent testing of 1 5 kV cables having electrochemi¬ cal trees demonstrate that dielectric strength can be signifi¬ cantly restored by (1) continuous flushing of the conductors with dry nitrogen and (2) by filling the conductor with selected liquid such as acetophenone and impregnating the cable insulation. When dry nitrogen is passed through the conductor of aged cables while they are maintained in water under voltage, drying of the insulation occurs from the inside, with resultant increase in voltage breakdown strength. Typical test results on 15 kV polyethylene insulated cable aged in the laboratory, having substantial electrochemical trees after aging and later flushed with dry nitrogen are given in Table 1.

IEEE Power

Engineering Review, August 1984