DESIGN OF EXPERT SYSTEMS FOR SHORT TERM LOAD FORECASTING AND VOLTAGE CONTROL

• Main Authors: HE, KUN-LONG, 何昆龍
• Other Authors: XU, YUAN-YU
• Format: Others
• Language: zh-TW
• Published: 1990
• Online Access: http://ndltd.ncl.edu.tw/handle/20061989137428176623

博士 === 國立臺灣大學 === 電機工程研究所 === 79 === Short term load forecasting and voltage control for power system dispatch/operation are investigated in the dissertation. The approach used in the work is to combine the experienced operators heuristic rules and mathematical algorithms to develop an expert system for each problem. The research works in the dissertation include (1) SHORT TERM LOAD FORECASTING OF TAIWAN POWER SYSTEM USING A KNOWLEDGE-BASED EXPERT SYSTEM, (2) SHORT TERM LOAD FORECASTING USING A MULTILAYER NEURAL NETWORK WITH AN ADAPTIVE LEARNING ALGORITHM, (3) FUZZY EXPERT SYSTEMS : AN APPLICATION TO SHORT TERM LOAD FORECASTING, (4) VOLTAGE CONTROL USING A COMBINED INTEGER LINEAR PROGRAMMING AND RULE BASED APPROACH. In the work on the application of expert systems to short term load forecasting, the experienced operators heuristic rules and pattern recognition technique are employed for day type identification. With the support of linear regression subroutine, the expert system takes the weather variables into account to conduct peak load and valley load forecasting. The load pattern for each day type and the peak and valley loads are then used to forecast the hourly loads. To further impove the accuracy of peak load and valley load forecasting, multilayer feedforward neural network is developed. A new learning algorithm is proposed to speed up the training process. A fuzzy expert system, which is suitable for on-line operation, is also developed to modify the values of forecasted peak load and valley load by using the most recent data and operators'' judgements. In the work on voltage control, an expert system, which combines, and an integer linear programming routine, is designed to yield proper corrective actions in roder to alleviate system voltage violations. To demonstrate the effectiveness of the proposed approaches, the aforementiond expert systems have been tested by using practical systems Taiwan Power company. It is found from the study are valuable aids to system operators for system dispatch and operation. 本論文針對電力系統運轉調度的所面臨的短期負載預測及電壓控制兩大問題進行深入 的研究。採行的策略為整合資深調度員之經驗法則，與適合之演算法，配合相關之資 料，建立解決各該問題之專家系統。共計完成：用於短期負載預測之法則庫專家系統 ，用於尖載低載預測之類神經網路設計，用於短期負載預測之乏晰專家系統，及用於 電壓控制之專家系統等項。 在短期負載預測方面：法則庫專家系統融合了資深調度員之經驗法則與圖形辨認（P attern Recognition）技術，進行日負載型態辨識。並配合以線性迴歸法（Linear Regression）為主體，考慮天候因素之尖載、低載預測法，可提供次日1 至24時之系 統負載預測類神經網路設計則採用多層類神經網路結構，有效的改進了法則庫專家系 統中之尖類載、低載預測能力。設計中並提出了一套可有效增進求解神經網路速度之 方法。乏晰專家系統採用乏晰理論，可根據最近資料以及調度員之經驗判斷，修正原 負載預測值，適合線上操作使用。 在電壓控制方面，主要融合了調度法則，資深調度員經驗，並應用靈敏度因子及整數 線性規劃法完成專家系統之設計，能有效的改善電壓異常現象。 上述各項專家系統均以台電系統之實際資料進行測試，其結果皆合乎實際需求，故對 電力系統之調度人員而言，為極佳之調度輔助工具。