Applied Cluster Rule NARMAX Method to Short Term Loading Forecasting

• Main Authors: Ching-Yi Lee, 李清義
• Other Authors: none
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/hd7xed

碩士 === 逢甲大學 === 自動控制工程所 === 91 === In general, short-term load forecasting (several hours to several weeks) is critical when applying to areas such as load management, economical dispatch, generation production setting, generation schedule, and pollution control. Hence, precise short-term load forecasting may help power planners in evaluating generation production cost and providing referable indication of the reliability of power supply which may avert a power control crisis and the waste of power energy. Among the existing methods of short-term load forecasting, stochastic time series has been most extensively used. Although capable of accurately predicting results, this method requires that a dispatcher has knowledge of complicated statistics when identifying the factorial number of model. By doing so, a precise load model can be developed. On the other hand, gradient search (e.g. Least Square Estimation, LSE) is adopted to estimate parameters. While easily influenced by the initial given solution, this method can direct the search for a local optimum solution, subsequently causing forecasting error. Also, during the forecasting, whether the forecasting value will overestimate or underestimate the real value can not be estimated beforehand. Hence, the reliability of this method is unsatisfactory. In light of above developments, this study applies the cluster rule NARMAX method with the Neural Networks for optimization , we can use fewer order terms and more combination terms to capture the dynamics of highly non-linear system. By doing so, the optimum solution of factorial number and parameter of the forecasting model can be searched for. Therefore, the NARMAX method can theoretically find out the optimum solution and improve the forecasting reliability when using the time series approach.