Research Of Forecast Method Of Electric Power Load

Electric power load forecast is one of important factor for electric power system works safely, reliably and economically. It has been an main direction in the field of electric automation. In the article, we have relied on theory of Artificial Neural Networks (ANN) and Genetic Algorithms (GA), using the equipment of remote ammeter statistic system, and made a thorough research in the method of electric power short-term forecast. The main research has been finished as follows.On the base of basic theory research of ANN, using the general approach ability of Multi-layer Feedforward Neural Network, we dissertated the theory and method of load forecast with ANN, and built the model at the same time.BP algorithm is the mostly wide-used method in training ANN. But it is an local optimize method in nature, enter into local optimal point easily and training speed slowly. Aiming to these questions, we made use of the global optimal ability of GA, combining the local optimal ability of ANN, composing the GA-ANN. Using GA training the constructs and weight of ANN, We have gain the purpose of increasing optimal speed and raising precision.In order to increase the precision of forecast, The history data needs high precision. In the article, we brought forward the way of filtering the history data with FIR filter, so it can decrease the affection of bad data. In practice, FIR filter wiped off the interrupt point successfully.Applying the ANN model in this article to forecast the load of No 1 oil extraction plant, we gain good result.