Accurate forecast of short-term electrical load is very important to the power system's security and economy. A new model is proposed which based on combining the wavelet transform and neural networks for load forecasting in this thesis. Some forecasting results are obtained for electrical load of Nanjing Area.
By analyzing the electrical load, we find that the load curve shows certain regularity. Then using the Self-Organizing Map network (SOM), the load sequence of one week can be divided into four load types. And by the good time-frequency characteristics of the wavelet transform, the load serial is firstly decomposed to different sub-serials using the Mallat’s pyramidal algorithm. Each sub-serial shows the different frequency characteristics of the load. Different artificial neural networks are constructed to predict each periodical sub-serial according to their characteristics. The network of each sub-serial mainly differs in selection of input variables of the network. To accelerate training neural network and to improve the convergence, an improved LM algorithm is adopted in artificial neural networks are used for each time interval (such as one net for each hour).In addition, the methods of abnormal data processing based on wavelet theory are presented dentally and simulated experimentally.
The results of Nanjing load forecasting show that the WVNN method possesses higher forecasting accuracy and better adaptability than artificial neural network(ANN) forecasting methods which considers day average and day type.For other time series forecasting problems, such as product price forecasting, the international crude oil price forecasting, and so on, the method is also with high reference value and guiding significance.
Keywords: day average temperature, day type, artificial neural network, wavelet transform, short-term load forecasting