Short-Term Load Forecasting Of Electrical Power System

Short-term load forecasting is the basis to achieve optimal operation of the power system. It has significant influence for electric power system security, reliability and economy. At present, the national power grid’s development has been increasingly demanding high to load forecasting. Therefore, using the intelligent algorithm of short-term load forecasting power system and improving the precision and stability of the load forecast have very important sense.According to the characteristics of power load, self-organizing fuzzy neural network (SOFNN) has been used in power system short-term load forecasting. This dissertation introduces in detail of the power system short-term load forecasting’s research content, and comprehensively summarizes the load forecasting methods, including the advantages and disadvantages of each method. Through the traditional neural network of the shortcomings of the algorithm, this dissertation proposes a new algorithm, namely the organization fuzzy neural network algorithm (SOFNN).The algorithm’s structure and the parameters learning method are simple, and it has good non-linear approximation ability, good forecast precision and generalization ability and other advantages, and also it can automatically determine the structure of neural model and then derives the model parameters. This dissertation makes load forecasting based on the data of EUNITE race, and pre-processes the data. It using the vertical processing method and horizontal processing method to repair and add the missing data of power load. The predicted results show that the SOFNN algorithm is better than the competition winner algorithm and better than fuzzy BP neural network algorithm as well. SOFNN algorithm using the maximum load average prediction value of a week to fix the date load prediction value of a day, the forecast precision is higher, the algorithm has very good practical value.