Based On Fuzzy Neural Network To Power System Short-term Load Forecast

In recent years, load forecasting is increasingly important with the power system market reform deepening. Short term load forecasting is a very important element in power system operation dispatch. It is the power system security and economic operation of the base, and is reasonable arrangements for its start and stops the unit to determine the fuel supply plan, because power trading is so important. This short-term load forecasting of predictive techniques power system's advantages and disadvantages of traditional and intelligent forecasting methods is introduced. Through analysis and comparison, fuzzy neural network algorithm is applied to short term load forecasting in this paper.Fuzzy neural network algorithm using fuzzy logic theory is good at dealing with some of the experiences of uncertain information and neural networks which have stronger ability to learn and can reduce the ambiguity using associative memory advantages. The factors will affect the load of fuzzy information processing. Fuzzy treatment temperature, weather conditions information as input is to create a load forecasting model. The network input and output is easier to capture the nonlinear relationship. This is based on the characteristics of short-term load forecasting, prediction. For the treatment of load characteristics on the basis of the analysis: The first, the history of the load on the power system load data of the sample data is used the method of efficacy rate for changing, and the rate of change in the pseudo-data will be removed. The average rate of change is used to compensate. The conventional data processing formula is used to repair for defect data; Secondly, the use of fuzzy logic theory fuzzes the information of temperature, weather conditions and other factors. It makes the network entered simply. The network input and output is easier to capture the nonlinear relationship between the amounts; Then, the selection of neural network model parameters and initial value is setting in-depth study, and with many trains in a large number of sample data. the design principles of the hidden layer, hidden nodes, the initial weights and learning parameters have been concluded that reasonably determine the structure of the network model and a fuzzy neural network load forecasting model; Finally, it uses SQL Server 2000 database management systems to developed software package based on the short-term load forecasting with high-level programming language Visual C # 2005 fuzzy neural network. It has friendly interface, high degree of program structure and strong transplantation. And using the history of the region and meteorological data with a load of Baoding verified the reasonableness of the package design and effectiveness. It made more perfect prediction results. Comparing with the traditional BP algorithm, it shows that it improves the prediction accuracy and stability of the load for the fuzzy neural network system which is based on short-term load forecasting model, it has important practical value.