Application Of Adaptive Network-based Fuzzy Inference System For Medium And Long-term Load Forecasting

Medium-term and long-term load forecasting is fundamental to electric power system planning, it is also an important component of the planning. With the aid of accurate load forecasting, the operation of the electric generating set in the power grid can be arranged economically and properly; the safety and stability of power grid can be improved; unnecessary storage capacity can be reduced; proper repair schedule of electric generating set can be made; regular social production and life can be guaranteed. An accurate load forecasting is also helpful to the reduction of generating cost and the improvement of social and economic benefit.On medium-term and long-term load forecasting, this paper put forward the model of medium-term and long-term load forecasting based on the Takagi-Sugeno adaptive Neural-Fuzzy inference system by integrate fuzzy theory with neural network. This model processed the fuzzy information by the technology of neural network, which made the fuzzy regulation and fuzzy membership functions of the fuzzy inference system can be generated automatically by the learning function, so it solved the problem that the regulation must be made by the expert experience artificially, the bottleneck of membership function and the input/output relation obtained by the neural network can not be accepted easily.Based on the data of the indicator of economic development and social power consumption of the AnXiang town, the level of power consumption of AnXiang town is forecasted using the Takagi-Sugeno adaptive Neural-Fuzzy inference system respectively based on lattice construction and clustering construction. After the fuzzification process of the input variables according to their membership functions, the fuzzy regulation and fuzzy membership function revised repeatedly by the learning function, the growth rate of power consumption can be obtained by the fuzzy inference. Therefore, the social power consumption can be forecasted. The example indicates that: the adaptive neural network inference system based on lattice construction need to identify the type of each variable and the number the variables in every input; but in the adaptive neural network inference system based on clustering construction, after the analysis of input/output data, the membership function and fuzzy regulation can be generated automatically; Therefore, In the medium-term and long-term load forecasting, the load forecasting model must be chosen according to the source and the analysis of the data to improve the practicability of the model.