Study On Power System Short-term Load Forecasting

With the development of China’s power industry,engineering technology is increasingly being implemented to the automatic control of power system.Power system load forecasting as one of the key technologies of automatic control,has immediate impact on reliability and economy of power grid.The short-term load forecasting determines the schedule of the power output.Forecasting error plays a significant role on generation enterprises’ economic benefit and entire social benefit.Therefore,load forecasting research is very important to power grid transmission.Firstly,this paper introduces the research progress of load forecasting,analyses and compares of the various methods of load forecasting technical characteristics.Secondly,specific features of the modeling and artificial neural network training methods,especially feed-forward neural network construction,the learning process data pre-processing and network model and genetic algorithm with BP neural network connection weights are optimized.This approach not only effectively overcomes the impact of complexity and data completion,but also enhances the optimization of the global search capability,while avoiding falling into "premature" convergence defects.And it further optimizes the simulation model,reduces the absolute value of the prediction error,and improves the prediction accuracy.In this paper,a tool box in MATLAB,named as Neural Network Toolbox has been utilized for programming.The load simulation has been conducted for a specific grid structure.Then to compare the output and the observed value from the next day,and then to validate the feasibility and accuracy so that can propose direction of adjustment.