Analysis And Rsearch On Inner Mongolia Electric Power System Load Forecasting

Electric power load forecasting is a necessary means for ensuring the secure, reliable and economical operation of power systems. The enhancement of load forecasting technique is directly advantageous for power systems to increase their economic returns and social benefits. This paper presents a load forecasting model based on the simulated annealing neural network algorithm which adopts the global optimization method of simulated annealing (SA) rather than the traditional gradient methods such as steepest descent in the artificial neural network (ANN) methodology. The optimum weighted values of ANN derived by the SA method are able to jump out of the local minimal traps and make the output of ANN a better approximation of the actual load model. The case study on the data of Inner Mongolia electric power system validates the improvement of SA incorporation into the ANN methodology. Compared to the traditional ANN methods, the simulated annealing-based neural network algorithm can raise the precision of neural network learning ability, decrease the training times obviously, and demonstrates a robust adaptability in its performance.