Application of Improved Particle Swarm Optimization-Neural Network in Long-Term Load Forecasting

摘要

The improved particle swarm optimization-neural network (IPSO-NN) can be achieved by improving four aspects of the classical particle swarm optimization (CPSO), such as the inertia weight, the learning factor, the variation factor, and objective function. By applying CPSO, the neural network (NN), and IPSO-NN into the long-term power load forecasting problem, the results show that IPSO-NN has not only better global searching ability and higher convergent accuracy than CPSO does but also shorter training time and faster convergent speed than NN does. In feasible running time, IPSO-NN owns the smallest mean error and the acceptable relative error within 3 %. Finally, this paper applies IPSO-NN in the long-term load forecasting of Langfang city from 2010 to 2019.