Study On Short Term Load Forecasting Based On Quantum Neural Network

As the basic tool of running and dispatching of power system, Short term load forecasting(STLF) is one of the most important daily contents of management and administration to power enterprise. Accurate STLF is the key to arrange generator maintain plan and energy development strategy. Reasonable and accurate STLF will directly affect the security, economical and power supply quality of the power system.Firstly, we introduce the main features of the short term load forecasting, and the main problem, challenge and factor of the short term load forecasting are also discussed. And then, we detailed descript the remained short term load forecasting methods. Comparative analysis including their advantages and limitations, application scopes and application condition are also presented among the above methods. At last, the basic data processing methods are all introduced which service for the bellow short term load forecasting methods.A short term load forecasting method based on quantum neural network is presented in the paper. Selection of the multi-layer activation function and the updating algorithm for the quantum neural network including are analysis. Based on the above theory, the detail steps for day load forecasting and hour load forecasting are presented. At last, simulation analysis is executed by using the actual grid data. Test results show that the forecasting method based on quantum neural network is better than the method based on RBF, and the forecasting effect of the quantum neural network is much better when temperature data is lack.A STLF method based on quantum genetic algorithm model is presented. The detail coding, genetic, mutation and cross operator algorithm are presented, and the specific implementation steps and principles are also discussed. And then, a novel cross operator method based on nonlinear ranking selection operators is presented which could restrain local optimal solution. At last, the performance of the quantum genetic algorithm model is analysis. Test results show that the forecasting method based on quantum neural network has fast training speed and global optimization ability, so it can be well used in short term load forecasting.