通过对BP神经网络输入负荷值的归一化处理,同时采用Levenberg-Marquardt(LM)算法,建立了一个改进了的BP神经网络,同时用它来对电力系统进行短期负荷预测.LM算法有效地提高了BP神经网络的收敛速度和负荷的预测精度.仿真结果表明,改进了的BP神经网络具有很高的预测精度和较强的适用能力.
By utilizing the normalization for the input load values of BP neural network and adopting Levenberg-Mar- quardt algorithm, this paper established an improved BP neural network and investigated the power system short-term load forecasting. Levenberg-Marquardt algorithm improves the convergence speed and the load forecast accuracy. The simulation re- suits show that the improved BP neural network can offer higher forecast precision and has greater applicability.