基于脉冲神经网络的短期负荷预测模型

Short-term Load Forecasting Model of Power System Based on Spiking Neural Network

针对电力负荷的非线性时间序列预测问题,本文构造了一种基于脉冲神经网络的短期负荷预测模型。该模型基于改进的脉冲神经元模型构建了多层脉冲神经网络,并通过脉冲时刻误差反传算法进行学习。模型借鉴了生物大脑基于脉冲放电的处理时空信息机制,将时间序列信息编码成脉冲序列,利用脉冲神经网络挖掘和处理多时间尺度时空放电模式的能力进行连续的负荷预测,最后通过解码神经元膜电压输出负荷估计值。为验证预测模型的有效性,本文基于某地区的实际电网负荷数据进行了预测实验,结果表明脉冲神经网络模型相对其他模型具有更高的准确性和稳定性。该研究为电力负荷预测的发展提供了一种新的可能途径。

To forecast the non-linear time series of power load,a short-term load forecasting model of power system based on spiking neural network(SNN)is constructed in this paper.Based on an improved spiking neuron model,a multi-layer SNN is built,which is further trained by a spiking time error back propagation algorithm.By imitating the brain’s spatio-temporal information processing mechanism through spiking activities,this forecasting model transforms the time series into spike sequences,and exploits the SNN’s capability in mining and processing the multi-time-scale spatio-temporal spiking pattern,thereby predicting the power load continuously.Finally,it outputs the estimated value of load through decoding the neuron’s membrane potential.To verify the effectiveness of the forecasting model,a pre⁃diction experiment is conducted based on the actual grid load data in one certain region,and results show that it has higher prediction accuracy and better stability compared with other models.The research in this paper provides a novel possible approach for the load forecasting of power system.