文化微粒群神经网络在用电量预测中的应用

Application of Cultural Particle Swarm Optimization Neural Network in Electric Load Forecasting

为了提高传统神经网络在中长期用电量负荷预测中的速度和预测精度,将文化算法、微粒群算法融入神经网络中,设计了文化微粒神经网络模型;将该模型用于我国某地区中长期用电量预测建模,采用了滚动时间窗技术处理输入输出数据,进一步优化模型数据输入量。该方法综合了微粒群算法的全局寻优能力和文化算法的演化优势。通过与传统的灰色预测模型以及实际数据对比,结果表明,结合滚动时间窗技术的文化微粒神经网络模型用于地区中长期用电量预测建模效果更佳,预测结果更能满足实际要求。

In order to improve the speed and forecasting precision of traditional neural network （NN）, a cultural particle swarm optimization neural network （CPSONN） was proposed by integrating culture algorithm （CA） and particle swarm optimization algorithm （PSO） into NN. The proposed model was used to construct a mid die-long-term electricity load forecasting model in an area of China. To further optimize the model data input, a rolling time window technique is used to deal with input and output data at the same time. This method synthesizes the global optimization ability of PSO and the evolutionary advantage of CA. Comparing with grey forecasting model and the actual field data, results show that the CPSONN with rolling time window technique is more effective for middle long-term load forecasting method in this region.