基于粒子群神经网络的热力站供热负荷预测

结合河北省秦皇岛市碧水园热力站的供热实际情况,提出了利用BP神经网络进行热力站供热负荷的预测。为克服标准BP算法收敛速度慢和易于陷入局部最小的问题,提出利用进化算法——粒子群算法进行神经网络初始状态的优化。在此基础上,进一步提出了混合粒子群算法和速度变异粒子群算法两种改进算法提高优化性能。计算结果表明,采用粒子群算法和BP算法相结合的办法,可以明显提高热负荷的预测精度。

Research on heat burden prediction and control of substation based on PSO algorithm

The use of BP neural network(NN)was applied to predict the heat burden of a heat substation in Qinhuangdao City,based on deep analysis of heat supplying characteristic.But their training,usually with back-propagation(BP)algorithm or other gradient algorithms,is often with certain drawbacks,such as,very slow convergence,easily stuck in a local minimum.A newly developed method,particle swarm optimization(PSO)model,was adopted to train perceptrons,and as a result,a PSO-based neural network approach was presented.For improving the predicting results,two improved PSO algorithm were presented also:Velocity Mutation PSO and Hybrid PSO.Both the two approaches were demonstrated to be feasible and effective by predicting heat burden and the identification of the heat exchanger system in substation.