引入电价与负荷相关系数的PSO-BP神经网络电价预测模型

Price forecast by PSO-BP neural network model based on price-load correlation coefficient

为了解决现有BP神经网络电价预测模型中由于非关联输入样本过多而影响学习效率、导致预测精度降低的问题,在分析电价与负荷相关性的基础上,提出了采用电价与负荷相关系数作为判断是否将负荷引入模型条件的新方法,并将相关系数引入PSO-BP神经网络电价预测模型,以降低模型非关联输入样本数,提高预测精度采用我国四川电力市场资料进行仿真计算,证明该方法具有良好的预测效果。

The price forecast accuracy of a BP neural network is often lowered due to the low learning efficiency caused by too many of non-related inputs. This paper develops a new method of using the price-load correlation coefficient as a condition to input the load to the model. The coefficient can be obtained through correlation analysis of the price and it is used in the PSO-BP neural network model to reduce the non-related inputs and improve the forecast accuracy. This new method is verified by a simulation using the data of Sichuan power market.