基于改进的并联灰色神经网络模型在电力需求预测中的应用

Application of Improved Model of Parallel Grey Neural Networks in Power Demand Forecast

为了提高电力需求预测的精度,分析现有人工神经网络和灰色预测方法各自的优缺点,将二者相结合提出一种并联灰色神经网络预测方法。新方法首先采用灰色模型、神经网络分别进行预测,而后给出了一种基于粗糙集理论确定权值的方法对加权系数加以确定,最后对预测结果加以组合作为实际预测值。用上述并联灰色神经网络模型对上海市的电力需求进行预测,模型精度和预测结果比较理想,优于单一预测模型。计算结果表明,该模型用于电力需求预测是有效可行的,适用于中长期需求预测。

Merits and demerits of both present artificial neural networks and grey forecast method are respectively analyzed for the sake of accuracy improvement of power demand forecast. The artificial neural networks and grey forecast method are combined to propose a forecast method of parallel grey neural networks. The new method firstly adopts grey model and neural networks for forecast and afterwards it presents a method for weight value conformation based on rough set theory; finally, it combines the forecast results as the actual forecast value. Power demand in Shanghai is forecasted by the model of parallel grey neural networks, and perfect model accuracy and forecast results are achieved and superior to that of single forecast model. The calculation indicates that the model is feasible for power demand forecast and acceptable for medium and long-term demand forecast.