基于神经网络和ARIMA模型的冷热电短期负荷预测

Short-term Forecasting of Cooling,Heating and Power Loads Based on Neural Network and ARIMA Model

冷热电负荷预测对终端供能系统的规划设计有重要意义,针对冷热电负荷预测方法中存在的变量多、时间开销大等问题,以5种典型建筑的冷热电负荷历史数据为基础,将Elman神经网络、自回归求和滑动平均ARIMA(autoregressive integrated moving average)模型和小波神经网络用于冷热电短期负荷预测。仿真结果表明:在冬夏典型日的冷热电负荷预测中,小波神经网络的最大平均绝对百分比误差为2.1%,计算速度适中,是较为合适的冷热电负荷预测方法;ARIMA模型的最大平均绝对百分比误差为4.1%,计算速度慢,但调试和确定参数的难度不大;Elman神经网络的最大平均绝对百分比误差为7.4%,但计算速度最快,网络参数少且调节简捷,适用于对预测精度的要求不太高,但需快速响应的场合。

It is of significance for the planning and design of a terminal energy supply system to forecast the cooling,heating,and power loads.To solve the problems in the combined load forecasting method,such as too many variables and large time cost,Elman neural network,autoregressive integrated moving average(ARIMA)model,and wavelet neural network are used to predict the short-term cooling,heating,and power loads based on the corresponding histori cal data of five typical buildings.Simulation results show that in the load forecasting on typical summer and winter days,the maximum mean absolute percentage error(MAPE)of wavelet neural network is 2.1%and the calculation speed is moderate,indicating that wavelet neural network is a proper load forecasting method.The maximum MAPE of ARIMA model is 4.1%and the calculation speed is slow,but it is not difficult to debug and determine the parameters.The maximum MAPE of Elman neural network is 7.4%,but the calculation speed is the fastest;moreover,the network parameters are fewer and the adjustment is simple,showing that Elman neural network is suitable for the situation where a moderate prediction accuracy but a fast response speed are required.