基于小波神经网络的燃气系统短期负荷预测

Short-term gas load forecasting based on wavelet-neural network

在燃气负荷预测中,由于日负荷的不稳定,仅以历史负荷为训练样本得到的人工神经网络难以满足日预测的精度要求。提出一种小波分析与BP神经网络相结合的预测方法。首先,将历史负荷序列进行小波分解成概貌序列和细节序列,并在此基础上利用概貌序列、细节序列,以及指数平滑和温度等多种因素训练BP神经网络,预测出未来燃气的日负荷。最后,对某市燃气负荷进行预测,验证了该方法的可行性与有效性。

In the forecast of gas-fired load, due to the instability of daily load, the historical load for the training samples of artificial neural networks are difficult to meet the requirements on the accuracy of prediction. This paper presents a new prediction method, which combined wavelet analysis with the BP neural network. First, to divide the historical load series into the approximate load series and detailed load series, and on this basis, using approximate sequence, details sequence, the exponential smoothing and temperature to train the BP neural network, in order to predict the future load of gas. Finally, to predict the gas load on the city to verify the feasibility and effectiveness of the method.