基于CNN和SVR相结合的电力负荷预测分析

Analysis of Power Load Forecasting Based on CNN and SVR

针对电力负荷预测提出了卷积神经网络和支持向量回归相结合的方法。首先将数据预处理成灰度图,作为算法输入数据;然后通过卷积神经网络进行特征提取,将电力负荷的影响因素重新混合,提取更高维的新特征;最后将新特征输入支持向量回归模型进行预测。通过试验对比,该方法实际效果良好。

Aiming at power load forecasting it proposes the method combining CNN and SVR. First step is to preprocess the data into the relevant gray-scale images which are taken as the model input. Then it extracts the new features through CNN. This step can mix the old features to get the new and comprehensive features. The last step is to put these new features into the model of SVR to do the final forecast. Through experimental comparison it shows that this new method can do better on the forecasting of power load than the other methods.