基于异源数据融合与并行计算的电力气象数据分析技术研究

Research on power meteorological data analysis technology based on heterogeneous data fusion and parallel computing

针对短期电力负荷进行预测时易受到不确定气象因素影响而导致预测准确率较低的问题,文中提出了一种基于气象数据融合与并行计算的电力负荷预测算法。该算法将时间和气象因子相结合来对预测的日气象数据进行关联度匹配,进而提升算法的准确率。通过对深度神经网络加以改进,并增加扩张卷积使模型具有更强的视觉野。同时结合Bi-GRU网络,利用其时序特征提取属性进一步增强算法的性能,再将算法部署至Spark并行平台以提高总体的运行效率。实验测试中,所提算法的MAPE、RMSE指标在所有对比算法中均为最优,且领先其他同类算法约0.2%和0.05,而增加运算节点后,算法的运行时间也会相应缩短,表明其具有良好的综合性能。

The short-term power load forecasting is easily affected by uncertain meteorological factors,and the forecasting accuracy is low.In order to improve the prediction accuracy,a power load forecasting algorithm based on meteorological data fusion and parallel computing is proposed.The algorithm improves the accuracy of the algorithm by combining the time factor and meteorological factor to match the correlation degree of the predicted daily meteorological data.At the same time,the depth neural network is improved,and the expansion convolution is added to make the model have a stronger visual field.Combined with Bi-GRU algorithm,the algorithm performance is further enhanced by using its temporal characteristics to extract attributes,and it is deployed to the Spark parallel platform to improve the overall operating efficiency.In the experimental test,the MAPE and RMSE indexes of the proposed algorithm are about 0.2%and 0.05 higher than those of other similar algorithms,which is the best among all comparison algorithms.After adding operation nodes,the running time of the algorithm will be shortened accordingly,indicating that it has good comprehensive performance.