基于网格化产汇流的径流式小水电发电负荷预测

Load forecasting of runoff small hydropower generation based on gridded and runoff yield

针对偏远山区径流式小水电发电负荷因来水量不确定性导致预测精度较低的问题,提出一种基于网格化产汇流模型和人工智能的小水电发电负荷预测方法。首先将小水电所在区域进行网格划分,引入径流曲线法(SCS-CN)计算每个网格上的历时降雨产流,再根据单位线原理并结合上游来水量构建小水电区域的网格化产汇流模型。最后,利用卷积神经网络(CNN)对产汇流数据及负荷数据进行高效特征提取,构造双向门控循环单元(BiGRU)网络模型的输入,建立基于降雨产汇流与CNN-BiGRU网络的径流式小水电发电负荷预测模型,通过降低来水量不确定性的影响来提高测试集上预测结果准确率。仿真结果表明,降雨经产汇流模型处理后,小水电发电负荷预测精度提高了7.55%;同时,与单一的GRU网络模型、组合的CNN-GRU网络模型相比,CNN-BiGRU网络模型在测试集上的预测精度分别提高了4.91%、2.39%。综上所述,降雨产汇流模型有效地提高了山区径流式小水电发电负荷的预测精度,为促进富集小水电地区清洁能源的消纳提供了理论依据。

Aiming at the problem of low prediction accuracy of runoff small hydropower generation load in remote mountainous areas due to the uncertainty of water inflow,a small hydropower generation load prediction method based on grid production and confluence model and artificial intelligence is proposed.Firstly,the area where the small hydropower is located is divided into grids,and the runoff curve method(SCS-CN)is introduced to calculate the rainfall and runoff on each grid.Then,the gridded runoff yield and concentration model of the small hydropower station area is constructed according to the unit hydrograph principle and combined with the water inflow from the upstream.Finally,the convolutional neural network(CNN)is used to efficiently extract features from the runoff data and load data,construct the input of the bidirectional gated recurrent unit(BiGRU)network model,and establish a runoff small flow based on rainfall runoff and CNN-BiGRU network.The hydropower generation load prediction model improves the accuracy of the prediction results on the test set by reducing the influence of the uncertainty of water inflow.The simulation results show that the accuracy of power generation load forecasting of small hydropower station is improved by 7.55%after the rainfall is processed by the runoff model;At the same time,compared with the single GRU network model and the combined CNN-GRU network model,the prediction accuracy of the CNN-BiGRU network model on the test set is improved by 4.91%and 2.39%respectively.To sum up,the rainfall yield and confluence model can effectively improve the prediction accuracy of runoff small hydropower generation load in mountainous areas,and provide a theoretical basis for promoting the consumption of clean energy in areas rich in small hydropower.