【摘要】 【目的】准确的径流式水电站出力预测对于拟定发电调度计划、电力保供策略至关重要。针对径流式水电站发电出力随机性强，直接预测精度低等特点，提出一种基于自适应变分模态分解和时间卷积网络(TCN)的组合预测模型。【方法】首先利用鲸鱼群算法(WOA)对变分模态分解(VMD)的参数进行优选，实现原始出力序列的最优自适应分解，然后对分解后的每个分量分别建立TCN模型进行趋势预测，最后将所得结果重构得到最终预测结果。【结果】结果显示：与其他模型相比，所提模型在相同条件下预测效果更优。在非汛期，所提模型决定系数R~2为97.08%、平均相对误差MRE为3.68%、均方根误差RMSE为10.05 MW;在汛期，所提模型决定系数R~2为93.71%、平均相对误差MRE为8.09%、均方根误差RMSE为32.96 MW。【结论】结果表明：(1)WOA-VMD方法能够有效地提取径流式水电站出力序列的特征，降低自身数据的不稳定性对预测结果造成的影响；(2)相比于VMD-TCN、TCN、LSTM、RNN、BP五种模型，所提出的WOA-VMD-TCN预测模型能有效提升水电站出力预测精度，为径流式水电站短期出力预测提供了一种新的、有效的建模思路。更多还原

【Abstract】 [Objective]Accurate power prediction of a runoff hydropower station is crucial for formulating generation scheduling plans and ensuring a reliable power supply strategy. Considering the strong randomness in the generation output of runoff hydropower stations and the low accuracy of direct prediction, a combined prediction model based on Adaptive Variational Mode Decomposition(VMD) and Temporal Convolutional Network(TCN) was proposed.[Methods]Initially, the Whale Optimization Algorithm(WOA) is employed to optimize the parameters of Variational Mode Decomposition(VMD), achieving optimal adaptive decomposition of the original output sequence. Subsequently, TCN model is individually established for trend prediction of each decomposed component. Finally, the obtained result are reconstructed to obtain the final prediction.[Results]The result shows that, compares to other models, the model established has varying degrees of improvement in prediction performance under the same conditions.[Conclusion]The result indicates:(1) The WOA-VMD method can effectively extract the characteristics of the output sequence of a runoff-type hydropower station and reduce the influence of the instability of its own data on the prediction result.(2) Compared to the five models of VMD-TCN, TCN, LSTM, RNN and BP, the proposed WOA-VMD-TCN prediction model can effectively improve the prediction accuracy of hydropower station power, providing a new and effective modeling approach for power prediction of runoff hydropower stations.更多还原