水电站径流和关联源荷的年时序场景概率预测

Annual Temporal Scenario Probabilistic Prediction of Runoff and Associated Source & Load of Hydropower Station

为在水电消纳优化调度中充分考虑径流和关联源荷功率的时序波动性和随机性的影响,减少弃水事件发生,提出一种水电站径流与关联源荷年时序场景概率预测方法,模拟生成日均径流量和关联源荷功率的年时序典型场景及其发生概率。通过自组织映射神经网络聚类历史旬场景生成旬典型场景,反映丰枯不同水文条件下径流和关联源荷旬内变化的差异性和典型性;基于马尔可夫转移概率、多场景条件概率和历史年“近大远小”相似性原则,构建旬场景随机生成模型,使模拟场景既能准确拟合实际径流和源荷功率年内时序变化的随机性、季节性和条件相关性等统计特征,又能体现年间趋势性变化规律;结合旬间波动量校验对年时序场景蒙特卡罗模拟,并通过k-means场景缩减获得径流和关联源荷的年时序典型场景及其发生概率。实际水电站算例结果表明,该方法具有精度高、适应性强、预测信息全面等优点。

To fully consider the temporal volatility and randomness of runoff and associated source & load in the optimal dispatch of hydropower and reduce spillage water, a probabilistic prediction method of annual scenarios for hydropower runoff and associated source & load was proposed to simulate the typical annual temporal scenarios of average daily runoff and associated source & load and their probability of occurrence. Several typical ten-day scenarios were generated by clustering with a self-organization mapping net(SOM). Then a ten-day scenario simulation model was built based on a Markov-chain probability matrix, a multi-scenario conditional probability matrix, and the similarity principle— "the closer historical year, the larger weight." It ensures that the simulated scenarios accurately fit the statistical characteristics of actual data(randomness, seasonality, and conditional correlation) for intra-year and reflect the trend evolution year-to-year. Combined with the fluctuation checks, annual temporal scenarios were simulated by the Monte Carlo method. Finally, the k-means scenario reduction was used to obtain typical annual temporal scenarios and their probability of occurrence. The results of an actual hydropower example show that the proposed method has the advantages of high accuracy, strong adaptability, and comprehensive prediction information.