梯度下降法在水库调度函数优化中的应用

The Application of Gradient Descent Method in Optimizing the Dispatching Function of Reservoirs

调度函数是制定水库中长期运行调度的常用方法,然而现行的通过回归分析法提取出的调度函数拟合精度不佳。针对这一问题,文中建立了以调度期内水库发电量最大为目标函数的优化调度模型,通过线性回归提取各时段水库调度函数系数作为初始值,再采用梯度下降法优化各时段调度函数系数,并在拟合精度差的时段引入流量分级策略,最终获得优化后的水库调度函数。实例计算结果表明,文中采用的方法不进可以提供精度较高的调度函数系数,还保证了优化调度的发电效益,证明了该方法的有效性和实用性。

The dispatching function is a common method for formulating long-term operation scheduling of reservoirs.However,the method used to extract the dispatching function by binary linear regression usually gets poor fitting accuracy.Aiming at this problem,the optimal scheduling model with the maximum generating capacity of the reservoir during the dispatching period was established.The reservoir scheduling function coefficients of each period were extracted as the initial values by linear regression.Then the gradient descent method was used to optimize the scheduling function coefficients of each period.The inflow classification strategy was introduced in the period with poor fitting accuracy,and the optimized reservoir scheduling function was finally obtained.The calculation results showed that the effective and practical method used in the paper provided higher precision scheduling function coefficients and ensured the optimal power generation efficiency of the dispatching.