导叶关闭规律非线性评价函数及多工况优化

Non-linear evaluating function for wicket-gate closing rate optimization and multi-mode optimum method

基于多目标优化基本理论,建立了一种针对导叶关闭规律优化问题的非线性评价函数,并根据导叶关闭规律优化的特点,采用遗传算法建立了导叶关闭规律的多工况优化模型.运用建立的评价函数与优化模型,对某实际电站进行了导叶关闭规律优化计算,并与采用线性评价函数的优化结果进行了比较分析.采用线性评价函数优化的导叶关闭规律,机组最大速率上升值为49.4%,与控制值50%接近,安全余量仅为0.6%,但蜗壳末端最大压力头和尾水管进口最小压力头的安全余量分别为4.717和2.812 m,相对较大;采用非线性评价函数,不需要试算权系数,仅需要给定各优化目标的控制要求即可得到优化的导叶关闭规律,虽然蜗壳末端最大压力头和尾水管进口最小压力头的安全余量分别减小到2.501和1.765 m,但机组最大速率上升值为46.1%,安全余量提高到3.9%,各目标的安全余量分配比较平均.基于非线性评价函数的多工况导叶关闭规律优化得到的结果虽然对单个工况不一定是最优解,但能保证所有工况的各优化目标均能满足控制要求且安全余量分配比较平均,比单工况优化结果更符合实际控制要求.

A non-linear evaluating function dealing with wicket-gate closing rate optimization was pre- sented, and the multi-operation point optimization method is proposed based on genetic algorithm. The wicket-gate closing rate of a real hydropower station was optimized by utilizing the proposed function and optimization model and the results were compared with those optimized by means of the existing linear evaluating function. For the optimized closing rate by using the linear evaluation function, the maximum speed rise is 49.4% , which is near the control value （50%） , and the safe margin is 0.6% only; but the safe margins of both the maximum pressure at the end of spiral casing and the minimum pressure at the inlet of draft tube are 4. 717 and 2. 812 m, which are relatively high. For the closing rate optimized with the proposed non-linear evaluation function, however, although the safe margins of both the maximum pressure at the end of spiral casing and the minimum one at the inlet of draft tube drop to 2. 501 and 1. 765 m, the maximum speed rise is 46.1% and the safe margin rises to 3.9%. The simulated results show that the optimized wicket-gate closing rate with the just proposed non-linear evaluating function can keep safety margin in a proper range for every optimization objective. Multi-operation point optimization method is proven to be aeeurate and the optimized closing rate is universal to different hydro-transient eases. Such a method plus the proposed non-linear evaluating function ean aehieve a relatively equalized distribution of safety margin for every optimization objeetive in different cases.