多级串联明渠调水工程多目标水位预测控制模型研究

Research on multi-target water level predictive control model of multi-stage cascaded open channel water diversion project

针对传统的明渠水位预测控制模型无法考虑闸门调控次数限制的问题,本文在以往的预测控制目标中加入了流量调整惩罚量,构造了多目标渠池水位预测控制模型;并采用带有精英排序策略的遗传算法来进行复杂优化问题的求解。将此模型应用于南水北调中线干线工程最后6级渠池的虚拟仿真模型中对模型可靠性进行分析,结果表明在两种测试工况中,本文的多目标预测控制模型相比于传统预测控制模型,能在保持相似的水位控制效果同时使得闸控次数降低43%和52%;而且采用遗传算法求解能考虑闸门死区带来的流量最小变幅约束问题,在需要提前进行流量微调的情况下生成更加合理的调控方案。本文结果也表明,基于水位状态预测模型构造多目标预测控制模型,并采用启发式算法进行优化问题求解,这一思路具有一定可行性。

Aiming at the problem that the traditional open channel water level predictive control model can-not consider the limitation of the number of gate adjustments,this paper adds the flow adjustment penalty to the previous predictive control targets and constructs a multi-target open channel water level predictive control model,then a genetic algorithm with elite sorting strategy is used to solve the complex optimization problem.The model is tested on a simulation model of the last six pools of the South-to-North Water Di-version Middle Route Project to analyze the model reliability.The results show that in the two test condi-tions,compared with the traditional predictive control model,the multi-objective predictive control model in this paper can maintain the similar water level control effect while reducing the gate control times by 43%and 52%;moreover,taking the genetic algorithm as the solution method can consider the minimum flow variation constraint caused by the dead zone of the gate and generate a more reasonable control plan when the flow needs to be fine-tuned in advance.This paper also shows the feasibility of constructing a multi-objective predictive control model based on a state prediction model and using heuristic algorithms to solve the optimization problem.