基于聚集度自适应反向学习粒子群算法在水库优化调度中的应用

Application of aggregation degree-based self-adaptive reverse learning particle swarm optimization algorithm tooptimal operation of reservoir

为高效、快速求解水库优化调度问题,提出基于聚集度自适应反向学习粒子群算法。此算法首先采用聚集度策略分析种群的聚散状态,并在此基础上,提出自适应反向学习策略,生成种群中心的反向解参与进化,引导种群改变聚散状态,进一步平衡算法的勘探与开发能力。将基于聚集度自适应反向学习粒子群算法与经典的和最新的高水平粒子群算法进行比较,在所测的基准函数中,本算法在5个基准函数上都取得最优解,验证了其对连续变量函数的优化能力强于所对比算法。在求解水布垭、隔河岩和高坝洲梯级水库优化调度问题上,本算法求得总发电量为86.335 71×10^8 kW·h,求解所需时间为721 ms,相较所对比算法的调度结果,总发电量最大提高了11.860 2×10^8 kW·h,所需计算时间最大降低了21 380 ms,由此验证了基于聚集度自适应反向学习粒子群算法对水库优化问题的可行性。

In order to efficiently and quickly solve the problem of optimal operation of reservoir, an algorithm of aggregation degree-based self-adaptive reverse learning particle swarm optimization is proposed herein.For the algorithm, the gathering and dispersing state of the population is analyzed with aggregation degree strategy at first, and then a self-adaptive reverse learning strategy is put forward on the basis of this for generating the reverse solution of the population center and participating evolution to lead the population to change the gathering and dispersing state for further balancing the exploringand developing capacities of the algorithm.Comparing the algorithmof aggregation degree-based self-adaptive reverse learning particle swarm optimization with the classical and latest high-level particle swarm optimization algorithms, this algorithm obtains the optimal solutions for the five benchmark functions within the measured benchmark functions;from which it is verified that the optimizing capacity of this algorithm for the continuous variable functions is stronger than all the compared algorithms.For solving the problems from the optimized operations of the reservoirs of Shuibuya, Geheyan and Gaobazhou, the total power generation obtained from this algorithm is 86.335 71×10^8 kW·h, with the necessary solving time of 721 ms, which is maximally increased by 11.860 2×10^8 kW·h with the maximum decrease of the time necessary for the calculation of 21 380 ms.Therefore, the feasibility of the algorithm of aggregation degree-based self-adaptive reverse learning particle swarm optimization for optimizing reservoir operation is verified.