论文部分内容阅读
提出了一种改进的多目标粒子群算法,该算法通过引入局部扰动的思想和变异操作来提高多目标粒子群算法局部搜索能力,并用非支配排序遗传算法的思想来进行外部档案的维护。对电网公司购电风险进行了详细分析,建立了以最小化条件风险值CVaR(Conditional Value-at-Risk)和最大化期望收益为目标的购电风险评价模型。该模型弥补了VaR(Valueat Risk)不能反映损失尾部信息的缺陷,可以防范小概率极端风险,降低了电网公司发生灾难性风险的可能,并且无需先验知识。用改进的粒子群算法对购电风险模型进行求解,每次可以求出一组最优解,并且最优解均匀地分布在最优前端上,为决策者的正确决策提供了参考。通过实验,证明了该算法及模型的有效性。
An improved multi-objective particle swarm optimization algorithm is proposed. This algorithm improves the local search ability of multi-objective particle swarm optimization by introducing the idea of local disturbance and mutation operation, and maintains the external file by the idea of non-dominated ranking genetic algorithm. This paper analyzes the power purchase risk of power grid companies in detail and establishes the power purchase risk evaluation model with the objective of minimizing the CVaR and maximizing the expected return. This model can make up for the shortcomings that Value at Risk (VaR) can not reflect the loss tail information, prevent the extreme probability of small probability and reduce the possibility of catastrophic risk for grid companies, and without prior knowledge. An improved particle swarm optimization algorithm is used to solve the purchase risk model. Each time a set of optimal solutions can be obtained, and the optimal solution is evenly distributed on the optimal front-end, which provides a reference for decision-makers’ correct decision-making. Experiments show that the algorithm and the model are effective.