基于深度强化学习的电网机组组合算法

Power Grid Unit Commitment Algorithm Based on Deep Reinforcement Learning

机组组合是电力系统经济调度中的关键问题。机组组合问题属于复杂混合整数非线性优化问题,大多数优化算法在求解速度、求解效率、计算复杂性等方面存在不足。随着人工智能技术的不断发展,智能算法在电力系统决策优化问题上逐渐得到应用,并取得较好效果。本文提出了一种基于深度强化学习的电网机组组合算法:首先构建以发电成本最小为目标函数的机组组合数学模型,并以功率平衡、机组出力限制和最小启停时间约束为约束条件,然后将机组组合问题在强化学习框架下完成描述,并对状态空间、动作空间、奖励函数进行定义,最后采用深度Q网络算法来训练求解机组组合问题模型。经算例分析及测试结果分析,验证了所提方法在机组组合问题求解上的有效性和优越性。

Unit commitment is a key issue in the economic dispatch of power system.For solving this kind of complex mixed-integer nonlinear optimization problem, most optimization algorithms have shortcomings in solving speed, solving efficiency and computational complexity.With the continuous development of artificial intelligence technology, intelligent algorithms have gradually been applied in decision-making optimization problems in power system and achieved better results.This paper proposes a unit commitment algorithm based on deep reinforcement learning.Firstly, a unit commitment mathematical model is constructed with the objective function of minimizing generation cost and the constraints of power balance, generation limit and minimum uptime and downtime.Secondly under the framework of the reinforcement learning, a mathematical description of the unit commitment problem is established.And the corresponding state space, action space and reward function are defined.Finally, the DQN algorithms is applied to train the model to solve the unit commitment problem.The well-trained model verified the effectiveness and superiority of the method proposed in solving the unit commitment problem by analyzing the test results.