基于最优价值网络的储能系统调度优化及仿真

摘要

The value of energy storage systems can be maximized if they store energy when the usable generation capacity is sufficient and release it if the capacity becomes small.The energy market can provide the clear price signals showing the shortage degree of generation capacity.A sensitivity analysis on the market price regarding to the output of storage systems can help consider the effect of storage systems on the market which however turns the optimization problem into a nonlinear type. A novel optimization framework is created using the latest achievement in the artificial intelligence research area and a series of value networks in order to solve the dynamic decision problem.A value network is used to estimate the decision value on each feasible state distributed in a 2D limited-resolution state space of time and storage status.The continuous increase of resolution will lead to an approximation of quasi-optimal dispatch solution.Based on the method the adequacy analysis for a power system consisted of energy storage and diversified types of generation is done in order to not only verify the effectiveness of the method but also demonstrate the simulated results on how the dispatching center reaches a win-win goal for users storage and generation by using energy storage systems to smooth price peaks and decrease wind or solar power curtailment loss.%储能系统只有根据系统可用发电容量的多寡来决定补充或释出能量的时机和规模，才能最大化储能系统的价值。市场机制能提供清晰的电价信号，反映可用发电容量的多寡程度。只有考虑市场价格关于储能出力变化的灵敏度，才能在优化中考虑储能系统对市场的影响，但由此产生非线性优化问题。借鉴人工智能领域的最新成果，建立了针对储能优化这一动态决策问题的全新优化框架，引入价值网络概念，用于评估时间—储能状态2维有界空间中任意路径的决策价值，以指导寻优；通过不断提高搜索精度，来逼近实际市场中储能调度的最优解。将该方法用于多源含储电力系统的调度分析，不仅验证了优化方法的有效性，也给出了调度部门在保证储能适度获益的市场设计下，利用储能容量平抑电价尖峰、充分利用弃风弃光量，使用户、储能、发电三方获益的仿真结果。