针对中长期发电优化调度问题周期长、规模大、随机性强等特点,提出一种基于近似动态规划(approximate dynamic programming,ADP)的多阶段优化决策模型.以预测场景表示日前市场煤价、流域来水、风速等随机变量,将远期合约购煤与市场购煤、水库用水等视为ADP框架下的阶段决策.分阶段决策降低了问题的求解规模和难度,提出的燃煤和水库蓄水的值函数近似策略解决了如何优化决策以保持阶段分解后的整体优化特性问题.通过在决策与近似值函数之间的迭代,求解出近似最优决策序列,进而获取发电优化调度方案.某省年度发电计划的计算分析结果表明,所提方法建模简洁,处理随机因素方便,与传统方法相比较,新方法在获得成本期望值偏差小于 0.5%的高质量随机解的同时,计算时间可大幅度下降,具有广泛的应用前景.
With respect to the features in medium/long- term optimal scheduling such as long period, large scale and strong randomness, a multi-stage optimized decision model based on approximate dynamic programming (ADP) was proposed. Prediction scenarios were used to represent uncertainties in day-ahead coal prices, water inflows and wind speeds. Decisions including forward contracts coal, day-ahead coal and water usage were recognized as stage decisions in ADP framework. Staged solution was used to reduce the size and difficulty of the problem. The proposed value function approximation strategy of coal and reservoir storage was used to make decisions and maintain the overall optimization after staged decomposition. Through the iterations between the decisions and the approximate value functions, the approximate decisions sequence was solved to obtain the optimal scheduling. The calculation results of annual scheduling in some provinces show that the proposed method can get the high quality stochastic solution and the deviation of cost expectation of less than 0.5%, and reduce the computation time substantially compared with traditional methods. Meanwhile, it is concise in modeling, convenient in dealing with uncertainties, and has a promising future.