【作者】 刘双全；

【导师】 张勇传；

【作者基本信息】 华中科技大学 ， 水利水电工程， 2009， 博士

【摘要】 随着我国电力工业的解制,水电系统从传统的水火电力系统中解放出来,调度目标和约束与传统的水电调度相比发生了很大变化。发电侧调度更多地将注意力投向了提高自身的利益同时兼顾电网的安全、稳定与经济运行。流域梯级电站的逐级开发使得我国的水电规模越来越大,也给水电调度带来了新的挑战,实行流域梯级电站的统一调度,实现梯级的最优开发,提高电网运行效率和效益显得尤为必要。梯级水电系统由于存在着复杂的水力、电力联系以及受径流不确定性的影响,其优化调度是一个极其复杂的优化问题,依靠调度员的经验来制定发电计划的传统调度方法已无法适应大规模水电系统的联合调度。本文结合“湖南沅水流域梯级电站优化调度系统开发工程”对梯级水电系统发电优化调度的模型和算法进行了研究,取得了一些有意义的研究成果,主要体现在以下几个方面:(1)首先引出本文的研究领域,回顾国内外水电优化调度领域的现状和研究进展,并对优化算法在水电系统优化调度中的应用进行了综述,在此基础上确立本文研究的重点。(2)考虑径流的不确定性,将自然径流看作受预报影响的白噪声随机变量,通过延长控制期来消除控制期末水库蓄能对水库运行方式的影响。当控制期足够长时,用于优化计算的约束条件和径流概率分布在各年份中趋于相同,基于此提出了当控制期足够长时水库的最优水位过程会出现周期循环的假设,建立了由年循环模型和实时调度模型组成的随机优化调度的数学模型。年循环模型只被离线求解一次,其结果用作实时调度的边界条件;建立了实时调度模型,其调度期包括了受实时信息影响的过渡期。结合实时信息的影响,水库从每时段初的实际水位出发,保证实时调度期末收敛于年循环最优水位,选取实时调度期第一时段的优化结果来指导水库的实际运行,通过各时段的滚动更新来获得水库运行方式,并给出了模型求解流程。(3)采用分级优化控制的思想,建立了考虑因素较全面的梯级水电系统短期发电优化调度模型。除了考虑常用的水量和电力平衡、库容和流量的上下限以外,还考虑了抽水蓄能电站的扩展,电站的非线性出力特性,水库出库流量变幅,电站出力变幅,电站运行区域的限制以及电站开停机次数和开停机持续时间的限制。对于非线性约束条件和目标函数采用一阶Taylor展开进行线性化处理,基于P分原理通过分解系统耦合出力约束将问题分解到各个梯级,对带上限的约束进行简化以减少约束矩阵的变量个数,求解过程中对人工变量不为零的情况按照预先定义的约束条件破坏优先顺序对不能满足的约束条件进行逐级破坏,确保优化计算总能得到一个有效解。在水库出库和电站出力确定的基础上,考虑电站开停机次数和持续时间约束,采用动态规划求得电站的运行区域计划。并分别给出了混合分解算法求解水电系统调度和动态规划求解电站运行区域计划的详细流程。(4)研究了长、短期优化调度的协调机制,由长期调度结果经线性插值得到短期调度的边界条件后从水库当前实时状态出发,对长期和短期调度进行滚动计算,实现整个梯级最优运行。最后通过对沅水流域梯级电站群联合优化调度的仿真,验证了本文提出的优化算法和建立的模型符合工程实际,在求解梯级水电系统发电优化调度问题上是可行且有效的。更多还原

【Abstract】 With the deregulation of China electric power industry, hydropower system breaks away from the conventional hydrothermal power system, which leads to significant changes of the operation objective and constraints compared to the traditional hydrothermal scheduling. Power companies pay more attention to their own generation interests while meeting the security, stability and economic operation demand of power grid. The progressive development of cascade hydropower stations makes ever larger scale hydropower systems in China, brings about new challenges for hydropower scheduling, and necessitates the unified scheduling to achieve the optimization development of cascade hydropower systems, and to improve the efficiency and benefits of power grid.Optimal scheduling of cascade hydropower systems is a complicated optimization problem affected by the hydraulic and electric coupling and the uncertainties of natural inflow, the conventional operation method based on dispatcher’s experience has no longer been able to fulfill the requirements of co-scheduling of the ever larger scale hydropower systems. Based on the project "Hydropower Optimal Scheduling System Development for Cascade Hydroplants on Yuan River Basin", the optimization model and solution algorithm for solving cascade hydropower system generation scheduling are studied, the research work mainly includes the following aspects:(1) First of all, the research field is introduced by reviewing the domestic and international state-of-the-art and advances in hydro scheduling optimization, the comments on optimization algorithms’ applications to this field are summarized, then the research focus is established.(2) Accounting for uncertainty, the natural inflow is described as a stochastic forecast-dependent white noise process. With the influence of the errors in estimating the end-of-study water value diminished by extending the study horizon long enough, the randomness of the standardized natural inflow reaches a stable status, and the constraints and probability distribution of inflow during each time interval tends to be the same every year, the assumption that the optimal reservoir storage trajectory will repeat annually when the study horizon is long enough is proposed. A stochastic model consists of two sub models, an annually cycling model and a real-time operation model, is established for the long-term operation optimization of hydropower system. The optimal annually cycling trajectories that serve as the boundary limits on the real-time operation model are derived for offline use by solving the annually cycling model only once. A real-time period (RTP) that covers a transition period (TP) influenced by real-time information is structured. The real-time operation model is to determine the reservoir storage trajectories, transferring from the observed water level at the beginning of a RTP to the annually cycling trajectories at the end of the RTP. Only the optimal storage derived at the end of the first time interval is used as the target to operate the reservoir, then the operation scheme is obtained to guide the reservoir operation by rolling computation. And the solution method to the model is also presented.(3) Inspired by the hierarchical optimization control mechanism, a comprehensive model for short-term generation scheduling of hydropower system is established. The involved constraints include the water mass conservation, the power balance, limits on reservoir storage and release, expansion of pumped-storage hydroplant, nonlinear plant-based power generation characteristics, limits on plant-based power and ramp, limits on plant-based discharge, limits on reservoir release ramp, limits on plant-based operational region, limits on plant-based start-up and shutdown frequency and duration. The nonlinear objective and constraints are successively approximated by first order Taylor series expansion, and p-decomposition-based algorithm is used to decompose the original problem into several sub ones by decomposing the coupling power balance, constraints with upper bounds are simplified so as to decrease the variable number in the coefficient matrix, a constraint violation priority is defined to deal with the situation that the artificial variables associated with the constraints are non-zero to ensure a feasible active solution is always derived. When the reservoir release and plant-based power are determined, the plant-based operation region schedule is obtained by solving a recursive Dynamic Programming process constrained by the requirements for the plant-based start-up and shutdown frequency and duration. The specific solution algorithm is described as well.(4) The coordination between long-term and short-term operation is studied. The storage boundary limits of the short-term optimal scheduling are derived from the long-term scheduling results by linear interpolation, then transferring from the observed water level, the coordination is implemented by rolling computation to attain the optimization operation of the cascade hydropower system.Finally, with application to the optimal co-scheduling of cascade hydroplants on Yuan River basin in Hunan, the models and methods proposed in this dissertation are practically verified to be suitable for generation scheduling of cascade hydropower systems, and the numerical results demonstrate the feasibility and effectiveness of the presented optimization models and solution algorithms.更多还原