近年来，世界各国一直在努力探索和发展清洁能源，以降低化石能源的消耗。尽管风电大规模并网能够促进电力与能源转型，但风电出力的不确定性对电力系统的调度运行和频率安全带来了诸多问题。如何在优化调度过程中考虑电网频率安全要求，避免发生严重的频率失稳事故，引发了研究人员的广泛关注。本文结合不确定性相关理论、优化理论、机器学习方法等专业知识，围绕考虑频率安全的电网与多能源系统的优化调度问题展开深入研究，主要成果如下：提出了一种考虑频率安全的电力系统静态经济调度模型。其中，频率安全由系统遭受大扰动故障后的频率暂态稳定程度来描述，基于时域仿真所得系统故障后的频率暂态曲线，提出了暂态频率安全指数和稳态频率安全指数，以评估不同调度方案下系统发生各种故障后的频率安全性。针对频率紧急控制策略中对负荷与机组切除容量的估计与切除优先级问题，分别通过自适应方法和灵敏度分析来确定。考虑到该优化模型的复杂性，选择了混合粒子群和灰狼算法进行求解。最后，通过对修改的IEEE-39节点测试系统进行仿真求解和对比分析，验证了所提模型的有效性。提出了一种考虑风电不确定性和频率安全的电网动态经济调度模型。其中，基于Copula理论和非参数高斯核密度估计对多风电场的风速相关性进行建模并得到联合概率分布函数，并结合场景分析法建立了多风电场风速的典型场景概率集。针对该优化模型，采用混合粒子群和灰狼算法进行求解。最后，在引入两个风电场后修改的IEEE-39节点测试系统上进行仿真分析，验证了所提模型的有效性。提出了一种考虑电网频率安全的电-气-热耦合系统协同优化调度模型。首先，对电力、天然气、热力系统以及耦合设备进行了建模，建立了电-气-热耦合系统动态经济调度模型，并利用商业求解器Cplex进行求解，在由引入两个风电场后修改的IEEE-39节点电力系统、20节点天然气系统和6节点热力系统耦合而成的测试系统上验证了模型的有效性。接下来，在优化目标中引入频率安全相关惩罚成本并基于深度极限学习机-自编码器对其进行快速准确估计。针对所建优化模型，采用混合粒子群和灰狼算法与商业求解器Cplex进行联合求解。最后，在39-20-6节点电-气-热耦合测试系统上进行了仿真分析，验证了所建模型的有效性。
In recent years, countries around the world have been trying to explore and develop clean energy, aiming to reduce the consumption of fossil energy. Although large-scale grid connection of wind power can promote power and energy transformation, the uncertainty of wind power output brings many problems to power system’s dispatching operation and frequency security. How to consider the frequency security requirements in optimal dispatching to avoid serious frequency instability accidents has aroused the wide attention of researchers. This thesis combines uncertainty related theory, optimization theory and machine learning methods to carry out in-depth research on the optimal dispatching of power system and multi-energy system considering frequency security. The main research achievements of the thesis are as follows:A static economic dispatch model of power system considering frequency security is proposed. Specially, frequency security is described by the frequency transient stability of the system after suffering a large disturbance fault. Based on the frequency transient curve obtained by time domain simulations, the transient frequency security index and steady-state frequency security index are proposed to evaluate the frequency security of a system suffering various faults under different dispatching schemes. In the frequency emergency control strategy adopted in frequency simulation, the capacity of load shedding or generator tripping is estimated by adaptive method, and the priority is determined by sensitivity analysis. The hybrid Particle Swarm Optimization and Gray Wolf Optimization (GWO-PSO) is selected to solve the optimization model. Finally, the simulations and comparative analysis are performed on a modified IEEE 39-bus test system to verify the effectiveness of the proposed dispatching method.A dynamic economic dispatch model of power system considering wind power uncertainty and frequency security is proposed. Specially, based on Copula theory and non-parametric Gaussian kernel density estimation, the wind speed correlation of multiple wind farms is modeled, and a joint CDF is obtained. The typical scenario probability set of wind speed of multiple wind farms is established using scene analysis method. Regarding the complexity of the optimization model, the GWO-PSO algorithm is applied for the solution. Finally, the simulations and comparative analysis are performed on a modified IEEE 39-bus test system with two wind farms to verify the effectiveness of the proposed model. A cooperative optimal dispatch model of electric-gas-heat coupling system considering frequency security of power system is proposed. First, the power system, natural gas system, heat system, and coupling components are modeled, and a dynamic economic dispatch model of the electric-gas-heat coupling system is established, which can be solved by the commercial solver Cplex. The simulations are performed on a test coupling system consisting of a modified IEEE 39-bus power system with two wind farms, a 20-bus natural gas system and a 6-bus heat system. Next, the frequency security related penalty cost is introduced into the optimization objective and is estimated quickly and accurately based on deep extreme learning machine-autoencoder. The GWO-PSO algorithm and the commercial solver Cplex are used to solve the model jointly. Finally, the simulations and comparative analysis are performed on the 39-20-6-bus electric-gas-heat coupling test system to verify the effectiveness of the proposed model.
