高比例可再生能源的强不确定性和低惯性从根本上改变了电力系统运行方式，使系统更频繁地产生复杂的安全稳定问题。大部分安全稳定规则难以解析化表达、难以理解、难以嵌入电力系统优化运行模型。针对上述问题，本文提出了基于稀疏斜决策树的电力系统安全稳定规则提取及优化运行决策理论与方法，实现了电力系统优化运行中经济性与安全性的协同。 本文首先明晰了安全稳定规则的三大基本要求：泛化能力强、易于理解和易于嵌入优化模型。提出了稀疏斜决策树技术及其训练算法，建立了通用的安全稳定规则提取及内嵌优化运行的理论与框架。利用迭代算法和大M法将规则转化为混合整数稀疏线性不等式约束内嵌优化运行，易于通过Cplex等求解器快速求解。 针对复杂的小干扰稳定规则难以嵌入优化运行的问题，提出了基于集成斜决策树的小干扰稳定规则提取及内嵌优化运行的决策方法。引入了小干扰稳定数据集主动学习仿真方法，显著降低了安全稳定仿真计算成本。提出了权重随机森林算法集成稀疏斜决策树以提升规则的泛化性能；提出了核心斜决策树的概念及辨识方法，以简化复杂的小干扰稳定集成规则。 针对电压稳定计算时间长、机理难以理解的问题，提出了基于全局优化斜决策树的电压稳定规则提取及内嵌优化运行的决策方法。全局优化斜决策树以最大化所有叶节点的纯净度及规则可靠性为目标函数。通过随机梯度下降训练参数，提升了稀疏斜决策树的学习性能及可扩展性。提出了计及电压的线性化最优潮流模型，将考虑电压稳定的优化运行模型转化为易于求解的混合整数线性规划模型。 最后，本文基于电力系统精细化运行模拟及安全仿真数据，提出了考虑安全规则的电力系统运行方式分析方法。该方法可以同时从定性和定量的角度深入分析安全稳定规则及可再生能源渗透率对电力系统运行方式的影响。青海和甘肃实际算例表明，提取的安全规则能够显著降低不安全的运行模式，此外，高比例可再生能源电力系统运行方式分散性增强、运行模式增多、时序变化频繁，需要准备足够的灵活性资源、保证充足的安全稳定裕度以应对运行方式的变化。 本文的工作扩展了电力系统安全优化运行及分析理论，为高比例可再生能源电力系统安全优化决策提供了关键模型与有效方法。希望本文的工作能够促进我国可再生能源建设、提升电力系统安全稳定运行水平。

The strong uncertainty and low inertia that stem from high renewable energy generation have fundamentally changed the power system operation mode, making the system more likely to face complex security and stability issues frequently. In addition, most of the security and stability rules are difficult to express analytically or embedded in the power system operation optimization. To this end, this paper proposes security rules extraction and operation optimization theory and method based on the sparse oblique decision tree to balance the economy and security in the power system operation. This paper first summarizes the three requirements of power system security and stability rules, including strong generalization ability, easy to understand, and easy to embed in optimization. Based on sparse oblique decision tree, this paper establishes a general theory and method for security and stability rules extraction and embedding in operation optimization. The big M method is used to directly transform the rules into sparse linear inequality constraints. The obtained security-constrained optimization model is mixed integer linear programming model, which can be quickly solved by commercial solvers. Due to the difficulty to embed small-signal stability rules in optimization, a method based on ensemble oblique decision tree for small-signal stability rules extraction and embedding in optimization is proposed. An active learning method is introduced to reduce the time and calculation cost of small-signal stability simulation. A weighted random forest method is proposed to ensemble sparse oblique decision trees to improve the generalization performance of rules; the concept and identification method of core oblique decision trees are established to simplify complex small-signal stability ensemble rules. Given the difficulty to calculate and understand the voltage stability, a global optimized oblique decision tree is proposed to learn the voltage stability rules. The global optimized oblique decision tree takes the purity index of all leaf nodes and the reliability requirements of the power system as the objective function. The tree is trained by a stochastic gradient descent algorithm to improve its performance and application to large dataset. A linearized optimal power flow model considering voltage and reactive power is proposed, and the operation optimization model considering voltage stability is transformed into mixed-integer linear programming model. On this basis, this paper proposes a data-driven power system operation mode and security analysis method based on the power system operation and security simulation data. From both the qualitative and quantitative views, this method can directly study the impact of renewable energy penetration and the extracted rules on power system operation modes. Empirical studies in Qinghai and Gansu power systems show that the extracted rules can significantly reduce the number of insecure operation patterns. In addition, the space dispersion, time variation, seasonal inconsistency of operation mode, and the number of operation patterns are all increasing under high renewable energy penetration. Thus, more flexible resources need to be prepared for frequent power system operation mode switching. Security and stability constraints need to be considered in operation optimization to reserve sufficient security margins. The work of this paper extends the theory of power system analysis and security-constrained operation optimization, which provides key models and effective methods for security-constrained decision-making under high renewable energy penetration. It is hoped that the work can promote the construction of renewable energy and guide the secure and stable operation of high renewable penetrated power system in our country.