Machine learning approaches to the unit commitment problem: Current trends, emerging challenges, and new strategies

Traditional power system operation and control decision-making processes, such as the unit commitment (UC) problem, primarily rely on physical models numerical calculations. With growing scale complexity of modern grids, it becomes more complicated to accurately formulate difficult efficiently solve corresponding UC problems. As a matter fact, plenty historical records well real-time data could provide useful information insights underlying grid. To this end, machine learning methods be valuable help understand relationship performance parameters, reveal rationality behind relationship, finally address problems in efficient accurate way. This article discusses current practices using approaches mixed-integer linear programming based The associated challenges are analyzed, several promising strategies for adopting effectively discussed article. In addition, we will also explore promptly steady-state nonlinear AC flow dynamics differential equations, so that they can integrated into guarantee security dynamic stability operations, compared DC constrained practice. Our studies show learning, model-free methods, is alternative or addition existing model-based methods. result, effective combination model expected derive solutions improve secure economic systems.