Study on hierarchical model of hydroelectric unit commitment based on similarity schedule and quadratic optimization approach

The short-term hydro scheduling (STHS) is mainly concerned with the accuracy and efficiency. Hydraulic unit commitment (HUC), the most important part of STHS, requires significant computational resources. Hence, we introduced a similarity search algorithm based on unsupervised machine learning to initialize the schedules of the HUC problem to reduce the solving difficulty. This paper aimed to minimize water consumption during the dispatch period. First, the dynamic time regularization (DTW) algorithm was used to measure the similarity of the historical load data and screen out the reasonable unit commitment schedules to be combined with the stochastic solutions. Subsequently, a hierarchical model was constructed for quadratic optimization. In the outer layer, the dual-population particle swarm optimization based on similarity results optimized on and off status of the unit, while in the inner layer, the DP was used to distribute the load. Moreover, the elite search strategy narrowed population quality differences. The results show that: (1) the model can improve economic benefits and ensure unit stability; (2) historical commitment solutions learned by the similarity algorithm exhibit constraint violations can be mitigated through secondary optimization, further optimizes the solution space; (3) ML algorithm can enhance HUC performance, especially for large-scale problems. •HUC scheduling is difficult to solve.•Introducing a similarity algorithm based on unsupervised machine learning to initialize schedules is an effective approach.•A quadratic optimization algorithm is constructed to resolve constraint violations and optimize the solution space.•A dual-population particle swarm optimization strategy is proposed to improve algorithm performance.•The proposed model enhances HUC performance, especially for large-scale problems.