A data-driven multi-energy economic scheduling method with experiential knowledge bases

The high penetrability of renewable energy and the increasing demand for hydrogen energy pose challenges to system scheduling and the accommodation of photovoltaic and wind power. Therefore, a data-driven multienergy economic scheduling method with experiential knowledge bases is proposed to enhance system operational efficiency and economic performance. Firstly, the Wasserstein generative adversarial network is used to enhance wind power and photovoltaic historical samples. The K-medoids clustering algorithm and & phi;-divergence are applied to achieve scenario reduction and acquire fuzzy sets, eliminating the need for assumptions about their probability distributions. Secondly, a kernel density estimation is used to improve the accuracy of characterizing the probability distribution of energy loads. Based on fuzzy sets, the distributionally robust optimization is used to achieve optimal scheduling in an integrated electricity-hydrogen-heat system. This technique enhances the accommodation capacity of photovoltaic and wind power while reducing operational costs. Finally, a decision-making technique based on an experiential knowledge base is studied, aiming to swiftly match corresponding decision variables by evaluating the similarity of labeled state variables. Simulation results show that the proposed method improves the decision-making speed by about 0.8 times and reduces operating costs and photovoltaic or wind power curtailment by at least 1%. The method can provide fast and efficient day-ahead economic scheduling for digitized integrated energy systems.