Stochastic Optimal Dispatching Strategy Of Integrated Energy System Based On Machine Learning

The global energy crisis and environmental issues necessitate the rapid development of renewable energy,and promote the transformation of the energy industry.The integrated energy system(IES)can achieve synergetic interactions among the energy systems to improve energy efficiency,significantly improve the utilization of renewables,and solve the environmental problems.Meanwhile,hydrogen is regarded as one of the most prospective energy carriers,and power to gas(P2G)technology has become a hot spot in the energy field.Moreover,the uncertainties of renewable energy and load bring challenges to the power system dispatching.Therefore,this paper aims to improve the utilization rate of renewable energy,reduce the uncertainties of renewable energy and load,and develop a more economical and stable plan for optimal dispatching in the IES.Stochastic optimal dispatching strategy of electricity-hydrogen-gas-heat integrated energy system(EHGHS)based on machine learning is presented in this paper.First,the structure of EHGHS and the energy conversion unit in the EHGHS is modeled.It analyzes the two-stage P2G technology instead of the traditional P2G technology,hydrogen fuel cell and combined heat and power cogeneration,and the coupling effect of electricity-hydrogen-gas-heat in the EHGHS.The synergetic interactions among electricity-hydrogen,hydrogen-heat and hydrogen-gas are also detailed described.Then since the uncertainties of renewable energy and load in the IES bring challenges to the energy system dispatching,an IES scenario generation method based on VAE-GAN and an improved spectral clustering method based on unsupervised learning in machine learning are presented to describe the uncertain characteristics of renewable energy and load in the EHGHS.The effectiveness and accuracy of the scenario generation model and the scenario reduction model are verified by case studies.The scenario generation method based on VAE-GAN can accurately describe the uncertainties of renewable energy and load in the EHGHS,and the improved spectral clustering method can extract more stable typical renewable energy and load scenarios.Finally,based on the results of the exctrated wind power,photovoltaic,electrical load and thermal load scenarios,an EHGHS stochastic optimal dispatching model with the lowest economic cost is established.The effectiveness,economics,and sensitivity of the stochastic optimal dispatching model are verified by case studies.It also studies the advantages of the EHGHS stochastic optimal dispatching model based on the machine learning scenario analysis method in reducing the overall operating cost,improving the utilization rate of renewable energy,and achieving synergetic interactions among the energy systems to improve the overall energy efficiency.