Abstract: In response to the dispatching challenges brought about by the randomness of the output of large-scale clean energy such as wind power and photovoltaic power and the operation constraints of cogeneration units based on heat to determine power generation, a virtual power plant combined heat and power economic dispatching model integrating a high proportion of renewable energy, cogeneration systems, electric-thermal energy storage and demand response mechanisms has been constructed. An improved multi-objective parallel molecular differential evolution algorithm based on the“proximity repulsion”principle of molecular dynamics is proposed, a double-layer parallel computing strategy with a fully asynchronous parallel architecture and dynamic migration of elite individuals is constructed, effectively overcoming the problems of premature convergence and insufficient computational efficiency existing in traditional swarm intelligence algorithms. The simulation results show that compared with the traditional differential evolution algorithm, it can achieve a global convergence success rate of 92.3%, and the computing time is significantly shortened. The constructed dispatching model not only fully consumes wind power and photovoltaic power but also meets the contracted heating demand,achieving a reduction in coal consumption and an increase in net income, which fully validates the practical value of the model and algorithm.

    Key words: virtual power plant (VPP); molecular differential evolution algorithm; wind/photovoltaic power generation simulation; combined thermo-electric dispatch; electric-thermal energy storage; dual-layer parallel computing