Surrogate Model-Based Multi-timescale Stochastic Control of Islanded Cascaded Hydro-Solar System Considering Non-Gaussian Uncertainty

In many mountain areas in Southwest China, cascaded run-of-the-river hydropower plants and photovoltaic plants with local loads are connected by ac network comprise fully renewable microgrids. These microgrids are connected to the external large power systems via long-distance transmission lines in normal operation mode. However, when the transmission line's planned or emergency outage occurs, the microgrids are forced to operate in islanded mode. Considering the spatial-temporal coupling and limited energy storage of the cascaded hydropower plants and the non-Gaussian uncertainty of solar power generation, the frequency control and intra-day energy management, i.e., power-sharing between the cascaded plants, can be challenging. In this paper, a multi-timescale control method is proposed to coordinate the secondary frequency control and intraday energy management. Low-dimensional polynomial models of the expectation and variance of the response of the secondary frequency control over the short timescale are obtained first. Incorporating this model and a dynamic river model, the long timescale energy management problem is formed as chance-constrained programming. The expectational cost function defined over the multiple timescales is minimized by solving this model in a receding-horizon manner. The proposed method is verified by simulations of a real-life hydro-solar system in Sichuan Province, China.