Abstract: For the short-term scheduling of cascaded hydropower stations, this paper constructs a network coordination short-term plant scheduling model to consider both peak shaving effect and power generation benefit, along with the factors of their interactive constraints. This new model is based on the water balance equation and takes the factors of unit output limitations, irregular vibration zones, and the backwater effect of tailwater as model constraints. Then, we develop an improved locust visual evolutionary neural network with adaptive parameter tuning to solve the model, through modifying the previous one by incorporating the visual residual mechanism, the half wave rectification mechanism, and local mean filtering. Further, a matrix-based state updating strategy is worked out to achieve the transition of states, resorting to the output of the visual neural network, the position updating strategy of the mountain gazelle optimization algorithm, and a chaotic mapping. Comparative numerical experiments have verified our new neural network has significant advantages in application: at least seven of thirteen benchmark examples, and at least two of three application scenes at the Beipanjiang cascade hydropower station in Guizhou.

Key words: cascaded hydropower, coordination of power plant and power grid, locust visual evolutionary neural network, mountain gazelle optimization, regulating power generation by water supply