Abstract: Applying the grey relational analysis (GRA) and Bayesian model averaging (BMA) method, this paper develops a new method for dispatching the power production of a hydropower station. We first determine decision variables and impact factor sets using GRA and the results of a deterministic optimal dispatch model, and then obtain rules for staged hydropower production dispatching using a multivariate linear regression model, a support vector machine, and a back propagation neural networks . Finally, the rules for monthly power dispatching are derived using BMA to take weighted average of the models’ results. Application in a case study of the Xinanjiang hydropower station shows that our method is more accurate and can achieve an efficiency of hydropower production comparable to that of deterministic optimal dispatch.

Key words: grey relational analysis, Bayesian model averaging, staged dispatching rules, hydropower
station, power production, model uncertainty