As the proportion of renewable energy power generation and the proportion of participation in market transactions continue to expand, the stochastic and volatile nature of renewable energy power generation makes it difficult for the current power coordination and balancing method to be applied to a high proportion of renewable energy power systems. This paper proposes a coordinated power balancing strategy for renewable energy systems based on spatio-temporal resource prediction, using ARIMA-LSTM model to predict renewable energy system loads and scenery resources, constructing a coordinated power balancing model, and solving it by using mixed integer programming method. The results show that the proposed ARIMA-LSTM model is not only compatible with the prediction of multivariate loads such as cold, heat, gas, electricity, etc., but also can be used for the prediction of wind speed, radiant illuminance, etc., and the MAPE error values in the prediction time period of one week are 6.05%, 1.88%, and 9.50%, which are lower than that of the models of ARIMA, LSTM, and Elman, and have better adaptability and prediction accuracy. The proposed stochastic power coordinated balancing model and algorithm are verified to be correct and effective by simulating and analyzing a high hydropower proportion system as an example.
