Learning Optimization of Power Grid Dispatch Considering Load Aggregator and Uncertainty of Both Source and Load

The aggregator has become an effective way for load-side units to participate in the power grid dispatch, and the uncertainty of both source and load brings a great challenge to the power grid dispatch. So the power grid dispatch optimization problem considering load aggregator and uncertainty of both source and load is studied in this paper. First, a two-stage load optimization model is established based on the load management mechanism of load aggregator. In the first stage, load aggregators decide the load reduction through the game during peak hours. In the second stage, each load aggregator independently dispatchs the loads inside. With this two-stage model, the load curve is optimized, and the peak-valley difference is reduced. Then a power gird dispatch optimization model considering uncertainty of both source and load on dispatching is established with the objective of minimizing the economic cost and wind abandonment. A corresponding learning optimization model based on a Markov decision process is established. At last, an improved particle swarm optimization algorithm and reinforcement learning method are used to solve the strategies of the load-side optimization and source-side dispatching models respectively. Experiment results show that the methods proposed in this paper can effectively reduce the cost of power generation and improve the level of wind power consumption.