Short-term Air Conditioning Load Forecasting and Optimal Scheduling Based on RIME Optimized CNN-BiGRU-Attention Model

Power load forecasting and optimal scheduling are the cornerstone to ensure the safety, stability and economic operation of power system, and their outcomes directly affect the plans of power decision-making departments. Traditional modeling and statistical methods ignore the influence of uncontrollable factors such as temperature and humidity environment, equipment conditions, and user habits on air conditioning load prediction. This paper proposes a CNN-BiGRU-Attention short-term air conditioning load forecasting algorithm based on RIME optimization. RIME optimization algorithm is used to optimize the learning rate of CNN-BiGRU-Attention model, the number of BiGRU neurons, the key value of attention mechanism and convolution kernel parameters. Data such as temperature, humidity, electricity price, and air conditioning load are used as input for training. Furthermore, the model is applied to construct a local microgrid including gas turbine, wind power and air conditioning loads with the goal of minimizing the operation cost of the power system. As a flexible load, air conditioning has power adjustable characteristics. Through user compensation and electricity price incentive, it realizes the regulation of electricity price and set temperature, changes the user’s electricity consumption behavior, and achieves the purpose of peak shaving and valley filling, flattening renewable energy output fluctuations and enhancing renewable energy utilization.