关键词: 空调器, 节能率估计, 自适应, 负荷估计, 长短时记忆网络

Abstract: At present, there is a lack of an intuitive method to quantify energy-saving effects. Although a few air conditioners can predict the amount of electricity saved, they mainly rely on laboratory data for statistics and estimation, which makes it difficult to dynamically adjust according to actual scenarios. To address this issue, a method for estimating the energy-saving rate of air conditioners based on a combination of load estimation and deep learning is proposed. Based on the principles of energy conversion and heat transfer in air conditioning systems, this method calculates the cooling capacity and load of the energy-saving algorithm. It uses Long Short-Term Memory (LSTM) networks to predict system parameters under normal mode, thereby estimating power consumption. The adaptive module, in combination with the Multilayer Perceptron (MLP) deep learning algorithm, can optimize parameters based on historical data, further enhancing the accuracy of the prediction results. Experimental results show that within a four-hour operation period, the deviation between the predicted electricity savings by this algorithm and the actual savings is kept within 8%. In addition, the effectiveness of each module is verified through ablation experiments.

Key words: Air conditioner, Energy-saving rate estimation, Adaptation, Load estimation, Long Short-Term Memory