Abstract:

A temperature and humidity prediction method based on similar days and Optuna-LightGBM is proposed to address the issue of low prediction accuracy of temperature and humidity in event driven intelligent control cabinets,which can’t provide the timely warning of abnormal temperature and humidity inside the cabinets.The similar day algorithm is used to select the suitable model training data sets.A temperature and humidity prediction model is constructed based on LightGBM.Using Optuna the model parameters are optimized.Finally,a threshold calculation method for warning parameters based on the curve inflection point detection is proposed.The temperature and humidity curve characteristics obtained from the prediction model are analyzed to achieve temperature and humidity warning.The experimental results show that the temperature prediction error MAPE of the proposed method is 0.35%,and the humidity prediction error MAPE is 0.73%,which can achieve the accurate prediction of temperature and humidity inside the cabinet and timely warning.

Key words: light gradient boosting machine (LightGBM), Optuna, similar day algorithm, environmental warning, temperature and humidity control system