Long-term forecasting and analysis of PM2.5, a significant air pollution source, is vital for environmental governance and sustainable development. We evaluated 10 machine learning and deep learning models using PM2.5 concentration data along with environmental variables. Employing explainable AI (XAI) technology facilitated explainability and formed the basis for factor analysis. At a 30-day forecasting horizon, ensemble learning surpassed deep learning in performance, with CatBoost emerging as the top-performing model. For forecasting horizons of 90 and 180 days, Bi-SLTM and Bi-GRU, respectively, exhibited the highest performance. Through an analysis of influencing factors by SHAP, it was observed that PM10 exerted the greatest impact on PM2.5 forecasting. However, this effect was particularly pronounced at higher concentrations of CO. Conversely, at lower CO concentrations, the impact of increased PM10 concentrations on PM2.5 was limited. Hence, it can be inferred that CO plays a pivotal role in driving these effects. Following CO, factors such as “dew point” and “temperature” were identified as influential. These factors exhibited varying levels of linear correlation with PM2.5, with temperature showing a negative correlation, while PM10, CO, and dew point generally demonstrated positive correlations with PM2.5.

In this work, energy time series forecasting competitions from the Schneider Company, the Kaggle Online platform, and the American society ASHRAE were considered. These competitions include power generation and building energy consumption forecasts. The datasets used in these competitions are based on reliable and real sensor records. In addition, exogenous variables are accurately added to the dataset. All of these ensure the richness of the information contained in the dataset, which is crucial for energy management. Therefore, (1) We choose to study forecast models suitable for energy management on these energy datasets; (2) Forecast models including popular algorithm structures such as neural network models and ensemble models. In addition, as an innovation, we introduce the Explainable AI method (SHAP) to explain models with excellent performance indicators, thereby strengthening its trust and transparency; (3) The results show that the performance of the integrated model in these competitions is more stable and efficient, and in the integrated model, the advantages of LightGBM are more obvious; (4) Through the interpretation of SHAP, we found that the lagging characteristics of the building area and target variables are important features.