Heating, Ventilation and Air Conditioning (HVAC) Systems have become prevalent in nearly all commercial buildings and office spaces. HVAC Systems are essentially required to maintain air quality and regulate temperature. However, these equipment experience failure due to poor maintenance strategies and as a result, consume high amounts of energy, causing unexpected downtime subsequently reducing the equipment lifespan. This work proposes an AI-based model to identify failures or faults in HVAC systems that supports predictive maintenance. The proposed system comprises three key functionalities designed to preemptively identify potential failures and highlight anomalies. This research utilizes the HVAC Power dataset sourced from Kaggle. Missing values were handled with a Histogram-based Gradient Boosting Regressor model to ensure data continuity. A Multivariate LSTM-based model is employed to predict active power consumption with high precision and less Mean Squared Error (MSE) of 0.0012. This work also addresses anomaly detection by using the Isolation Forest algorithm, with graphical representation of anomaly detection output. Extreme Gradient Boosting algorithm is applied further to confirm classification performance, that has delivered an F1 Score of 91.41% and an area under the ROC Curve of 0.95. Unlike other such approaches, Explainable AI is integrated to optimize the model by analyzing SHAP plots, which contributes to greater interpretability. This integration results in a reduction in MSE by 0.0045, representing a meaningful refinement of the model’s predictive performance.