Analyzing risks involved in smart energy sustainability entails identifying, assessing, and mitigating diverse types of risks that include financial, operational, and environmental factors that can affect the dependability, efficiency, and ecological friendliness of energy systems in smart cities. The study uses a high-frequency dataset from smart meters in Mathura and Bareilly districts in India collected over 2 years from May 2019 to October 2021 which contains millions of data points. To forecast energy consumption patterns and reveal possible risks we used machine learning models like linear regression, random forest, gradient boosting, and extra tree classifier. By using several machine learning algorithms such as multiple linear regression (MLR), classification trees (CTs), random forests (RFs), and support vector machines (SVMs) this paper developed an empirical model to establish an interrelationship between district heating systems investments’ influence on the performance improvement variables for sustainable development goals. Notably, the ensemble learning approach had a remarkable precision rate of 94.69% indicating its importance in forecasting and managing demand for power. Moreover, the findings provide insights that could help policymakers and service providers improve urban energy sustainability and efficiency.