Optimizing Power Supply with AI-Predicted Demand for IoT and Machine Devices

Many devices operate on a standard 220 V AC supply but may not always require the full power for theirspecific tasks, leading to energy waste. This paper proposes an AI-based system to predict the minimumpower needed for such devices based on their type and use case, and then supply only that amount tooptimize energy usage. We collected and preprocessed data on device types, use cases, and powerconsumption, then trained several machine learning models to predict the minimum power required. TheGaussian Process Regression model achieved the highest accuracy with an R² score of 0.9583. We discussthe implementation challenges, such as ensuring prediction accuracy and system safety, and outlinepotential energy savings through case studies. This approach promises significant energy efficiencyimprovements and reduced operational costs.