AI Driven Solar Microgrid With Predictive Load Management

Author

Prof Atul Malode, Prathamesh Bhagwat, Harsh Meshram, Bhavik Pipare, Aniket Bhalkar


Abstract

 Such small systems, which are solar-assisted and battery-backed, must account for limited stored energy, variable availability of renewable energy, and loads of varying priority. This paper introduces a prototype of a powerfully simple, AI-assisted solar microgrid for predictive load management, derived from the technical implementation TinyML-Based Weather-Aware Smart Load Management System. The solar microgrid is constituted of an ESP32 embedded controller, a 3S2P battery pack, voltage-divider circuits, a current and power monitor INA226, a temperature sensor DS18B20, Wi-Fi weather-data acquisition, an OLED display, an IP-based web dashboard, a 100W DC-to-AC inverter, and three sets of relays for load control. The AI-assisted load management approach employs a TinyML classifier on ten electrical, thermal, weather, and time features to determine the energy condition of the load as being either HIGH, MEDIUM, LOW, or CRITICAL. Load states can range from Normal, Emergency, and Critical loads, while voltage and temperature rules independent of the model can take precedence. The project report proposes a synthetic dataset of approximately 5,000 samples, demonstrating a validation accuracy of 95.9% and a test accuracy of 96.08%. Due to the synthetic nature of the dataset, the accuracy figure is an approximation of the model's success on the prepared dataset and does not predict field performance. The paper offers an overview of the architecture, technologies, equations, test conditions, and limitations of the prototype.


Keywords

Battery management, embedded artificial intelligence, energy management, ESP32, Internet of Things, microgrid, predictive load management, TinyML.



Full Text:

Download Paper PDF


References


[1] R. Rathinam and S. Dayalan, “Demand Side Load Management by Using Priority Based Load Shedding Algorithm with and without Renewable Energy Generation,” 2019.

[2] N. Maitanova, J.-S. Telle, B. Hanke, M. Grottke, T. Schmidt, K. von Maydell, and C. Agert, “A Machine Learning Approach to Low-Cost Photovoltaic Power Prediction Based on Publicly Available Weather Reports,” 2020.

[3] M.-F. Ng, J. Zhao, Q. Yan, G. J. Conduit, and Z. W. Seh, “Predicting the State of Charge and Health of Batteries Using Data-Driven Machine Learning,” 2020.

[4] X. Sui et al., “A Review of Non-Probabilistic Machine Learning-Based State of Health Estimation Techniques for Lithium-Ion Battery,” 2021.

[5] A. Moin, M. Challenger, A. Badii, and S. Günnemann, “Supporting AI Engineering on the IoT Edge Through Model-Driven TinyML,” 2022.

[6] “Home Energy Management System Based on Genetic Algorithm for Load Scheduling: A Case Study Based on Real Life Consumption Data,” 2023.

[7] “Towards Energy-Aware TinyML on Battery-Less IoT Devices,” 2023.

[8] A. Ajitha et al., “Design and Implementation of Smart Home Energy Management System for Indian Residential Sector,” 2024.

[9] S. Panda et al., “Priority-Based Scheduling in Residential Energy Management Systems Integrated with Renewable Sources Using Adaptive Salp Swarm Algorithm,” 2024.

[10] B. K. Jha, A. Tiwari, R. B. Kuhada, and N. M. Pindoriya, “IoT-Enabled Smart Energy Management Device for Optimal Scheduling of Distributed Energy Resources,” 2024.

[11] Espressif Systems, “ESP32 Series Datasheet and Technical Documentation.”

[12] Analog Devices, “INA226 High-Side or Low-Side Measurement, Bidirectional Current/Power Monitor: Data Sheet.”

[13] Maxim Integrated, “DS18B20 Programmable Resolution 1-Wire Digital Thermometer: Data Sheet.”

[14] Edge Impulse, “Embedded Machine Learning Documentation.”

[15] OpenWeather, “Weather API Documentation.”

Share your valuable work from Social Media Buttons