IoT based Solar Power Forecasting cum Intelligent Scheduling Algorithm for Smart Microgrid System

Abstract

The efficient operation of solar-powered microgrids depends on accurate forecasting and optimal scheduling to ensure reliability, cost-efficiency, and seamless integration of the grid. This study presents an integrated approach for the optimal scheduling of a solar PV microgrid with a Battery Energy Storage System (BESS), utilizing machine learning and optimization techniques within an IoT framework. The forecasting module employs a two-stage preprocessing pipeline—Kaufman Adaptive Moving Average (KAMA) and Infinite Impulse Response (IIR) filter—for smoothening solar and load data, followed by a Bidirectional Long Short-Term Memory (BiLSTM) network for accurate prediction of solar generation and load demand. These forecasts inform a Mixed-Integer Linear Programming (MILP) model, which optimally schedules the PV and BESS operations to minimize costs while maintaining energy balance and system stability. Validation with real-world data confirms high forecasting accuracy and effective cost reduction, highlighting the practicality and scalability of the proposed solution for smart microgrid management.The novelty of this work lies in the synergistic integration of three diverse analytical domains—time-series smoothing (KAMA), signal processing (IIR Filter), and deep learning-based sequence modeling (BiLSTM)—to develop a robust forecasting framework for Solar Power Generation and Load Demand prediction. While each of these techniques has been applied individually in other problem settings, their combined utilization for renewable energy forecasting forms a distinct and innovative methodology.