Intelligent Solar Forecasts - Modern Machine Learning Models & TinyML Role for Improved Solar Energy Yield Predictions

Aug 3, 2026

Advancing sustainable energy relies on reliable forecasting to enable efficient solar energy management, reduce costs, and support stable grid operation. This paper presents a tiny machine learning approach for real-time, low-cost solar energy yield prediction on resource-constrained edge IoT devices. It evaluates four deep learning models for solar farm yield forecasting while analyzing how hyperparameter tuning affects accuracy and robustness. The study discusses implementation challenges and potential benefits for residential and industrial control systems.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages19

Reading Time48 min

AccessFree to download

Introduction

Motivation

Abstract and Index Terms

Methodology and Model Evaluation

Hyperparameter Tuning Impact

Implementation Challenges and Opportunities

Implications for Energy Management

This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and content may change prior to final publication. Citation information: DOI 10. 1109/ACCESS.2024.3354703 Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000 . Digital Object Identifier Intelligent Solar Forecasts: Modern Machine Learning Models & TinyML Role for Improved Solar Energy Yield Predictions ALI M. HAYAJNEH1 ,(Member, IEEE), FERAS ALASALI1 ,(Member, IEEE), ABDELAZIZ SALAMA2 ,(IEEE Member), WILLIAM HOLDERBAUM3 ,(Member, IEEE) 1Department of Electrical Engineering, Faculty of Engineering, The Hashemite University, P.O. Box 330127, Zarqa 13133, Jordan. 2Department of Electrical and Electronic Engineering, University of Leeds, Leeds, UK. 3 School of Science, Engineering and Environment, University of Salford, Salford M5 4WT, U.K. Corresponding authors: William Holderbaum ([e-mail: w.holderbaum@salford.ac.uk](e-mail: w.holderbaum@salford.ac.uk)) and Ali M. Hayajneh (e-mail: [alihayajneh@hu.edu.jo](alihayajneh@hu.edu.jo)). This work is supported by funding from the Scientific Research and Innovation Support Fund, Ministry of Higher Education & Scientific Research, The Hashemite Kingdom of Jordan, under grant number (ENE/1/02/2022) . The work is partially funded by the University of Salford, UK, and Royal Academy of Engineering, UK, under grant number DIA-2021-18 . ABSTRACT The advancement of sustainable energy sources necessitates the development of robust forecasting tools for efficient energy management. A prominent player in this domain, solar power, heavily relies on accurate energy yield predictions to optimize production, minimize costs, and maintain grid stability. This paper explores an innovative application of tiny machine learning to provide real-time, low-cost forecasting of solar energy yield on resource-constrained edge internet of things devices, such as micro-controllers, for improved residential and industrial energy management. To further contribute to the domain, we conduct a comprehensive evaluation of four prominent machine learning models, namely unidirectional long short-term memory, bidirectional gated recurrent unit, bidirectional long short-term memory, and simple bidirectional recurrent neural network, for predicting solar farm energy yield. Our analysis delves into the impacts of tuning the machine learning model hyperparameters on the performance of these models, offering insights to improve prediction accuracy and stability. Additionally, we elaborate on the challenges and opportunities presented by the implementation of machine learning on low-cost energy management control systems, highlighting the benefits of reduced operational expenses and enhanced grid stability. The results derived from this study offer significant implications for energy management strategies at both household and industrial scales, contributing to a more sustainable future powered by accurate and efficient solar energy forecasting. INDEX TERMS Solar power forecasting, Time series forecasting, Internet of things, Deep neural networks. I. INTRODUCTION A. MOTIVATION SOLAR photovoltaic (PV) integration into global power systems has increased significantly over the past decade. The majority of these PV facilities are deployed in lowvoltage (LV) and medium-voltage (MV) networks, presenting distinct challenges for integrating renewable energy sources (RES) as distributed generation (DG) . In distribution networks (DN), these difficulties include reverse power fluxes, voltage violations, and grid stability [1], [2] . González-Sotres et al. [1] conducted a study to assess the impact of forecasting on centralised voltage control for solar generation in distribution systems. Their findings emphasised the significance of accurate forecast data for achieving optimal control settings and highlighted the need for improvements in forecasting tools for pre

FAQs

What forecasting capability does the paper focus on for solar energy management?

The paper focuses on real-time, low-cost forecasting of solar energy yield to improve production optimization, cost reduction, and grid stability.

Which machine learning models are evaluated for predicting solar farm energy yield?

The study evaluates four prominent models: unidirectional LSTM, bidirectional GRU, bidirectional LSTM, and a simple bidirectional recurrent neural network.

How does hyperparameter tuning affect forecasting performance in this work?

The analysis investigates the impacts of tuning model hyperparameters on prediction accuracy and stability, providing guidance to improve robustness.

Aug 3, 2026

Upload by Ava Thompson