Improving Short-Term Forecasting Models for Optimised Photovoltaic Curtailment Strategies

Rodriguez Ruiz-Canela, Carlos (2025) Improving Short-Term Forecasting Models for Optimised Photovoltaic Curtailment Strategies. Integration Project, Industrial Engineering and Management.

Preview	Text FinalReportCarloss5129087.pdf Download (1MB) | Preview
	Text Toestemming.pdf Restricted to Registered users only Download (153kB)

Abstract

This project addresses the critical need for accurate short-term photovoltaic (PV) forecasting to support FIRN Energy’s operational decision-making. Building on a prior ARIMAX-based approach, the work introduces enhanced forecasting pipelines for both minute-ahead and day-ahead horizons. A re-implemented ARIMAX(1,0,1) model was tuned for minute-ahead predictions, capturing rapid fluctuations in PV output using an autoregressive structure and exogenous environmental data. A Random Forest model was trained across data from over 20 FIRN sites for day-ahead forecasting, demonstrating superior generalisation and predictive accuracy compared to site-specific or previous statistical models. All models were developed in Python, enabling seamless integration with FIRN’s infrastructure. Results show significant accuracy gains across evaluation metrics (MAE, RMSE, MAPE, R2), with the day-ahead Random Forest achieving up to 68% MAPE reduction over the original ARIMAX baseline in worst-case sites. The final deliverable includes all forecasting code, input data templates, and API connectivity scripts, offering FIRN a scalable, ready-to-deploy solution.

Item Type:	Thesis (Integration Project)
Supervisor name:	Cucuzzella, M. and Taheri, M. and Vacchini, E.
Degree programme:	Industrial Engineering and Management
Thesis type:	Integration Project
Language:	English
Date Deposited:	21 Jul 2025 14:48
Last Modified:	21 Jul 2025 14:48
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/36449

Actions (login required)

View Item