Beijing Institute of Technology Press Co., Ltd
image:
Feature selection-based irradiance forecast for efficient operation of a stand-alone PV system
view more
Feature selection-based irradiance forecast for efficient operation of a stand-alone PV system
Credit: Green Energy and Intelligent Transportation
Researchers have developed a feature selection-based solar irradiance forecasting method to improve the operation of stand-alone photovoltaic systems. The approach uses a bidirectional long short-term memory hybrid network to forecast solar irradiance and then applies the forecasted data to estimate the optimum tilt angle of photovoltaic panels, helping increase PV output power.
Solar irradiance forecasting is important because photovoltaic power output depends directly on the amount of solar energy reaching a panel. In stand-alone PV systems, accurate forecasts can help operators understand the likely availability of solar energy and make better decisions about system configuration and operation. When forecasting is poor, PV systems may operate less efficiently, especially in settings where grid support is limited or unavailable.
The tilt angle of a PV module is another key factor in energy production. A panel that is not oriented effectively may receive less solar irradiance than it could under a better angle, reducing power output even when the solar resource is available. Determining the optimum tilt angle, or OTA, can therefore be an important step for improving PV system performance.
The new study connects these two tasks by using forecasted solar irradiance data to determine the optimum tilt angle. According to the article, the researchers first use a bidirectional long short-term memory, or Bi-LSTM, hybrid network to forecast solar irradiance. Bi-LSTM models are useful for time-series forecasting because they can learn sequential patterns in both forward and backward directions, helping capture relationships in meteorological and irradiance data.
A feature selection step is used to identify input parameters that improve the accuracy of solar irradiance forecasting. This is important because not all available input variables contribute equally to prediction quality. Selecting more informative features can reduce unnecessary complexity and help the forecasting model focus on the factors most relevant to solar irradiance behavior.
After forecasting solar irradiance, the study estimates the optimum tilt angle of the PV module by applying the forecasted data to the ASHRAE solar irradiance model. ASHRAE refers to the American Society of Heating, Refrigerating and Air-Conditioning Engineers. By combining a machine-learning forecast with a physical irradiance model, the method aims to connect data-driven prediction with practical PV panel orientation decisions.
The researchers compared the performance of the Bi-LSTM hybrid network with observed solar irradiance data and with existing forecasting models reported in the literature. They also evaluated the impact of optimum tilt angle by comparing solar irradiance received on tilted and horizontal surfaces. This comparison helps show whether improved forecasting and tilt-angle selection translate into better physical energy capture, rather than only better numerical prediction.
The work was experimentally implemented using a PV module setup at Thiagarajar College of Engineering in Madurai, Tamil Nadu, India. According to the article, the optimum tilt angle obtained by the proposed method produced higher PV output power than other tilt-angle approaches reported in the literature. The study also states that the proposed methodology achieved higher PV output power in both simulation and experimentation.
Further validation will still be needed across different climates, seasons, PV module types, mounting constraints, and stand-alone system architectures. Even so, the study offers a strong indication that feature-selected irradiance forecasting can support more efficient PV operation when paired with optimum tilt-angle estimation. For stand-alone photovoltaic systems, such methods could help improve power output and make solar energy use more reliable in off-grid and distributed energy applications.
Green Energy and Intelligent Transportation
10.1016/j.geits.2025.100308
Experimental study
Not applicable
Feature selection-based irradiance forecast for efficient operation of a stand-alone PV system
12-Jan-2026
Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.
Media Contact
Ning Xu
Beijing Institute of Technology Press Co., Ltd
xuning1907@foxmail.com
Beijing Institute of Technology Press Co., Ltd
Copyright © 2026 by the American Association for the Advancement of Science (AAAS)
Copyright © 2026 by the American Association for the Advancement of Science (AAAS)
