Machine Learning Algorithm for Assessing Photovoltaic Panels Partial Shading Losses

Strategic Insights -

Researchers at the University of Porto have developed a machine learning algorithm to identify and assess the impact of partial shading on photovoltaic panels. According to the study, partial shading can reduce production and accelerate aging of these panels. The algorithm, which utilizes K-means clustering and a Long Short-Term Memory neural network, demonstrates flexibility and scalability without requiring prior dataset knowledge from the end user.

Key Takeaways:

  • The machine learning algorithm was developed to assess the impact of partial shading on photovoltaic panels, which can reduce production and accelerate aging.
  • The algorithm employs K-means clustering to analyze differences between expected and measured power, grouping data based on these deviations.
  • The Long Short-Term Memory model demonstrates flexibility and scalability without requiring prior dataset knowledge from the end user.
  • The algorithm uses photovoltaic panel electric circuit models to calculate predicted power based on measured panel irradiance, current, and voltage.
  • The study presented experimental data from both models, with the K-means model achieving a closer approximation to reference values.
  • The research was conducted by Armando Luis Sousa Araujo and Tiago Francisco Pires at the University of Porto.
  • The study utilized a real case study to demonstrate the effectiveness of the algorithm.

Statistics:

  • The study concluded that partial shading can reduce photovoltaic panel production by [x]% and accelerate aging by [x] years.
  • The Long Short-Term Memory model achieved a classification accuracy of [x]% in identifying periods of partial shading.
  • The K-means model achieved a closer approximation to reference values, with a mean absolute error of [x]%.
  • The study utilized a dataset of [x] samples from a real photovoltaic panel system.

Sources:

  • Machine Learning Algorithm for Assessing Photovoltaic Panels Partial Shading Losses based on Inverter Data by Armando Luis Sousa Araujo and Tiago Francisco Pires, U.Porto Journal of Engineering, 2025,11(1).
  • U.Porto Journal of Engineering, Universidade do Porto.
  • DOI: 10.24840/2183-6493_0011-001_002742.