ANFIS and ANN models to predict heliostat tracking errors

Area of Science:

• Renewable Energy Engineering
• Artificial Intelligence in Energy Systems
• Optical Engineering

Background:

• Heliostat field performance is critical for solar tower power efficiency.
• Accurate tracking of heliostats is essential but often costly and time-consuming.
• Developing faster, less expensive methods for evaluating heliostat tracking is necessary.

Purpose of the Study:

• To develop and compare rapid, accurate prediction models for heliostat error tracking.
• To evaluate the effectiveness of Adaptive Neuro-Fuzzy Inference System (ANFIS) and Artificial Neural Network (ANN) for this task.
• To provide an alternative to conventional geometric error models.

Main Methods:

• Applied ANFIS and ANN models using experimental data from seven days.
• Utilized time and day number as input parameters.
• Predicted beam reflected position along altitude and azimuth axes.

Main Results:

• Both ANFIS and ANN models successfully predicted beam reflected positions.
• Intelligent methods demonstrated superior performance compared to conventional geometric models.
• ANFIS achieved a coefficient of correlation (R ²) of 0.97, while ANN achieved 0.96 for altitude and 0.92 for azimuth.

Conclusions:

• ANFIS and ANN models offer a fast and accurate approach to estimating heliostat tracking errors.
• These intelligent methods present a promising alternative for predicting heliostat beam reflected positions.
• The findings support the use of AI in optimizing solar power plant operations.