Electrochromic device modeling using an adaptive neuro-fuzzy inference system: A model-free approach

• ANFIS models are employed successfully to model the EC device.
• The results of the experiments proved the validity and interpretability of new models.
• The proposed models showed good performance for input noisy data.
• The time-evolution of Tlum in relation to the area of EC device is demonstrated.
• ANFIS models can act as virtual luminous transmittance sensors.

This paper presents a new approach for the modeling of an Electrochromic (EC) device. The proposed system relies on an adaptive network-based fuzzy inference system or equivalently, Adaptive Neuro-Fuzzy Inference System (ANFIS). The ANFIS network has used 33 experimental data sets of which, 24 data sets were taken as training data and 9 data sets were taken as testing data. The ANFIS performance statistical indices mean absolute error (MAE), root mean square error (RMSE), and non-dimensional error index (NDEI) are found to be close to zero and coefficient of determination (R2) and linear correlation coefficient (ρ) and variance account for (VAF) are found to be close to one. Some interpretability issues regarding the ANFIS models, such as rule consistency, rule separation, and rule completeness are discussed. Simulation examples are provided to illustrate the effectiveness of the proposed approach. This study also includes experiments that confirm the good performance and the potential of the ANFIS models for datasets with noise. The proposed models can be seen as virtual luminous transmittance sensors.