SolarPro Performance Simulator

Photovoltaic Residential Modelling & Analysis

Model Empirical + TDR
Split 2 Years Train
Status Ready
What does this simulator do?

This platform implements the PVUSA (Photovoltaics for Utility Scale Applications) empirical regression model to predict photovoltaic system energy production based on weather conditions. It incorporates a comprehensive Transitory Deterioration Rate (TDR) model that accounts for three degradation mechanisms:

1. Thermal Degradation (Arrhenius): Models accelerated aging due to elevated cell temperatures using activation energy principles (Ea ≈ 0.65 eV for EVA encapsulants).

2. Humidity Degradation (Hallberg-Peck): Captures moisture-induced corrosion and delamination through combined temperature-humidity acceleration factors.

3. Soiling Losses (HSU Model): Simulates particulate accumulation (PM10/PM2.5) and rain-based cleaning events using validated deposition velocities.

The workflow uses train/test temporal splitting to calibrate model coefficients on historical data (default: 2 years) and validate predictions on future unseen periods, preventing data leakage and ensuring realistic performance assessment.

TDR Model Configuration (Thermal + Humidity + Soiling)
Enable TDR
TDR Parameters: These values control how the model simulates long-term performance degradation. Activation energies (Ea) determine temperature sensitivity, weights balance thermal vs. calendar aging, and soiling parameters model dust accumulation between rain events. Default values are based on peer-reviewed literature for crystalline silicon modules in Mediterranean climates.
Train/Test Split
Years (calibration data)
Thermal Degradation (Arrhenius)
eV (0.6-1.1 for EVA)
K (STC = 298.15 K)
fraction (2%)
fraction/year (0.5%)
0-1 (thermal vs calendar)
Humidity Degradation (Hallberg-Peck)
eV
n (typical 2-3)
contribution factor
fraction/year
Soiling (HSU Model)
ug/m3
ug/m3
mm rain
fraction (first year)
Workflow Steps: The simulation follows a six-stage pipeline: (1) Load plant configuration from JSON, (2) Fetch hourly weather data from Open-Meteo API, (3) Calibrate empirical coefficients using training data only, (4) Generate predictions across all periods, (5) Validate using separate train/test metrics, (6) Generate week-ahead forecast using Open-Meteo Forecast API.
1
Load Plant Configuration
Ready

Reads hierarchical JSON structure containing building metadata, array specifications (tilt, azimuth, module count), and temporal boundaries for the simulation period.

2
Fetch Weather & Power Data
Ready

Retrieves hourly GTI, temperature, wind speed, humidity, and precipitation from Open-Meteo API. Applies wind correction to roof height and calculates cell temperature using Skoplaki model.

3
Calibrate on Training Data
Ready

Performs Huber regression on filtered training samples (GTI > 50 W/m², midday hours) to extract c₁-c₄ coefficients. Uses IQR and Hampel outlier removal for robustness.

4
Generate Predictions (All Data)
Ready

Applies empirical regression model to full dataset, then multiplies by combined TDR factor: TDR(t) = [1-D_thermal][1-w_h·D_humidity]·SR(t), where SR is the soiling ratio from HSU model.

5
Validate (Train + Test + Combined)
Ready

Computes R², RMSE, MAE, MAPE, and cumulative error separately for training period, test period, and combined dataset. Generates scatter plots and residual diagnostics.

6
15-Day Forecast
Ready

Fetches 15-day weather forecast from Open-Meteo Forecast API and applies calibrated PVUSA model with TDR to predict future energy production. Uses current system degradation state as baseline.

Validation Results: The metrics below quantify model accuracy across different temporal partitions. Training metrics show calibration fit quality, test metrics demonstrate generalization to unseen data, and combined metrics represent overall performance. R² > 0.85 indicates strong predictive power for operational forecasting.

Validation Results (Train / Test / Combined)

TDR ENABLED
Daily Energy Comparison
Export Data
System Log