Model total photovoltaic power from panel measurements
The updated model estimates total_pv_power from environmental, orientation and individual-panel measurements. It uses 7,080 records from the Photovoltaic Generation Dataset with Orientation Diversity.
1. Industrial challenge
Energy analysts can inspect the relation between measured panel outputs and aggregate power. Because the five panel powers are themselves model inputs, this saved configuration is an aggregate-power modelling demonstration rather than an independent future generation forecast.
Updated measurements
Use the five-panel orientation dataset.
Aggregate output
Inspect measured versus estimated total power.
Forecast distinction
Separate available measurements from future predictors.
2. Data set
The adapted Zenodo dataset contains 21 columns: 20 inputs and total_pv_power. Inputs include calendar encodings, weather descriptors, radiation, incidence-cosine descriptors and pv1_power through pv5_power.
Source: Photovoltaic Generation Dataset with Orientation Diversity. Dataset license: CC-BY-4.0. The downloadable ZIP includes the adapted data and attribution notices.
| Dataset measure | Saved value |
|---|---|
| Analysis unit | photovoltaic observation |
| Records | 7,080 |
| Raw variables | 21 |
| Encoded model inputs | 20 |
| Model outputs | 1 |
| Training roles | 4248 |
| Validation / selection roles | 1416 |
| Testing roles | 1416 |
| Unused roles | 0 |
| Field | Role | Type | Categories |
|---|---|---|---|
| day | Input | Numeric | |
| cos_day | Input | Numeric | |
| hour | Input | Numeric | |
| cos_hour | Input | Numeric | |
| air_temperature | Input | Numeric | |
| dew_point | Input | Numeric | |
| wind_speed | Input | Numeric | |
| air_pressure | Input | Numeric | |
| weather_condition | Input | Numeric | |
| solar_radiation | Input | Numeric | |
| pv1_incidence_cos | Input | Numeric | |
| pv1_power | Input | Numeric | |
| pv2_incidence_cos | Input | Numeric | |
| pv2_power | Input | Numeric | |
| pv3_incidence_cos | Input | Numeric | |
| pv3_power | Input | Numeric | |
| pv4_incidence_cos | Input | Numeric | |
| pv4_power | Input | Numeric | |
| pv5_incidence_cos | Input | Numeric | |
| pv5_power | Input | Numeric | |
| total_pv_power | Target | Numeric |
In all 7,080 rows of the supplied CSV, total_pv_power equals the exact sum of pv1_power through pv5_power. Those five component powers are model inputs, so this fit demonstrates aggregation of measured outputs rather than an independent forecast.
3. Model
The model has 20 encoded inputs and 1 outputs. No architecture-selection experiment is recorded in this project. The diagram shows the topology used by the saved model.
| Layer | Input shape | Output shape | Activation |
|---|---|---|---|
| Scaling | 20 | 20 | |
| Dense | 20 | 3 | Tanh |
| Dense | 3 | 1 | Identity |
| Unscaling | 1 | 1 | |
| Clamping | 1 | 1 |
4. Training strategy
The saved training configuration uses QuasiNewton with NormalizedSquaredError.
Quasi-Newton method results
| Measure | Value |
|---|---|
| Epochs number | 25 |
| Elapsed time | 00:00:00 |
| Stopping criterion | Loss goal |
| Training error | 0.001 |
| Validation error | 0.001 |
5. Model selection
No model selection experiment is recorded for this version. The validation subset guides fitting where a training report is present; it is distinct from the held-out test rows.
6. Testing analysis
The figures and tables below refer to the current project’s saved testing analysis. The subset uses testing role 2; it contains 1416 source records.
total_pv_power goodness-of-fit parameters
| Measure | Value |
|---|---|
| Determination | 0.999 |
7. Model deployment
Open the downloaded project in Neural Designer, inspect the dataset roles and preprocessing, then review the saved task report. Use the same input schema and category order when calculating outputs. The ZIP contains the exact current .nd, its source data and the applicable dataset notices.
Workflow: source measurements → schema and availability checks → model output → domain review. Keep model versions, validation evidence and incoming-data monitoring together.
8. Scope and limitations
The target is closely related to the individual panel powers used as inputs; the very high test fit should be interpreted in that context. A future-generation forecast would require predictors known at the forecast origin and time-ordered evaluation. Do not transfer the original page’s model weights or performance figures to this dataset.
References
- Photovoltaic Generation Dataset with Orientation Diversity. Mehmet Cunkas; Mustafa Arslan; Hasan Huseyin Cevik. 10.5281/zenodo.19245713
- Dataset terms: Creative Commons Attribution 4.0 International. Full attribution and transformations are included in
LICENSES/DATASET-LICENSE.txt. - Current Neural Designer project and saved task report, snapshot 6 October 2026.
