davecook/solar-panel-yield-2026

Reference calculator for PV soiling loss and water-fed pole cleaning ROI in Canada.

v0.1.0 2026-04-12 22:47 UTC

Requires

• php: >=8.0

MIT

• Dave Cook

README

Reference calculator, datasets, and working paper on photovoltaic (PV) soiling loss, water-fed pole (WFP) cleaning engineering, and cleaning ROI for PV systems in Canada.

What this is

A companion toolkit to the working paper Solar Panel Cleaning Yield Recovery: Soiling Loss, Water-Fed Pole Engineering, and Cleaning ROI for Photovoltaic Systems in Canada (v0.1.0, April 2026). The paper and this repository cover:

• Soiling physics: how fast transmittance drops as dust accumulates, by climate zone and panel tilt
• Water-fed pole engineering: deionized resin capacity as a function of inlet TDS, PV geometry to pole-length mapping, brush technique, and the fall-protection argument for grade-based operation
• Regional solar resource: monthly irradiance for six Ontario cities
• Cleaning economics: annual yield recovered vs cleaning interval for residential, commercial, and utility-scale systems
• Regulatory crosswalk: OSHA, CSA, IEC, ASTM, Health Canada, NRCan, ECCC

Repository layout

datasets/        7 CSVs + generator + HF dataset card
pdfs/            Working paper (31 pages) + generator
docs/            Sphinx documentation (ReadTheDocs)
engines/         8 language implementations:
  python/        FastAPI + CLI reference engine (PyPI)
  rust/          Actix-web server (Crates.io)
  java/          Maven-packaged JAR (Maven Central)
  ruby/          Gem (RubyGems)
  elixir/        Mix package (Hex.pm)
  php/           Composer library (Packagist)
  go/            Go module (pkg.go.dev)
  nostr/         Long-form article publisher (6 relays)
notebooks/       Jupyter notebook (GitHub browse)
kaggle/          Same notebook + Kaggle metadata
observable/      ObservableHQ notebook markdown
public/          PHP form + JS calculator
SOURCES.md       Bibliography

Quick start (Python)

pip install solar-panel-yield-2026
python -c "from engine import calculate, _canonical_vector; print(calculate(_canonical_vector()))"

Or run the reference engine against the canonical test vector from the source tree:

python engines/python/engine.py sample

Or start the HTTP server:

python engines/python/engine.py serve --port 8000
curl -X POST http://127.0.0.1:8000/calculate -H 'Content-Type: application/json' -d '{
  "system_capacity_kwp": 5.0,
  "ghi_annual_avg": 3.54,
  "days_since_cleaning": 30
}'

Cross-language parity

All eight engines produce identical output for the canonical test vector:

Input	Value
system_capacity_kwp	5.0
ghi_annual_avg	3.54 (Sudbury)
days_since_cleaning	30
soil_class	medium
climate_zone	temperate
panel_tilt_deg	15
electricity_price_cad_per_kwh	0.14
cleaning_visit_cost_cad	120.00
tap_water_tds_ppm	180
panel_height_ft	12

Output	Value
current_loss_pct.expected	4.3485 %
lost_cad_per_day.expected	0.0862 CAD
clean_annual_kwh	5,168.4
never_clean_annual_loss_cad	86.83
wfp_pole_ft	10
wfp_cartridge	1L_cartridge
water_per_cleaning_l	15.2
ro_prestage_recommended	false

Datasets

Seven CSVs published under CC BY 4.0; see datasets/README.md.

Working paper

31-page working paper, CC BY 4.0. Build from source:

pip install reportlab
python pdfs/generate_pdfs.py

License

• Code: MIT (see LICENSE)
• Datasets, PDF, dataset card: CC BY 4.0

Maintainer

Dave Cook, Binx Professional Cleaning, North Bay, Ontario, Canada. Contact: [e-mail].