Skip to content
b edited this page Feb 16, 2019 · 36 revisions

General Idea

The performance of solar photovoltaic (PV) systems can degrade over time, we aim to provide a tool to check the performance. The user fills in as input location and specifications of the system. Our app will then calculates if the performance is acceptable.

We created a web app that allows users to check the status of their solar installations by evaluating if the energy production is at a normal state. Our application aims at providing a quick verification of the monthly energy production of rooftop solar panels. The user will provide the following input:

  1. Location of the solar panel: address
  2. Size of PV: area in m2
  3. Capacity kWp
  4. 12 months of production data

The app looks up the estimations from sonnendach, compares the the input with the reference values and returns:

  1. Location on map
  2. Status: ok or not ok
  3. Monthly comparison of actual vs. reference values as a bar chart

Methods and tools

  • web app running on amazon s3
  • basic statistics
  • visualization with d3.js

Screens

First prototype

(1) the user input interface

(2) the output results

Comparison to a test system

We tested our workflow on data from various systems. The following plots shows a comparison between the actual production (measured) of a PV system and the data from sonnendach. As shown below differences vary month to month and amount to 4.53 kWh/m2 is this case.

To determine if the system is not working properly we perform statistical significance tests, assuming that the measurements and reference data from sonnendach are normally distributed. Below an illustration of the concept, showing the relation between measured (i.e. Gesamt-Stromerzeugnung) and sonnendach, with a linear regression and confidence intervals.

Possible extensions

Data

  • Allowing automated / continous data input (similar to sonnenertrag.eu). This would enable warnings.
  • Historic data of real installations to identifiy the variance and detect outliers
  • Collecting feedback of users if the prediction was correct, after they had a professional check. This would allow learning and finetuning the

Frontend

Logic

  • Historic comparison. Currently sonnendach only provides the last 12months.
  • Adding the age of the installations to take into account the natural degrading of panels (e.g. 0.8% per year)
  • Finetuning benchmart algorithm to take into account installation specifics (e.g. paneltype, converter, wirkungsgrad).
  • Take into account shading effects of for example trees

Business

  • Providing the service as a subscription with warnings

Links Misc

Data

  • Ueberlandstrasse 2d, 8953 Dietikon: here

  • Test system

Clone this wiki locally