Applications are open! Apply until 15 January for German-language Master’s programs and selected English-language Master’s programs starting in the summer semester.

Details here

TUHH > Studying > During your Studies > Career Center > Fishing for Experience > Archiv Projektaufgaben

CarbonFreed - Development of an AI-Tool to Extract Data out of Single Line Diagrams

We are a young and determined crew with a mission to initiate innovative projects and develop digital products which will contribute to the decrease of carbon emissions in our atmosphere to enable a better, greener future for everybody on this planet. To achieve this we strive to exhaust recent technological capabilities as well as trendsetting software applications.

In the long run, we do not want to focus on a single domain but rather constantly screen the energy transition market for new challenges which can be tackled using innovative technologies like Artificial Intelligence.

Currently, we concentrate on contributing to the expansion of the renewable energy sector by providing electrical planning assessments and certificates for medium-sized power systems (e.g. solar systems). These documents are demanded by law if one is planning to install such power systems and hence indispensable when it comes to pushing clean energy forward.

This topic also led to the origin of our first project: The development of a software solution that automates the elaborate preparation of such documents and thus decreasing the time required significantly. This relieves professional engineers of repetitive tasks, which is crucial due to the skilled worker shortage, while simultaneously increasing the expansion of renewable energy resources.

We operate with a “hands-on mentality” and scrutinize existing structures with the goal to get something going. Because there is no other way if we want to sustainably change our world for the better.

Situation

We are currently developing an AI-supported browser-based application to provide automated grid code certification for solar power plants. The application will collect all necessary data from the customer, check for completeness, and start selecting the relevant data for the certification process. At this stage, the customer can be reminded in case essential data are missing so that they can upgrade their data set with the missing information. In the next step, the algorithm will extract all information from the data set which is needed for the compliance evaluation. For instance, screening of the type of solar inverters planned to be used, the total installed capacity, or the type of grid protection units used in the power plant. Finally, the algorithm compares the customer data against the norm VDE-AR-N 4110 which describes the grid code requirements for grid connection to the medium-voltage network in Germany. The results get documented in an evaluation report which is the output of the application and subsequent delivery to the customer and grid operator.

Problem

In Germany, it is required by law to have a power plant certificate. Without such a certificate you won’t get grid access and therewith can’t receive the feed-in tariff. Current waiting times are up to 9 months and given the scarcity of Electrical Engineers the Certification bodies can’t just scale up their workforce to cope with the workload. We want to play our role in reducing the waiting times by automating the certification process as much as possible.

Important information can be captured from Single Line Diagrams (SLD). The SLD contains information on the medium voltage connections as well as the low voltage section of the solar power plant.

Today the relevant information needs to be extracted manually by an experienced Electrical Engineer. This takes time and can’t be easily scaled up and adjusted to the new demand.

Aims of the project

While developing the entire tool with professional software engineers the evaluation of the SLD shall be done separately. We want to automate the extraction of data from the SLD using Computer Vision. Since the signs of the components are standardized and normalized in the VDE DIN a computer vision algorithm can be trained to recognize the relevant data. Components to be found by the software are e.g., grid protection units, power plant controller, the current/voltage sensors and their respective input measurement data and the circuit breakers activated by the grid protection units. The results shall be summarized in a pre-defined template (e.g. .docx, .json).

Scope

Main objective of the project is to develop the above-mentioned functionality to extract the information automatically. The tools/software/frameworks to be used are secondary. It can be done with Open AI frameworks (pre-developed, pre-trained AI algorithm) or Python-based frameworks. We want to integrate it later in our software, so some interfaces need to be designed.

The work to be done shall be a proof of concept. We can try out different ways to achieve the goal to see which way gives the best results.

Target group (students)