Here you will find factsheets, publications and results related to the AISOP project.
This document comprises the annual report submitted to the Swiss funding authority at the end of 2024.
This report provides an overview of the activities on ML-driven demand side anomaly detection and identification. This task of the project focuses on anomalies like newly installed PV systems and other notable data shifts, aiding DSOs by providing an alarm signal.
Dimitrios Papadopoulos presented his M.Sc. Thesis entitled “Low Voltage Load Forecasting Using Ensemble Methods”. In his work, he studied forecast day ahead energy consumption using data from smart meters. The models studied included Feed Forward Neural Networks, Convolutional Neural Networks, Generalised Additive Models, and Gradient Boosted Regression Trees. Feed Forward Neural Network and Generalised Additive Models are combined using simple and weighted averaging. The procedure is repeated, using Convolutional Neural Network and Generalised Additive models. The results are compared with Gradient Boosted Regression Trees model because it is considered an accurate ensemble model. Finally, models trained on one smart meter data set, are applied to forecast data from different smart meters to assess how well are the models generalising to different datasets.
Our paper entitled “A framework for data-driven decision support for operational planning in active distribution networks” was presented during the Theme 2: Network operation and control supporting increased hosting capacity session of the CIRED 2024 Vienna Workshop on June 20th. Were we give an overview of the operational planning framework along with examples of data analytics we are developing. In this framework a DigitalProcessTwin (DPT) with Single Source of Truth (SSoT) facilitates data management from IoT sensors and model outputs, and workflows define analytics and forecasting tasks to support operator decisions. Examples of load forecasting and anomaly detection in power quality data are presented.
Note that the paper is published in the IET Conference Proceedings. Volume 2024, Issue 5. https://doi.org/10.1049/icp.2024.1979 but it may be hard to access, please contact us if you need it. On the download link to the right you can find our poster.
The CKW Group is a distribution system operator that supplies more than 200,000 end customers in Central Switzerland. CKW publishes anonymised and aggregated data from smart meters that measure electricity consumption in canton Lucerne.
We pre-processed part of these data and sorted it per smart meter ID. It is stored as parquet files named with the id field of the corresponding smart meter anonymised data. Example: 027ceb7b8fd77a4b11b3b497e9f0b174.parquet
Part of these data are archived in the cloud public cloud space of AISOP project https://os.zhdk.cloud.switch.ch/swift/v1/aisop_public/ckw/ts/batch_0424/batch_0424.zip and also a zenodo public record from where it can be easily downloaded.
AISOP organised the track titled “AI for Energy Utilities” at the AMLD – Applied Machine Learning Days EPFL on Tuesday, March 26th. The track included a list of presentations by TSO, DSOs and a generative AI start-up. See details here.
It also included a poster session, where we presented “A DATA CO-PILOT FOR ELECTRIC DISTRIBUTION UTILITIES TO SUPPORT GRID SITUATIONAL AWARENESS”.
The paper “Measurement-based Locational Marginal Prices for Real-time Markets in Distribution Systems” presents a method to calculate distribution locational marginal prices (DLMPs) by means of online estimation of linear sensitivity models that map bus voltages to power injections and to line power flows. It uses only synchrophasor measurements collected at buses interfacing market participants and in lines of interest and it can be found in the IEEE Transactions on Power Systems: 10.1109/TPWRS.2024.3369476.
A pre-print can be easily accessed here.
This document comprises the annual report submitted to the Swiss funding authority at the end of 2023.
As distribution grids incorporate more renewable energy sources and demand becomes more flexible, more information about the current and future state of the grid becomes vital for operating the grid in a cost-effective way. Digitalization is therefore essential, as it facilitates data acquisition and processing. Digital process twins (DPTs) can help energy utilities to better manage the operation of their grids. Read about AISOP’s approach in this newsletter.
The paper by Mojgan Hojabri, Severin Nowak, and Antonios Papaemmanouil published in the open access journal Energies can be found at https://doi.org/10.3390/en16166023. It presents a comparison of classification algorithms for detection of intermittent faults of distribution network cables. The cases investigated with the IEEE 33 bus system include faults with high and low impedance, and short and long duration.
Presentation given by Antonios Papaemmanouil to the AI4Grids Symposium – AI in Distribution Networks on 26. September 2023.
This document provides an overview of the digital twin, including its history and definition. In addition, it introduces a new term in digital twin technology, the digital process twin, and explores its application in the power system based on the aims of the AISOP project.
ETG taskforce report, Ulf Hager et al
This document describes processes for quality control, risk management, data management and the communications strategy in the AISOP project
This document comprises the annual report submitted to the Swiss funding authority at the end of 2022.
AISOP will create an AI-assisted decision support system for the electric distribution system operators (DSOs) to drive decarbonisation that is underpinned by advanced digital technology. We create a new “operational planning” framework for distribution system operators, inspired by similar approaches that are already in place for transmission system operation.
Project Duration: 01.05.2022 – 01.05.2025
Total Budget: € 2,105,684.
Funding: € 1,197,891.
Project Coordinator: Lucerne University of Applied Sciences and Arts.