Introduction: The growing share of distributed energy resources (DERs) and electric vehicles (EVs) in low-voltage (LV) grids is a challenge for distribution system operators (DSOs) as the volatility increases and higher power peaks are expected due to the simultaneity of EV charging processes in particular. To reduce possible necessary grid reinforcement, the grids could be operated closer to their operational limits. To make this possible, more transparency in LV grids is required, which is largely nonexistent because there are only very few measurements on medium-voltage/low-voltage transformers and the LV grids themselves. In the German Pilot of the EU project InterConnect, experts from Fraunhofer IEE, in collaboration with KEO Interactive, EEBus, Stromnetze Hamburg (SNH), and the University of Kassel, are at the forefront of intelligent EV charging testing. Their main focus is to establish a stardized approach for handling diverse data sources, estimating the grid states, leading to enhanced efficiency in charging of EVs, and increased security and reliability of the power system.

 

Methodology of State Estimation: Artificial neural networks (ANNs) are inspired by biological neural networks and can adaptively recognize patterns in data sets. In the German Pilot of the EU project, based on the existing measurements (input features), e.g., voltage, current, and active power, ANN is used to estimate the desired operating parameters (output labels), e.g., bus voltages and line/transformer loadings, to detect possible voltage band violations or overloadings, see Figure 1 a) and b)..

 

Implementation and Validation: The proposed method is implemented mainly based on the Python package pandapower [1] and Pytorch [2], i.e., the grid model and grid calculations are performed in pandapower environment, and the ANN is constructed using PyTorch. To verify the effectiveness, the proposed method has been applied to different SimBench [3] LV grids, i.e., the bus voltages and line loadings are estimated by the trained ANN based on the given measurements. The estimation accuracy in the LV grid is highly dependent on the specific characteristics of the grid, such as the proportion of DERs, the presence of volatile load customers, and the density of installed meters. In the observed simulations, the 99%-percentile of the maximum voltage error is less than 1% p.u. for all buses, while, considering all lines, the 99%-percentile of the maximum line loading error is less than 10%, mostly less than 5%. In real-world grids, they could be higher.

 

Use Case: In the German pilot phase, the project partner KEO equips the five selected hotels in Hamburg with intelligent charging systems to measure and transmit power usage data. The collected charging power data, along with grid measurements from SNH, are sent to the beeDIP-platform[1] developed by Fraunhofer IEE, see Figure 1 c). The proposed method is implemented as a beeDIP-microservice to estimate the corresponding grid states using the collected measurements. To prevent voltage violations or overloads caused by charging, an iterative calculation process is employed. This process limits the charging power, minimizing congestion issues within the grid as much as possible. The field test started in 2023.

 

Figure 1 Flow chart of the scheme: a) training phase; b) validation phase; c) application in InterConnect

 

Conclusion: In the process of power grid digitization, artificial intelligence plays a vital role in leveraging existing grid data to support power grid operation. In the EU project InterConnet, ANN is used to predict the impact of EV charging on the grid state.  It enables effective EV charging management while supporting power grid stability.

 

[1]   L. Thurner, A. Scheidler, F. Schäfer, J.-H. Menke, J. Dollichon, F. Meier, S. Meinecke, and M. Braun, “pandapower—an open-source python tool for convenient modeling, analysis, and optimization of electric power systems”, IEEE Transactions on Power Systems, vol. 33, no. 6, pp. 6510–6521, 2018.

[2]   A. Paszke, S. Gross, S. Chintala, G. Chanan, E. Yang, Z. DeVito, Z. Lin, A. Desmaison, L. Antiga, and A. Lerer, “Automatic differentiation in PyTorch”, 2017.

[3]   S. Meinecke, D. Sarajlić, S.R. Drauz, A. Klettke, L.-P. Lauven, C. Rehtanz, A. Moser, M. Braun, “SimBench—A Benchmark Dataset of Electric Power Systems to Compare Innovative Solutions Based on Power Flow Analysis”, Energies 2020, 13, 3290.

[4]   B. Requardt, “Architekturen und Verfahren für modulare Pilotsysteme und Erweiterungen von Netzleitstellen“, Ph.D. dissertation, Dept. e2n., Univ. Kassel, Kassel, Germany, 2021

[5]   B. Requardt, S. Wende – von Berg and M. Braun    “Plattform für Pilot Systeme im Netzoperationsbetrieb”, 16. Symposium Energieinnovation, Graz 2020

 

[1] beeDIP is a pilot system platform that operates on microservices architecture [4, 5]. Its primary purpose is to facilitate operation management by offering a modular software platform. This modular architecture allows for seamless integration and connection of new third-party or custom functions to the existing control room software. To ensure compatibility and uniformity, beeDIP employs data integration services that convert required data from various sources, such as TASE.2, into the common information model standard. These standardized data sets are then made available to the grid calculation services, such as optimal power flow and network constraint management.

Author: