Nils Bornhorst’s research while affiliated with University of Kassel and other places

The increasing penetration from intermittent renewable distributed energy resources in distribution grid brings along challenges in grid operation and planning. To evaluate the impact on the grid voltage profile, grid losses, and discrete actions from assets (e.g. transformer tap changes), quasi-static simulation is an appropriate method. Quasi-static time series and Monte-Carlo simulation requires a tremendous number of power flow calculations (PFCs), which can be significantly accelerated with a parallel High-Performance Computing (HPC)-PFC solver. In this paper, we propose a HPC-PFC-solver-based grid simulation (parallel simulation) approach for a multi-core CPU platform as well as a greedy method, which can prevent the errors caused by simultaneous parallel simulation. The performance of the proposed approach and the comparison is demonstrated with two use cases.

Power systems are rapidly and significantly changing due to the increasing penetration of distributed energy resources (DERs) and the rapid growth of widespread grid interconnections. An increasing number of grid operators is thus interested in the reduced equivalent representation of a large, interconnected power system to reduce the amount of required computational resources and data exchange, e.g., between grid operators. However, state-of-the-art grid equivalents become more and more inapplicable since they are analytically calculated for one specific grid state. They cannot properly be adapted to grid state changes and the behavior of the increasingly used controllers, such as reactive power controllers of DERs. Therefore, an innovative grid equivalent based on artificial neural networks (ANN) is proposed which overcomes the drawbacks of the state-of-the-art grid equivalents as follows: 1) Using supervised ANNs with feedforward and recurrent architectures, power systems can be equivalently represented adaptively and thus more accurately. 2) A feature selection method identifies the elements in the grid with high sensitivity on the boundary enabling a reduction of grid data required for the ANN-based equivalent. 3) To guarantee data confidentiality and cybersecurity, an additional unsupervised ANN, an Autoencoder, is used for obfuscation of the data which is required to be exchanged among grid operators, while the relevant information of the original data is preserved, maintaining the estimation accuracy. The ANN-based approach is analyzed and evaluated with two German benchmark grids and representative scenarios. The results demonstrate that the proposed approach outperforms the state-of-the-art radial equivalent independent method.

In the past decades, power systems worldwide have increased in size and complexity due to the increasing penetration of distributed energy resources and the rapid growth of widespread grid interconnections. The frequent grid state changes and the use of local controllers make the determination of appropriate grid equivalents challenging. Conventional grid equivalent techniques are based on one specific grid state. Larger deviations from this grid state leads to a reduced estimation precision of the grid equivalent. To tackle this issue, an artificial neural network (ANN) based approach is proposed in this paper for modeling an interconnected power system with a high share of distributed energy resources with reactive power control. In the proposed approach, the ANN is used as a grid equivalent, i.e., the ANN learns the relationship between the grid states (operating points, switching states, and controller parameters) and the power exchange at the interconnection, such that the effects of the external grid area at the boundary lines are accurately estimated. The performance of the ANN-based approach is compared to that of the state-of-the-art REI equivalent and the extended Ward equivalent.

Im Rahmen des Projekts RPC2 (Reactive Power Control 2) wird eine netzebenen- und netzbetreiberübergreifende Betriebsführung vom Fachgebiet Energiemanagement und Betrieb elektrischer Netze (e2n) der Universität Kassel und dem Fraunhofer IEE mit den Verteilnetzbetreibern (VNBs) LEW Verteilnetz GmbH (LVN) und AllgäuNetz (AN) entwickelt und angewendet. Ziel ist es, das gemeinsame Blindleistungsmanagement der beiden VNBs zu optimieren, wobei Informationen über das jeweilige Nachbarnetz ausgetauscht werden müssen. Dazu ist es erforderlich, statische Netzäquivalente mit Hilfe einer Netzreduktionsrechnung zu ermitteln um den Datenaustausch möglichst gering zu halten und Netzdaten zu anonymisieren. Konventionelle Netzäquivalentmethoden sind bspw. das Ward-Äquivalent [1] und das REI-Äquivalent (REI: radial, equivalent, independent) [2]. Die Genauigkeit dieser Ansätze ist jedoch bei Netztopologieänderung und signifikanter Netzzustandsänderung begrenzt [3], falls keine erneute Netzreduktionsrechnung durchgeführt wird. Aufbauend auf dem Ansatz von künstlichen neuronalen Netzen (KNN) für eine Zustandsschätzung [4] wird in dieser Veröffentlichung eine innovative Netzäquivalentmethode basierend auf KNN (KNN-Äquivalent) vorgeschlagen. Basierend auf der Open-Source-Software pandapower [5] sind die REI-Methode und die KNN-Methode zur Bildung vom Netzäquivalenten implementiert worden und anhand der Fähigkeit bewertet worden, die Knotenspannungen des Originalnetzes, d.h. des nicht reduzierten Netzgebiets, während der Simulation (mit Leistungs- und Netztopologieänderung) zu verfolgen. Die Ergebnisse zeigen eine hohe Genauigkeit der Methode des KNN-Äquivalents. Zudem wird die Rechenzeit erheblich reduziert.

Benchmark grids are power system parameter datasets, that are suitable for testing, publishing and comparing network planning and operation solutions and algorithms. A large amount of benchmark grids already exists. However, due to continuous development of new technologies and associated power systems changes, e.g. inverter-coupled distributed energy resources (DER) or controllable medium voltage (MV)-low voltage (LV) transformers, extended or new benchmark grids are required recurrently. To easily generate appropriate future benchmark grids, detailed descriptions of benchmark grid generation methodologies are required. Existing benchmark publications, however, lack such descriptions. Therefore, in this paper, an appropriate and comprehensible methodology to generate current and future benchmark grids is proposed. First results from the first application of the methodology in the project SimBench demonstrate its ability to generate an open-source, up-to-date, benchmark dataset that can be upgraded in future using the methodology.

2nd SimBench Poster at the Energy Research Network - Poster (German) presented at the 2nd presence meeting of Energy Research Network at the project group “Comparability and Transparency”, describing the generation of HV grid data from open-source geodata

The increasing number of distributed generators connected to the distribution system at the low and medium voltage level requires a reliable monitoring of distribution grids. Economic considerations prevent a full observation of distribution grids with direct measurements. First state-of-the-art approaches using artificial neural networks (ANN) for monitoring distribution grids with a limited amount of measurements exist. These approaches, however, have strong limitations. We develop a new solution for distribution system monitoring overcoming these limitations by 1. presenting a novel training procedure for the ANN, enabling its use in distribution grids with a high amount of distributed generation and a very limited amount of measurements, far less than is traditionally required by the state-of-the-art Weighted Least Squares (WLS) state estimation (SE), 2. using mutliple hidden layers in the ANN, increasing the estimation accuracy, 3. including switch statuses as inputs to the ANN, eliminating the need for individual ANN for each switching state, 4. estimating line current magnitudes additionally to voltage magnitudes. Simulations performed with an elaborate evaluation approach on a real and a benchmark grid demonstrate that the proposed ANN scheme clearly outperforms state-of-the-art ANN schemes and WLS SE under normal operating conditions and different situations such as gross measurement errors.

The SimBench poster (German) from the first presence meeting of the research network for energy presents the general SimBench methodology and the application at the multiple voltage levels