The progressive retirement of conventional power plants is decreasing the total system inertia and making the frequency control more delicate. In this context, voltage-sensitive loads will have a more pronounced impact on frequency. Therefore, there is an interest in understanding and quantifying how voltage dynamics impact the frequency response. For this purpose, a linearization of the system and the analysis of the obtained transfer functions are in focus. The paper aims to complement the traditional eigenvalue analysis, capture a system's response to large disturbances, and estimate the frequency overshoots. The study also shows that controllers affecting the load voltage, such as Power System Stabilizers (PSSs), can impact frequency response during a large disturbance. The impact of PSS on system responses and its transfer functions is carried out by applying a general control configuration. Both conventional PSS1A and multi-band PSS designs are analyzed and compared. Thereby, the paper explicitly interlinks a PSS design, and its parameters with properties of the system's frequency response reflected in the system transfer function. Illustrating enhancements facilitated by the presented approach, the explicitly derived PSS designs demonstrate the potential to reduce frequency overshoot. As an example, this reduction is achieved by improving a common mode damping or changing the position of the dominant zero of the closed-loop system.

Deployment of distributed generation (DG) is happening at an increasing rate driven by both the need for the reduction of humanity's carbon footprint and economic opportunities. The trend of increasing penetration levels of DG has also introduced interdependence between the responsibilities of system operators and the owners of the generating units. To fully accommodate distributed generation in the system, these interdependencies and the accompanying operational challenges need to be addressed. At the same time, the development of information and communication technologies within power systems has enabled access to immense information collected from the grids, such as the wide-area synchrophasor measurements. To utilize the increasing amounts of collected information and address the challenges of DG integration, novel applications can be developed by using data-driven methodologies.  This thesis aims to investigate if such data-driven approaches are sufficiently reliable, timely and accurate to address the DG integration challenges appearing across the responsibility areas. 

To explore the capabilities of data-driven approaches, two impacts of DG were studied in this thesis. First, different types of islanding detection methods were proposed. The data-driven methods for computing sensitivity parameters as seen from a distribution-level generator were developed and integrated within a proposed local islanding detection method. It was shown that such estimation methods enable a reliable and timely islanding detection by the use of field measurements. Next, synchrophasor-based remote islanding detection methods were developed as a tool for the situational awareness of system operators. It was shown that the dimensionality of the voltage angle measurements can be utilized to distinguish between islanding and other types of events using a data-driven approach. 

The other impact of DG that was addressed is the modelling of distribution networks in dynamic studies of a transmission network. A large number of existing approaches to dynamic modelling were identified pointing to the need for a model structure selection procedure. Thus, a data-driven modelling approach that selects an optimal model structure and evaluates the uncertainty of the models' outputs was proposed. The models' uncertainty was then used to select only informative events for parameter identification and to thereby reduce its computational burden. The performance of the method was demonstrated by using it to derive an optimal equivalent model of a modified CIGRE network.

Through the aforementioned contributions, the thesis shows that the increasing observability of the systems enables data-driven methodologies to address the challenges of DG integration. 

Utbyggnaden av distribuerad generering (DG) sker i en ökande takt, drivet av behovet av att minska mänsklighetens koldioxidavtryck och av ekonomiska möjligheter. Trenden med ökande penetrationsnivåer av DG har också infört ett ömsesidigt beroende mellan aktörerna i elsystsmet, inkluderande både producenter och nätägare. För att fullt ut kunna hantera DG i elsystemet måste dessa ömsesidiga beroenden och de åtföljande operativa utmaningarna åtgärdas. Parallellt med detta har utvecklingen av informations- och kommunikationsteknologier möjliggjort tillgång till enorma mängder data som samlats in från näten exempelvis synkroniserade fasvinkelmätningar. För att utnyttja de ökande mängderna insamlad information och möta utmaningarna med DG-integrationen, kan eventuellt nya applikationer utvecklas med hjälp av datadriven metodik. Denna avhandling syftar till att undersöka om de datadrivna tillvägagångssätten är tillräckligt tillförlitliga, snabba och precisa för att ta itu med de utmaningar som uppkommer på grund av DG-integrationen.

För att utforska fördelarna med datadrivna tillvägagångssätt studeras i denna avhandling två effekter av DG. Först presenterades olika typer av metoder för ödriftsdetektering. Datadrivna metoder för att beräkna känslighetsparametern för en generator på distributionsnivå utvecklades och integrerades i en lokal detekteringsmetod för ödrift. Det visades att sådana uppskattningsmetoder möjliggör en tillräcklig tillförlitlig och snabb detektering av ödrift genom användning av fältmätningar. Därefter utvecklades synkroniserad fasvinkel-baserade metoder för fjärrdetektering av ödrift som ett verktyg för systemoperatörerna. Resultaten visar att dimensionaliteten hos spänningsvinkelmätningarna kan användas för att skilja mellan ödrift och andra typer av händelser även då data-drivna metoder används. 

Den andra effekten av DG som togs upp är modelleringen av distributionsnät för dynamiska studier av transmissionsnät. Ett stort antal befintliga tillvägagångssätt för dynamisk modellering identifierades som pekade på behovet av en automatiserad metod för val av modellstruktur. Således föreslogs en modelleringsmetod som väljer den optimala modellstrukturen och utvärderar osäkerheten i modellernas resultat. Modellernas osäkerhet användes sedan för att endast välja informativa händelser för parameteridentifiering och därmed minska dess beräkningsbörda. Prestandan för den datadrivna metoden demonstrerades genom att härleda en optimal ekvivalent modell av ett modifierat CIGRE-nätverk.

Genom de tidigare nämnda bidragen visas det i avhandlingen att systemens ökande observerbarhet tillsammans med den datadrivna metodiken kan användas för att ta fram applikationer som kan möta utmaningarna med DG-integration.

Significant penetration levels of distributed energy resources increase the likelihood of continued operation of a power system island after an islanding event. It is important to employ adequate islanding detection methods to mitigate the adverse effects of unintentional islanding and possibly transition to a safe islanded mode of operation. This paper focuses on the computational aspects of a class of methods that utilizes a change in the sensitivity parameters as an indicator of islanding events. It is shown that the inherent properties of the measurement signals cause numerical issues for the computation of the sensitivity parameters. The paper also analyses three algorithms for estimation of the coefficients that overcome the numerical issues. The performance of the algorithms has been demonstrated using synthetically generated measurements. In addition, the data from field experiments has been used to further illustrate the practical viability of the algorithms. 

The likelihood of an unintentional power system islanding is increased in systems with significant penetration of distributed generation. To mitigate the adverse effects of islanding, a quick and reliable islanding detection method is needed. This paper first analyzes covariance matrices of a linearized power system model, and relates them to the principal component analysis of experimentally obtained covariance matrices. Additionally, a new model-independent islanding detection method is proposed that uses measurements of voltage angle differences between multiple locations in the system. The angle differences are first preprocessed to remove the effects of nonstationarity. Thereafter, a probabilistic model of principal component analysis is trained using the acquired measurements. The principal and residual spaces extracted from the measurements are used to discriminate between islanding and other events in the system. The applicability of the proposed method is demonstrated by using real measurements gathered from several locations in a transmission grid.

Modelling of distribution networks has received a renewed interest due to the appearance of dynamic phenomena arising from the increasing penetration levels of distributed generation. The conflicting implications of a model’s parsimony and flexibility, required model accuracy, available a priori information, and the purpose of a model, are inherent in system modelling. To tackle these issues, different models and identification methods have been proposed with the aim of representing the response of distributed generation to system events. The number of proposed models points to the fact that no single model can represent all possible dynamic behaviour of distribution networks. To facilitate, and even automate, model structure selection, this paper proposes a method that sequentially processes system events and classifies them as informative or noninformative for parameter identification. The method then uses informative events for parameter identification while the noninformative ones are used for optimal model structure selection and validation. Such selected and validated models can then be used in operation for dynamic security assessment of a transmission system. The performance of the proposed method is demonstrated in the case of equivalencing of a distribution network with photovoltaic and wind generation. Author

This paper describes new additions to the Rapid Parameter Identification Toolbox (RaPId), which is an open-source MATLAB toolbox for parameter estimation using models developed with the Modelica language and exported with the functional mock-up interface (FMI) Standard. These additions include an updated graphical user interface (GUI), an optimization method utilizing multiple starting points for a gradient descent optimization, and examples for different cyber-physical system applications such as the Duffing-Holmes equation modeling in a form of electrical circuit and a hydroelectric power plant modeling.

The growing presence of distributed generation in power systems increases the risk for the unintentional creation of electrical islands. It is important to apply reliable and quick is landing protection methods. At the same time, the deployment of phasor measurement units facilitates the usage of data-oriented techniques for the development of new wide-area protection applications, one of which is islanding protection. This paper presents a Bayesian approach to detecting an islanding event, which utilizes measurements of voltage angles at the system's buses. A model of mixtures of probabilistic principal component analysers has been fitted to the data using a variational inference algorithm and subsequently used for islanding detection. The proposed approach removes the need for setting parameters of the probabilistic model. The performance of the method is demonstrated on synthetic power system measurements.

Substation automation systems are characterised by a high degree of functional integration, which can lead to a centralised substation protection and control (CPC) architecture. Most CPC architectures utilise merging units to interface with current and voltage transformers, which causes a high-communication load on the process bus in case of large substations. This study proposes a CPC architecture based on distributed signal processing units (DSPUs) to overcome those drawbacks by publishing the results of the signal processing algorithms directly. The reduction of the communication load through the usage of DSPUs has been shown in a case study, which uses a 16-bay transmission substation topology as a reference.

This paper presents the latest improvements implemented in the Open-Instance Power System Library (OpenIPSL). The OpenIPSL is a fork from the original iTesla Power Systems Library (iPSL) by some of the original developers of the iPSL. This fork's motivation comes from the will of the authors to further develop the library with additional features tailored to research and teaching purposes. The enhancements include improvements to existing models, the addition of a new package of three phase models, and the implementation of automated tests through continuous integration.

The increased integration rate of inverter-interfaced devices is affecting the frequency response of the modern power systems. This leads to an increase of the variability of the power generation and to a reduction of the total system's inertia. This evolution of the system necessitates the prediction of frequency metrics, so that the frequency stability of the system can be guaranteed and that necessary mitigation measures can be taken. This paper proposes a method to predict the frequency nadir by using a Neural Network (NN) approach. As the approach uses measurements during a first short time period after the event, it more accurately predicts the frequency nadir compared to using static values. Several inputs for the NN are examined and when the appropriate ones are selected, a highly accurate prediction is accomplished.