Power systems are becoming increasingly complex, handling rising shares of distributed intermittent renewable generation, EV charging stations, and storage. To ensure power availability and quality, the grid needs to be monitored as a whole, by wide area monitoring (WAM), not just in small sections separately. Parameter oscillations need to be detected and acted upon. This requires sensors, data assimilation and visualization, comparison with models, modelling, and system architectures for different grid types.
This hands-on reference for researchers in power systems, professionals at grid operators and grid equipment manufacturers, as well as for advanced students, offers a comprehensive treatment of advanced data-driven signal processing techniques for the analysis and characterization of system data and transient oscillations in power grids. Algorithms and examples help readers understand the material. Challenges involved in realistic monitoring, visualization, and analysis of actual disturbance events are emphasized.
Chapters in this second edition cover WAM and analysis systems, WAM system architectures, modelling of power system dynamic processes, data processing and feature extraction, multi-sensor multitemporal data fusion, WAM of power systems with high penetration of distributed generation, distributed wide-area oscillation monitoring, near real-time analysis and monitoring, and interpretation and visualization of wide-area PMU measurements.
Inspec keywords: power system interconnection; distributed power generation; power grids; phasor measurement; data analysis; spatiotemporal phenomena; power system stability; sensor fusion
Other keywords: sensor fusion; power system measurement; monitoring; data analysis; spatiotemporal phenomena; power grids; power system interconnection; power system stability; interconnected power systems; phasor measurement; distributed power generation; wide area monitoring
Subjects: Power system management, operation and economics; Data handling techniques; Power system measurement and metering; General electrical engineering topics; General and management topics; Distributed power generation; Signal processing and detection; Power system control
The development of advanced wide-area measurement systems (WAMS) based on synchrophasor technology provides unprecedented views of power system dynamic behavior with increased resolution and accuracy [1-4]. In addition to the growth in the amount of data, the variety of measurement devices has also increased. In this context, advances in the development and installation of inexpensive, low-voltage recording devices have resulted in the deployment of many sensors that transmit data to specialized data concentrators. As the size and complexity of power grids continue to increase, real-time monitoring, protection, and forecasting of dynamic processes become increasingly important [5, 6]. This increase in the volume and variety of the data requires advances in methodology to understand, process automatically, and summarize the data.
Fast, high-quality synchrophasor measurements of voltage and current phasors using signals from a global positioning satellite can enhance wide-area visibility greatly and result in enhanced system security and reliability [7-9].
Advanced applications in wide-area monitoring encompass implementing situational awareness systems, including disturbance alerts, event location triangulation, oscillation detection, early warning systems, power system oscillatory tracking, and other advanced features [9-11]. At the core of these systems are intelligent sensing and signal processing and communication techniques to make optimal use of wide-area data.
This chapter provides an overview of fundamental principles in wide-area monitoring. Models, applications, and areas of improvement in real-time system monitoring and key research directions in data management are described and highlighted.
Several power utilities have recently designed and implemented advanced wide-area monitoring systems and data processing strategies to enhance grid stability and reliability. These strategies encompass the implementation of situational awareness systems, including disturbance alert, event location triangulation, oscillation detection, early warning systems, data archiving, and other advanced features [1-3].
Wide-area monitoring systems will continue to evolve as software, sensing, and communications technology advance and signal processing tools improve. Various forms of wide-area monitoring systems have been developed to give early warning of power system disturbances [4, 5]. In order to integrate data from multiple sensors, specialized data fusion and communication techniques must be integrated into the existing wide area monitoring system (WAMS) architectures. The design methods must incorporate both fault-tolerant strategies and intelligent data fusion techniques to enhance reliability and safety and also to improve the performance of global monitoring systems.
This chapter gives a broad overview of various wide-area monitoring architectures, including sensor development and conceptualization, data processing and data fusion, and anomaly detection algorithms. Inspired by recent work on data fusion techniques, advanced WAMS architectures are introduced that represent a combination of measuring, monitoring, and analysis architectures. Critical concepts in multisensor modeling, estimation, and fusion are reviewed.
A framework for integrating data from multiple sensors to produce actionable intelligence is identified. Advantages, challenging aspects, and recent advances in the design and implementation of WAMS architectures are reviewed, and areas, where research is needed to advance the use of WAMS data, are highlighted and described.
Challenges posed in developing distributed data fusion algorithms are also outlined.
The development of wide-area measurement systems (WAMS) provides unprecedented views of the system with increasing resolution and accuracy, coupled with the capabilities of measuring new variables [1, 2]. Central to developing advanced monitoring systems that improve the current predicting capabilities is the investigation of data correlations in both space and time.
Modeling spatio-temporal measured data presents a unique set of problems as it often exhibits spatio-temporal dependence, nonlinearity, and heterogeneity [3-7]. To capture the space and time variability, it is desirable to have an adequate distribution of sensors. Different studies have incorporated spatial information within the framework of WAMS [3, 8, 9]. Few attempts, however, have been made to integrate spatial and temporal approaches for investigating wide-area phenomena. In [8], spatio-temporal analysis methods have been used to examine dynamic trends and phase relationships between key system signals from measured data. These results, however, are intuitive; integrated spatio-temporal approaches are needed to quantify and understand patterns of system behavior in large-scale systems.
Predictions from spatio-temporal information from measured data can be used to study and evaluate patterns of variation in system dynamics and forecast the time evolution of transient processes [2]. Spatial models, in particular, may also serve as a basis for developing more complicated models that are better able to represent the observed oscillations at unmonitored locations [3]. Accurate spatial information can provide precise predictions of dynamic features such as propagating rates and the extent and distribution of mode propagation and may improve monitoring systems' detection and tracking performance [6].
This chapter introduces promising new approaches for estimating and predicting a multivariate spatio-temporal process from observational data. A general mathematical framework is provided in which spatio-temporal analysis of large data sets can be performed. In this approach, spatio-temporal models incorporating the dynamics of only a few temporal modes are developed. The technique can be used to monitor, model, and forecast wide-area physical processes such as inter-area oscillations and reveal a system's key features.
Methods for extracting the dominant temporal and spatial components of variability in measured data are investigated and numerical issues are addressed.
Power system data are often corrupted by non-Gaussian, nonlinear, and nonstationary artifacts and noise. High levels of ambient noise, in particular, result in nonstationary signals, which may lead to inefficient performance of conventional data processing methods.
Extracting robust parameters from such signals and providing confidence in the estimates, are, therefore, challenging and require an adaptive filtering approach that accounts for artifact types [1-3].
The extensive development of signal processing methods for measured data during the last decade has been guided by the study of modal properties of dominant interarea modes. Much of the literature on signal processing focuses on linear analysis. Linear, stationary methods are successful when carefully applied, but they lack the general applicability offered by a data-driven approach.
Traditionally, modeling techniques have dealt with complex behavior by trying to apply linear models to windows of observation exhibiting nearly stationary or linear characteristics.
Under stressed operating conditions, it may be possible to parameterize system behavior or even decide on the appropriate form of a linear model. Only very recently, techniques that account for nonlinear and nonstationary behavior have begun to percolate into power system data processing theory. In parallel with this, new statistical methods for identifying trends, quasistationary behavior, and other predictability measures continue to be developed. Explicit treatment of these issues has led to different data processing approaches with the ability to process more general system behavior.
Despite the prevalence of such a large number of signal processing methods and their success in several practical applications, modal analysis in power system data remains a difficult challenge.
This chapter examines nonlinear and nonstationary data processing methods. Multivariate analysis methods are also considered, and the concept of feature extraction and selection is introduced. The methods are contrasted to earlier standard analysis approaches in the existing literature. Examples are used throughout to illustrate various points.
Power system data are multiscale and multivariate in nature. The increasing availability of wide area measurement systems capable of producing large amounts of multidimensional data has made the use of multivariate data analysis methods more commonplace.
The preceding chapters described methods for analyzing multisensory multitemporal sensor data. Modern wide area measurement systems rely on data assimilation to estimate initial and boundary data, interpolate or smooth sparse or noisy observations, and to evaluate observing systems and dynamical models. The realization of practical data assimilations systems, however, is challenging due to both the high dimensionality of the data and the communication and computational requirements.
Multivariate processes arise when several related time series processes are observed simultaneously over time instead of observing just a single series [1]. Existing wide area monitoring system (WAMS) architectures provide only partial state information; as a result, the information provided by individual sensors is incomplete, inaccurate, and/or unreliable.
This chapter examines the feasibility of using multisensor data fusion techniques for monitoring and analyzing power system oscillatory behavior. A common conceptual mathematical framework for integrating multiscale data to improve situational awareness is provided. The framework includes techniques to classify and extract dynamic patterns from multisensor multiscale data. Outlier detection and methods to evaluate the statistical significance of the results obtained from the different methods are also discussed.
The methods are implemented and compared in terms of their ability to fuse data from multiple sensors.
Accurate diagnosis of system health is a vital step in wide-area monitoring. Advanced event characterization is crucial for improving the detection, identification, and description of system health and the development of corrective measures. Large interconnected power systems and their areas or regional systems are highly complex and variable structures that defy predictions. Monitoring these systems in the face of uncertainty and variability remains a daunting challenge.
The last two decades have borne witness to an explosion of interest in developing power system monitoring and analysis techniques [1]. By monitoring the time evolution of crucial system parameters, monitoring techniques can be used to trigger remedial control actions and alarms and aid in developing situational awareness tools [1-3].
Central to this framework is the diagnostic and prognostic signal processing and measurement techniques used to detect and diagnose power system health [4, 5]. Inappropriate monitoring strategies can lead to irrelevant or poor system characterization, which, in turn, can have profound operational and economic impacts.
Power system monitoring encompasses a variety of activities that involve event detection and classification and assessment of power system health status [6]. The inclusion of spatiotemporal dynamics is needed in order to identify localized and propagating features in measured data as well as to compress system information. It has been realized that these measurements may contain moving patterns and traveling waves of different spatial scales and temporal frequencies [7].
Furthermore, because wide-area measurements are characterized by nonlinearity and high dimensionality, a challenging task is to find ways to reduce system dimensionality to a few modes and link these modes to the underlying dynamical/physical behavior involved.
In this chapter and in Chapter 7, several tools to assess power system health are developed and tested. Methods for evaluating changes in measured oscillatory response are examined, and new approaches for use in wide-area system monitoring are presented.
Issues related to the robustness of the methods in the presence of measurement noise and multiple events are discussed.
Wide-area monitoring of systems with high penetration of distributed renewable energy is complex because of the size of modern power systems and the various transmission and distribution levels at which distributed energy resources (DERs) are installed. The wide geographical spread and continuous growth of DERs increase system complexity, resulting in new stability, operational, and security challenges [1-4]. Simultaneously, this trend is accompanied by significant and often unpredictable changes in power transfers caused by power market policies [5], sudden changes in renewable generation output power [6], and more flexible transmission systems and generation systems, including the installation of HVDC interconnections, large offshore wind farms, and increasing distributed generation, particularly wind and solar PV [7, 8]. Together these issues pose new security and operational challenges that must be addressed using advanced monitoring and control techniques.
Key drivers of these challenges are increasingly variable and uncertain operating scenarios and reduced system strength and inertia caused by the large-scale integration of inverter-based resources. In addition, there is a trend toward a more distributed generation often located in remote system regions.
This chapter discusses different topics in wide-area monitoring of power systems with geographically distributed generations, such as frequency and voltage monitoring techniques, PMU placement in hierarchically distributed power systems, the identification of coherent reactive power sources, and the problem of model reduction of nonlinear and unstable systems. Likely directions for further developments in the field of wide-area monitoring are identified.
Recent studies that use high-order singular value decomposition techniques are also examined to evaluate the performance and accuracy of wide-area monitoring techniques in systems with significant renewable power generation and storage technologies.
Interest in distributed monitoring techniques for monitoring power system oscillations continues to increase due to the enormous scale of modern interconnected power grids and the increased integration of a wide variety of emergent distributed energy resources into the existing monitoring systems.
As modern interconnected power systems grow in scale and complexity and the geographical distribution inherent to renewable generation becomes more prominent, spatial, and temporal patterns analysis becomes more difficult to analyze and interpret [1-5]. Furthermore, more data and complexity will result from analyzing dynamic interactions between subsystems.
Several limitations are associated with the application of essentially centralized structures for monitoring wide-area system behavior. These include: (1) the inability to recognize disturbances and prevent cascading failures, particularly if they co-occur at different system locations, (2) reliability issues associated with the processing of multichannel data, (3) efficiency since centralized monitoring does not account for the geographically dispersed renewable generation that is often monitored separately, and (4) performance as centralized architectures may not take advantage of advances in information and communication technologies [2, 6-10].
These highly distributed energy resources have prompted the research and development of hierarchical-distributed architectures and techniques to store, analyze, and monitor critical system behavior at various transmission levels. A common alternative is to split the observed measurements into blocks of variables exhibiting common characteristics and analyze system performance using distributed monitoring techniques [11, 12]. Ideally, these subsystems should use hierarchically decentralized processes and control monitoring methods, thus indirectly reducing massive amounts of data transmission through the communication network [13, 14].
This chapter introduces the use of distributed wide-area monitoring techniques and provides an intuitive understanding of multivariate analysis techniques. A framework for distributed monitoring of power system oscillations using multiblock (MB) analysis techniques and higher-order singular value decomposition (HO-SVD) is introduced to understand, characterize, and visualize power system global behavior.
Timely and accurate monitoring of dynamic system behavior is essential to improve wide-area situational awareness and system reliability.
The preceding chapter directed attention to formulating analytic methods to assess power system health. This chapter examines the use of near real-time analysis techniques to detect, locate, and characterize power system disturbances and monitor power system oscillatory dynamics.
By tracking key system parameters, the system health can be timely and accurately estimated, and meaningful metrics for monitoring and prognosis can be derived. Techniques to detrend and denoise measured power system data are outlined and tested on actual phasor measurement unit (PMU) data. The critical issues for future research in health monitoring and anomaly detection and prognostics using dimensionality reduction and clustering techniques are discussed. Tools for real-time visualization and monitoring are also reviewed. Emphasis is placed on the development of multiscale, multivariate data analysis techniques.
Examples are used throughout to illustrate various points.
The analysis of multiple sets of data usually of different types or modalities is a challenging problem in power system stability analysis. Examination of system disturbances may involve a large number of measured signals with composite record lengths on the order of several minutes or hours [1-5] and be complicated by noise, trends, and other artifacts.
In addition, comparisons are also needed against model simulations, dynamic probing tests, and previous events [4-6]. Dynamic analysis of measured data faces two major challenges. First, records collected on the wide-area monitoring systems (WAMS) are high-dimensional, complex, and may be contaminated by noise from different sources [6-9]. Second, spatiotemporal data are heterogeneous and highly variable, especially when anomalies resulting from system disturbances change over space or time and exhibit irregular or intermittent behavior.
In this chapter, measured data from an actual system event are used to investigate the ability of wide-area monitoring techniques to monitor and visualize system behavior. Several multisensor data fusion-based forecasting architectures are investigated and tested. The applications covered include the assessment and use of various signal processing techniques in measured synchrophasor data.
Practical methods for obtaining approximations to system behavior are discussed, and the accuracy of the models is evaluated using standard metrics. Visualization techniques are also presented.
The experience in the analysis of collected data from phasor measurement units (PMUs) is presented, and critical factors that influence the system's operation are examined. The issues of data collection, conditioning, and extraction of the primary oscillation frequency are discussed.
No Abstract available.
No Abstract available.
No Abstract available.