Regarding secondary reticulation, automation has completely changed how Low-Voltage (LV) and Medium-Voltage (MV) switchgears work. A giant leap forward in switchgear technology, automation has simplified operations while improving power distribution networks’ safety and dependability. Notwithstanding these evident advantages, the industry’s conflicting reactions show that adopting new technologies is complicated and fraught with opportunities and threats.
MV and LV switchgears were traditionally handled by hand as a fundamental component for managing power flow and safeguarding electrical circuits from overloads and malfunctions. Manual manipulation is required to control or switch circuits in these mechanical systems under varying loads. The integration of automation in switchgear systems paved the way for the increasing demand for power and the complexity of systems, which necessitated more efficient, dependable, and safe control methods.
As part of the country’s energy policy, a cleaner, cheaper, and more efficient power grid is a priority for India’s government. Concurrently, the impending mobility revolution will significantly affect the grid’s utilisation. Because of this, our distribution grid system will look very different. As for the generating side, renewables will supplant fossil fuels, now the primary source of conventional energy, in due time. As a result, the production of energy is becoming increasingly decentralised. Producing energy in a centralised fossil fuel power plant is reliable and easy to regulate. Wind and solar power, on the other hand, are examples of renewable energy sources that might disrupt production due to their unpredictable and weather-dependent output patterns. There will be both surpluses and shortages in output. Energy demand and consumption peaks will rise due to electric mobility. Mobility is closely tied to the power sector, and some mobility applications have distinct consumption patterns. These changes pose a serious threat to the distribution infrastructure when taken together.
Renewing grid assets increases system flexibility, which system operators must achieve to meet this issue. Although Germany’s grid is now one of the most stable globally, it will ultimately require updates to several essential components. The rise of decentralised power generation and increased electric mobility significantly affects switching patterns and peak performance in the medium voltage range, resulting in an unusually high need for repairs in this sector. We must set up a reliable monitoring system to keep the new and old parts from breaking down.
Grid operators must make network operations as flexible as feasible by improving asset communication and monitoring to achieve grid stability. Switchgear and circuit breakers, two crucial components of power grids, can no longer be serviced at predetermined intervals due to the high frequency and irregularity of switching events. Figure 1 depicts the three basic approaches to industrial equipment maintenance: reactive, preventative, and predictive. Reactive maintenance involves waiting for machinery or tools to break down before acting. Here, the system is used to its total capacity throughout its lifespan, but catastrophic breakdowns might happen, leading to costly repairs and even danger.
The inherent condition of the equipment is disregarded during scheduled preventive maintenance activities initiated by statistical metrics, such as operational hours or elapsed time since the previous maintenance. There exists a risk of over-maintenance, as shown by the excessive lubrication of moving parts, although they often have a substantial operational lifespan at maintenance locations.
Predictive maintenance integrates data from condition monitoring with metrics for system efficiency and other metrics to foresee potential breakdowns or inefficiencies. As soon as there is a change in the monitored status of the equipment, the maintenance schedule is adjusted accordingly. Without fail, this maximises the system’s lifespan. The ideal supply grid maintenance method is predictive since it maximises cost, labour, and environmental efficiency. In contrast to preventive maintenance, predictive techniques rely on intelligent algorithms to find the best and safest maintenance plan. These algorithms may draw from various simulations or artificial intelligence techniques, including machine learning and expert systems. AI is most commonly used in predictive maintenance in the industrial sector.
Three significant issues with medium voltage switchgear may be identified using predictive maintenance principles. The primary challenge is identifying appropriate sensors that consistently and robustly measure the crucial physical parameters throughout the switchgear’s lifespan. Switchgear is also operated in extremely harsh environments all over the world. Therefore, the sensors need to be able to handle that. The absence of quantitative data also presents a problem. Temperature readings taken continuously over the switchgear’s extended lifetime are highly unusual, if not non-existent. Switching operations are carried out for breaker drive monitoring just a handful of times annually, primarily for maintenance. As a result, the measurement data needed to construct AI/ML algorithms is scarce for both application scenarios. The scenario is even worse in breaker drive monitoring since switching operations only take tens of milliseconds. Finding meaningful patterns in the measured data and creating trustworthy prediction systems are thus formidable challenges.
Grid operators must provide and assess precise data to develop an effective Computerised Maintenance Management System (CMMS) for predictive maintenance of distribution grids. This aligns with the goal of operators to raise consciousness about condition monitoring and predictive maintenance. Data capture relies heavily on sensor technology and Remote Terminal Units (RTUs). An example of a cost-effective method for detecting failures early on is temperature monitoring using infrared. Data analysis employs Artificial Intelligence (AI) techniques and a fragmented infrastructure for comprehensive data examination to facilitate and initiate decision support from complex industrial data sets. Condition monitoring and predictive maintenance exemplify innovative, cost-efficient solutions integrating industrial AI methodologies with advanced sensor technology (Figure 2). In addition, GIS-enhanced data analysis speeds up the maintenance procedure.
Figure 3 shows the logic. Both existing and new sensors establish the technological underpinnings of this method. For example, mechanical systems connected to switchgear systems are used as condition monitoring platforms to process the data produced by these sensors. We find out how the various pieces of the switchgear are doing right now by using the platform. Depending on the situation, you may choose what to do regarding operations and maintenance. The data from the condition monitoring platform is processed by predictive algorithms using machine learning approaches. Linking the data to other data sources, such as switchgear or sensors, allows for the prediction of future changes in asset status. The result is the ability to map out a better strategy for maintaining each piece of switchgear. Developing and testing a feasible, scalable business model for the condition monitoring and predictive maintenance of medium voltage switchgear is essential before industrial stakeholders implement such technological breakthroughs.
Distribution Grid Assets and Monitoring Technology
Switchgear is an essential element of electrical infrastructure, serving both as a controller and a safeguard. Switchgear is commonly employed to interrupt an electrical circuit, either for troubleshooting or to modify the circuit. The types of events in switchgear are varied. This article discusses a crucial component of medium-voltage switchgear (Figure 4) and the circuit breaker.
Medium-Voltage Switchgear
Figure 4 shows the layout of tens of panels used in medium voltage switchgear installations within enclosed structures. Air is a common insulating substance, which gives designers more leeway when building and prolonging the lineup. An MVS panel’s primary functions and specifications are as follows: The switchgear ensures safe operation by individuals, is serviceable and compact, allows for disconnection and grounding of components, offers long-term functionality for several decades, and limits the heating of current-carrying elements, effectively isolating electrical failures, such as arc flash, within its confines. These considerations predominantly dictate the fundamental configuration of modern medium-voltage switchgear. The electrical system is contained within a metal container, and interlock devices routinely secure the doors. Segregation walls shield the switchgear from its adjacent switchgear, and a chimney built into the switchgear can divert hot gas from an arc flash. An air blast duct can also be installed on the switchgear. The high-voltage-carrying components (such as insulation, opening and shutting, and current-carrying components) and low-voltage control equipment are often housed in separate sections of switchgear, with the former including cable, breaker, and bus bar sections (Figure 5). They can be removed from the switchgear to maintain or replace primary protective devices, such as circuit breakers. The switchgear incorporates sensors, current and voltage transformers, and additional features. Usually, it has an earthing switch for secure service operations and configuration changes.
A central bus bar system often extends through all panels in a medium voltage configuration. The core system consists of three horizontal bus bars, one for each phase. The central system is linked to vertical feeder bus bars inside each panel for electrical connections to the components within each panel. Each panel has several potential configurations, including incomer, feeder, bus coupler, etc. The precise configuration of a panel is significantly influenced by its rated voltage (7.2 kV-36 kV) and current ratings (630-3150 A), necessitating the use of panels with diverse topologies.
Monitoring of Breaker Drives
The principal role of a circuit breaker in switchgear is to disconnect defective components from the power grid and interrupt fault currents, thus safeguarding the electrical system from damage. Mechanical study identifies drive, linkage, pole, and housing as the four primary components of a circuit breaker (Figure 6a). Spring-driven mechanisms are frequently utilised when the drive subsystem provides the energy for opening and closing operations. The linkage represents the connection between the drive and the pole, containing the electrical contacts interrupting fault currents. The pole is encased in a particular insulating substance, and the metal enclosure encircles the drive and connection.
Research done in India on failure data from electrical components in the medium-voltage distribution grid indicates that the circuit breaker is the most susceptible component of this type of switchgear. The breaker motor and operating mechanism account for almost 90% of mechanical failures in circuit breakers (Figure 6b).
Several potential failure scenarios for circuit breakers are summarised in the IEEE guideline. Possible causes, effects, traits, and monitoring choices are outlined for each failure scenario. Drawing on references, the authors isolate the breaker motor as the primary point of failure for modern medium-voltage circuit breakers. The most recent research on breaker drive monitoring has led to many conclusions. Observing the contact travel time during opening and closing operations is one method to ascertain whether the breaker drive requires maintenance. Additional information on the breaker’s health may be gleaned from the precise measurement of the contact travel. Another common practice is to monitor the vibration signals at a single point on the circuit breaker as it opens and closes. Comparing it to a healthy state allows for the detection of mechanical irregularities. Failure detection can be aided by signal processing approaches such as wavelet analysis and short-time fast Fourier transforms. The condition monitoring and diagnostics of the breaker drive remain unresolved, as per the evaluation of the technical maturity of monitoring options. Many mechanical components, such as joints, bearings, springs, dampers, lever arms, sheet metal, rubber stops, and electrical connections, constitute the complex kinematic chain that connects to the poles and embodies the breaker drive.
Thermal Monitoring
Joule heating is the phenomenon wherein electric current traverses a conductor, generating thermal energy (Figure 7). As the resistance of electrical contacts increases due to various defects (e.g., degradation, loss of connections, or corrosion), their existence can be identified by monitoring temperature. An increase in current produces more heat, which can shorten the lifespan of electrical equipment and hasten its degradation.
Technology: Sensors
Condition monitoring and predictive maintenance are becoming more appealing to electrical equipment and machinery operators. The main reasons for this are to prolong the life of the equipment, save operational expenses, and prevent catastrophic breakdowns. Sensor data recording the required physical quantities is critical for predictive and condition assessment systems. Three things come to mind when thinking of medium voltage switchgear: (1) the current temperature, (2) the gears and mechanisms of the controls and switches, and (3) the level of discharge.
Infrared Radiation Detectors for Remote Temperature Measurement
Among the most critical factors for electrical switchgear design are thermal concerns. Predictive and condition monitoring systems rely heavily on ongoing thermal state evaluations. Surface acoustic wave sensors, radio frequency identification sensors, and wireless sensors are some methods used to monitor the contacting temperatures of electrical equipment. Among the many benefits of contactless technologies are those of Infrared Thermography (IRT). Because it is non-contact and located in areas with low magnetic or electric fields, the measurement is also accessible from electromagnetic interference – so that it won’t mess with the equipment’s dielectric needs. It is also unnecessary to turn off an electrical system to check it. Furthermore, IRT may include a broad expanse, unlike point-measuring sensors. This substantially reduces the necessity for sensors.
The output voltage of thermoelectric sensors, such as thermopiles, is directly proportional to the amount of infrared light detected. They often consist of a series of thermocouples positioned on a fragile membrane. The cold junctions of the thermocouples are positioned atop a heat sink to provide a substantial temperature differential between the hot and cold junctions when the membrane temperature is altered by incoming infrared radiation. Figure 8a shows that thermopiles are mechanically stable and do not need a moving part, such as a shutter or chopper, to function.
Sensors for Breaker Drive Monitoring
The kinematic chain from the working mechanism to the poles is subjected to endurance testing throughout breaker drive development. These evaluations measure travel curves, velocity, torsion, contact pressure, rebound, and vibrations. Consequently, these figures must underpin the reliable and resilient monitoring of the breaker drive. A linear transducer or potentiometer at the pushrod is advised to immediately measure the travel curve and the position of the moving contact. The travel curve measurement for the opening and closing operation of a circuit breaker is depicted in Figure 9a. Determining the breaker drive’s opening and shutting speeds from the travel curve is standard practice. Conversely, rotational transducers assess opening and closing velocities by calculating the trip curve based on the rotation of the primary shaft.
Technological Aspects: Predictive Maintenance using Machine Learning
Much talk has recently been about AI and autonomy in the political, corporate, and trade groups and organisations. As with Industry 4.0, artificial intelligence appears to be the newest industrial craze. It improves industrial systems’ supply chain management, automation, and production optimisation. In addition, businesses are improving industrial system autonomy through AI techniques.
Machine Learning for Predictive Maintenance
What follows is a description of the several common prediction objectives. Predicting the health condition of a system, such as sound, poor, or worse, is one goal; the latter state often denotes a malfunctioning system. A further purpose of this forecast is to determine how much of a system’s usable lifetime remains. Utilise a support vector machine to predict the probability distribution across various health states. An RUL prediction’s associated probability is used to weight a weighted sum of each health state’s average historical RULs, which is then combined with each state’s average RUL of the historical data. Use a multiple-binary classifier strategy, wherein each classifier indicates whether something is healthy for a separate prediction horizon. The chance of unforeseen breakage and the lifespan going unutilised are used to calculate the maintenance cost for each forecast horizon. The RUL that is returned is equal to the least cost prediction horizon. Another approach is to forecast the system’s health index, which reveals how a system is degrading. Obtain a feature representation of the time series using a Recurrent Neural Network (RNN). The k-nearest neighbour method used this feature vector for its training. During the prediction phase, the Remaining Useful Life (RUL) is determined by computing a weighted average of the RULs from the k most similar health index curves identified during training.
Utilisation of AI for Monitoring Switchgear
The considerable volume of research on AI-based electrical equipment monitoring is unsurprising, given the current prominence of AI. Several methods for keeping tabs on electrical machinery use IRT. One example is using binarised IRT pictures in a substation to train a Support Vector Machine (SVM). These pictures feature Zernike moments, which are polynomials orthogonal to the unit disc. Utilise NSCT to enhance IRT images of rotating machinery, inputting various features derived from the image histogram into many machine learning algorithms, including Support Vector Machines (SVMs) and feed-forward neural networks (NNs). Try using a support vector machine and a neural network to categorise malfunctioning phases in switchgear, utilising a set of characteristics retrieved from in-route thermal photos of electrical equipment.
Sneha Kumari is from Dayananda Sagar College of Engineering, Bengaluru, Karnataka, India
Dr. Sujit Kumar is an Assistant Professor in Dayananda Sagar College of Engineering, Bengaluru, Karnataka, India
