Article: Effective Partial Discharge Analysis and Monitoring for MV Switchgear Health Assessment

By Giacomo Ciotti

Introduction

In recent years, advancements in diagnostic technologies have driven a significant shift in maintenance strategies, moving from Time-Based Maintenance (TBM) to Condition-Based Maintenance (CBM). This transition enables more efficient asset management by tailoring maintenance decisions to the actual condition of equipment rather than relying on fixed schedules (Fig 1).

Fig 1 Different maintenance approaches.

Given the increasing aging of many electrical assets and the high criticality of certain of them, such as MV switchgear and cables in specific industrial environments, there is a growing focus on assessing the health condition of their electrical insulation systems. In fact, insulation degradation is one of the primary factors affecting the reliability and longevity of electrical equipment and advanced diagnostic techniques, such as Partial Discharge Analysis (PDA) and PD Monitoring (PDM), are becoming essential tools for identifying potential failures before they lead to costly downtimes or catastrophic faults.

Nowadays, PD monitoring for MV switchgear can be implemented using modern systems that are easy to set up, manage, and capable of triggering alarms only when dangerous PD activity is detected. A critical point for these systems may be the reliability of the automatic data analysis, but machinemonitor® offers an advanced monitoring solution that combines automated analytics with expert human supervision. This hybrid approach leverages the power of AI data analysis while ensuring continuous oversight by experienced engineers. By integrating these capabilities, the system minimizes false alarms and reduces the risk of data misinterpretation, providing a highly reliable assessment of insulation health.

Online PD system composition

Online Partial Discharge (PD) monitoring systems typically consist of three main components:

• Acquisition unit: the core of the monitoring system. It samples and digitizes analog input signals, automatically processes and classifies data, stores relevant information, and, when necessary, generates diagnostic alarms.
• PD sensors: installed on the electrical assets being monitored. For switchgear, these are typically High Frequency Current Transformers (HFCTs), which are placed around ground leads at the cable termination, or TEV antennas, which are attached to external metallic panels.
• Synchronization sensor: provides the acquisition unit with a stable 50 or 60 Hz sine wave reference to ensure accurate signal processing.

Depending on the switchgear configuration, HFCTs and TEV antennas can sometimes be installed without requiring a power outage, minimizing the impact of the monitoring system deployment.

With its advanced computational capabilities, acquisition units are always the most expensive component of an online monitoring system. To address the cost-effectiveness demands of the MV industry, machinemonitor® has designed this unit with up to 32 input channels, reducing the cost per monitored panel as much as possible.

Noise and disturbances rejection during data acquisition and analysis

MV switchgear is used in various sectors of electrical energy distribution, including power generation facilities, mines, refineries, ports, and railway transportation, among others. Despite differences in configuration and installation environments, all these contexts share a common challenge: the presence of electromagnetic disturbances and noise, which can interfere with online PD measurements. To ensure accurate monitoring, it is essential to have hardware or software solutions capable of identifying and filtering out such disturbances.

Machinemonitor’s PD monitoring systems address this challenge by employing software frequency filters that limit the acquisition of noise signals, ensuring the system remains sensitive only to the relevant to PD activity. Additionally, once the data is collected, an AI-driven algorithm distinguishes between real PD signals and unwanted noise or disturbances. This advanced approach allows the system to trigger alarms based not only on the amplitude of detected signals but also on their nature and origin. Since noise and external disturbances do not stress the insulation system, they do not pose a threat to the integrity and reliability of the monitored switchgear, even if they exhibit high amplitude or repetition rates. Conversely, smaller PD signals with growing trend in amplitude and repetition rate indicate potential insulation issues and must be promptly identified and reported to the customer.

Data Interpretation: A steep amplitude increase doesn’t always mean a red alarm

Fig 2 shows an increasing trend in the detected amplitude signals. From a first analysis it might seem like a potentially dangerous evolution that may require immediate decisions. Monitoring systems must instead be able to identify firstly the source of the signals through the analysis of the so-called Phase-Resolved PD (PRPD) pattern and raise alarms only when real harmful PD phenomena are identified.

Fig 2 Example of a steep increase in the detected signal amplitude. The chart shows the signal amplitude and the relevant moving average (within a window of 10 acquisitions).

Table 1 reports the PRPD pattern related to the acquisitions marked as ‘1’, ‘2’, and ‘3’ in Fig 2. From the analysis of these charts, it is evident that the signals are related to background noise and electronic commutation disturbances only. The third column on the right in Table 1 shows the results of the automatic AI signal source identification. These results prove that the monitoring system can correctly identify and discard the disturbances and avoid triggering alarms on harmless phenomena.

Table 1 PRPD pattern charts

Conclusion

Advanced PD monitoring provides an indispensable tool for identifying potential issues early, preventing unexpected failures, and optimizing maintenance efforts.

The system introduced in this article has demonstrated exceptional capabilities in automated data processing. By integrating sophisticated AI-driven analytics with expert human oversight, it effectively distinguishes genuine PD activity from noise and disturbances, ensuring accurate and reliable assessments. This hybrid approach not only minimizes false alarms but also enhances trust in the technology, making it a robust solution for maintaining the health and longevity of MV switchgear in different industrial environments.

Giacomo Ciotti will be joined by a host of electrical engineering experts and HV practitioners at the Machines & HV Assets. View the agenda or the full speaker lineup.