Monitoring Power Quality in Industrial Compressors

Introduction

This tutorial covers the practical application of power and current monitoring sensors to track the power quality of industrial compressors. We demonstrate which parameters are crucial, such as the cosine φ (displacement power factor) and the impact of harmonic distortion, providing a detailed insight into these power quality metrics.

figure 1: Displacement (cos φ) and Harmonic Distortion.

The ultimate goal is to predict potential failures, such as overheating, and offer a gateway to the optimization of industrial equipment.

---

Technical Background and Problem

Industrial compressors, especially high-power models like the 3-phase 200kW compressor used in this case study, can cause power quality issues such as harmonic distortion.

In the upstream transformer cabin, the presence of higher-order frequencies increases core losses due to eddy currents, resulting in elevated transformer temperatures. Special K-factor transformers are used to withstand these heating problems, but the heat losses persist.

figure 2: Eddy currents cause energy losses, such as here resulting in thermal tripping of a generator.

Additionally, other equipment connected to the same power grid, such as small electronic equipment power supplies or capacitor banks, may experience a higher failure rate due to the stresses induced by the unexpected harmonic currents oscillating between the 200kW compressor and the passive components in their power supplies.

---

Sensor Deployment and Capabilities

One of our clients asked to install a power quality monitoring system at the entry point of one of their compressor rooms. The goal of this monitoring system is to gain insight in both the cumulative energy consumption and also the possible infrastructure improvements learned from the captured sensor data.

For this, two sensors were deployed on the compressor room electrical supply, each with their specific focus:

iQunet GridMate AG1 Power Quality Monitor: This LoRaWAN-enabled sensor measures aggregate data such as average grid voltage, RMS and peak current, cosine φ, true power factor (TPF) and distortion power (THD) on all three phases on a 10 minute base interval. This provides us the long-term data essential for the total energy usage as well as the amount of displacement power and the harmonic contents.

figure 3: AG1 LoRaWAN Power Monitor.

Wireless Current Waveform Sensor: Positioned on one of the phases, this sensor captures high-speed snapshots (4kS/s) of the current waveform and its spectrum every 10 minutes. It delivers detailed insights into the time-domain and frequency spectrum and helps to identify the various sources of distortion and intermittent spike events (such as the upstart of the compressor).

figure 4: iQunet wireless current clamp [model ADMOD-CURR].

The combination of these sensors allows for not only real-time analytics but also provides the historical data necessary for the early detection of potential faults. In the next chapter, we will delve into this aspect further.

---

Initial Findings

Within the first few hours, initial data from the AG1 monitor revealed a considerable level of distortion (approx. D=3x20kVArd at 250 Hz) compared to 170 kW of active power at the 50 Hz fundamental. Although the cosine φ seemed to be well-corrected at around PF=0.98, the distortion power was the main factor for reducing the true power factor to around TPF<0.9.

figure 5: Power phasor as shown in the dashboard of the iQunet edge server.

Waveform Analysis and Root Cause

The current waveform sensor (ADMOD-CURR) also identified high harmonic distortion as seen in the system’s spectral footprint, with significant spectral components at the 5th (250 Hz) and 7th (350 Hz) harmonics of the fundamental (50 Hz).

figure 6: Spectrum plot of the compressor current draw, as shown in the iQunet edge computer dashboard.

Additionally, the time-domain waveform revealed the characteristic ripple caused by a 6-pulse 3-phase rectifier at the DC-bus input stage of the compressor VFD.

figure 6: Time-waveform of the compressor current, as shown in the iQunet edge computer dashboard.

In-depth Technical Audit

The 5th harmonic was notably significant, making up about 30% of the main component current in the spectrum plot. From the ratios of the harmonics, it can be determined that a damping choke is used to “just” comply with the IEEE-519 standard when operating at the nominal power of the compressor.

For more details on this see the “ABB Technical Guide to harmonics with AC drives” link .

figure 7: Excerpt from "ABB Technical Guide to harmonics with AC drives".

At power levels deviating from the nominal operating point, however, it is observed that the tuning mismatch of the damping choke results in an increasingly worse THD performance characteristic, leading to non-compliance with even the regulatory requirements.

More information on the problems of the suppression of harmonics with passive components can be found here [powerquality.blog] .

Figure 8: History of Cosine φ and True Power Factor. Although Cosine φ regulation is effective, harmonic distortion (THD) increases as compressor power decreases. The very high THD levels (TPF<0.8) are believed to be caused by 11th harmonic oscillations in the power grid caused by other on-site capacitor banks (further investigation needed).

---

Results and Benefits

Insights

The two sensors did provide immediate, detailed insights into power consumption patterns, enabling a deeper understanding of the root causes of heat losses. The high distortion power is one of the primary issues that must be addressed.

While phase compensation or harmonics suppression with capacitor banks may offer some (marginal) improvements, there is also the very real increased risk of damage caused by unexpected resonances between multiple on-site systems.

The customer has now all the necessary information to calculate whether the cost of a more advanced VFD (e.g. with hybrid harmonic active filter) is justified to improve to the current situation. The decision process will not only involve the heat losses in the upstream transformer cabin, but also the impact of HF harmonics on the lifespan of nearby electrical and mechanical components, such as parasitic bearing currents in rotating equipment connected to the same grid.

See “ABB Technical guide No. 5 - Bearing currents in modern AC drive systems” link .

Figure 9: Bearing damage due to parasitic currents [credits: ABB].

Predictive Maintenance

By analyzing the current spectrum in a regular inverval, the sensors also enable the early detection of changes in the electronic or mechanical behavior of the compressor. This proactive approach can significantly reduce unplanned downtime.

Figure 10: Example of Anomaly Detection using Machine Learning. The step indicates a sudden change in behaviour of the machine, which needs attention but, because the deviation is stable, does not require immediate action.

Machine Learning Applications

iQunet offers an optional service to automate anomaly detection. Small variations in the machine’s operating state cause related changes in the spectral footprint of electical current or mechanical vibrations. While these changes are difficult to detect with the naked eye, a custom-trained machine learning model can provide a reliable early-warning system for critical assets, without the need for expert personnel to analyze the sensor data.

---

Conclusion

This case study has demonstrated the capabilities of modern sensor technology in tackling power quality issues in industrial settings.

By providing detailed measurements, the combination of the appropriate voltage and current sensors enables precise registration of energy usage, along with the identification of the root causes of energy losses due to harmonic distortion currents. For the more experienced user, it offers valuable insights into the installed equipment, down to the ability to determine the VFD characteristics of attached machinery through black-box analysis.

Finally, armed with this knowledge, the customer can take informed steps to implement future operational improvements and monitor emerging anomalies as a foundation for predictive maintenance.

For more detailed technical insights and support, explore our documentation and case studies, or contact our support team.