The underutilisation of data continues to be an oversight in many organisations. There are several causes for this, one obvious cause comes from not connecting remote monitors to an IT infrastructure and thus preventing transmission of the data. Another not so obvious cause comes from a lack of awareness of the valuable data is being generated by equipment classified as non-monitoring such as relays and digital fault recorders. Along with the general data captured, advancements in data science have enabled manufacturers to reuse existing parameters measured by a device and extract further information. The information that is extracted from the data generated by such equipment enables organisations to gain insights into the connected critical assets.
The data and information gathered can be analysed alone by subject matter experts or it can form part of a bigger data set. The premise of integrated condition monitoring (ICM) is the combination and advanced analysis of the data generated by separate equipment connected to the same asset. By combining the data from monitoring and non-monitoring equipment a more complete picture of the situation past and present can be created thus enabling a deeper understanding of the condition of critical assets, as well as the root cause and impact of events or grid infrastructure conditions.
Advancements in interoperable data analysis applications and equipment helps organisations simplify information extraction. By utilizing data science and algorithms, software applications can streamline ICM data analysis and focus attention on the information that truly matters thus unlocking the full potential of the data gathered. In this paper, we will explore some examples of how ICM can be achieved due to advancements in equipment and data analysis applications. We will also show how adopting ICM helps organisations improve operational efficiencies.
All businesses share a common goal; implement operational efficiencies whilst maintaining a reliable service for their customers. However, in today’s environment, achieving this goal is increasingly challenging due to the abundance of aging assets across the globe and businesses focused heavily on reducing spend, hence the common phrase “do more with less”.
Data drives the decisions businesses take to implement and achieve the expected efficiency and reliability results. Data derived from the critical components and assets at the core of the service provided to their customers and responsible for their operational process delivery. There has never been more data in the world than exists right now. It is estimated that 2.5 quintillion bytes of data is created each day at our current pace, with that pace accelerating due to new connected devices and the Internet of Things (IoT).
However, this abundance of data must be derived and refined into precise, strategic and actionable information. Data is derived using communication networks and refinement comes in the form of its analysis and interpretation.
In this paper we will focus on the concept of integrated condition monitoring (ICM) which can play a key role in the capture, accurate analysis and interpretation of data. ICM is the mechanism by which the data from multiple sources and assets is captured and analysed together to provide a holistic picture of an assets condition at an individual and fleet level.
ICM also focuses on situational information as part of the assessment in order to better understand the impact of operational fluctuations and external events. When deployed at a fleet level, ICM can provide a more accurate picture of the operational risk exposure of an organisation based on the potential of assets failing and the overall impact of the unexpected downtime as well as repair/replacement costs. ICM can also help organisations develop a successful condition-based maintenance strategy.
Data Analysis Challange
Unfortunately, underutilisation of data continues to be an oversight in many organisations which hinders their ability to achieve their mission statements. There are several causes for this, the most common and obvious cause comes from not connecting remote monitors to an IT infrastructure and applications where it can be fully utilised thus preventing the transmission and analysis of the data at the right time by the right people.
By considering this important point at the start of deploying remote monitoring, this oversight can be easily overcome. If the remote monitoring has already been deployed, fortunately, it is typically straight forward to retrospectively setup and establish communications with the remote devices.
However, in both instances, engagement between the organisation’s IT department, the end users of the data and the vendors is critical. Between the IT department and the end user, an architecture needs to be developed that shows where the data needs to travel from and reside, and to which zones or network substructures it must navigate to get to where it needs to be. Once that is understood, the IT department and the selected vendor or vendors can then develop a topology that can be used to provide the communication path as per Figure 1. This could be as simple as connecting an ethernet cable and opening specific network ports on an existing IT infrastructure or it could require a more comprehensive setup that requires deployment of dedicated wired/wireless communication devices and a network for transmitting the data.
Another not-so-obvious cause of data underutilisation comes from a lack of awareness of the valuable data is being generated by equipment classified as non-monitoring such as relays and digital fault recorders along with other systems. Relays are used to protect an asset when an event occurs by tripping. For many people, as long as the relay trips and prevents damage to the connected asset, the relay has done its job. However, the data captured by the relay during the normal operation of the asset as well as when an event occurs contains valuable information.
Asset monitoring and protection equipment manufacturers are continually evaluating the sensing capabilities and data captured by their relays to seek out new information that can be derived. Advancements in data science have enabled manufacturers to reuse existing sensing and parameters measured by a device to extract further information. The extra information that is extracted from the data generated by such equipment enables organisations to gain more insights into the connected critical assets.
Types of Data
When it comes to ICM there are many different types of data being analysed, below is an example of 4 categories.
- Symptomatic Data – This is data gathered on an asset which is a symptom of an issue rather than the result or cause of a problem.
- Fault Data – This is data captured when a fault occurs causing an event such as a trip, it can be the result or cause of a problem.
- Environmental Data – Data captured about the external environment which the asset is operating over which there is little or no control.
- Operational Data – This is data measured and captured on the operation of the asset such as load, energisation, tap changes, thermal, cooling control etc.
Symptomatic and fault data can sometimes get confused. For example, when dissolved gases start to generate in an oil filled transformer it is symptomatic of several issues such as insulation breakdown, loose connections, overheating, damaged or worn components etc. Whereas fault data would be considered data such as partial discharge, which can be the result of another issue such as insulation breakdown or bare metal contact, but its presence can cause further issues if not addressed.
Symptomatic data can often be a precursor to a serious fault and hence be detected well before fault data is generated and captured. However, symptomatic data can also be caused by benign changes in the operation of the asset, environmental variation or it can even be the expected behaviour of an asset.
Here in lies the challenge, without downtime and invasive internal examinations, how can you identify if the symptoms recorded are an indication of an issue? If so, how serious is the issue, the damage being done, what is the root cause and is the asset in danger of failing unexpectedly?
The good news is by applying an ICM methodology and cross-corelating the multiple sets of data available it is possible to better understand what is happening inside the asset.
ICM Application Example
The following example is a real-world application of ICM that started in 2018. It focused on transformer condition assessment by correlating DGA from a 9-gas online monitor, temperature data from RTDs and electrical data obtained from the protection relay. Thus allowing the assessment of gas values and transformer electrical parameters during different loading conditions, faults, tripping and transformer tap positions etc. providing pre-fault and post-fault records, energization records as well as historical records (learned data records) of electrical, thermal and DGA data (operating history).
The transmission utility had a 150 MVA, 220 – 132 kV power transformer installed and operating since 1976 which was producing abnormal DGA results. Due to the critical loading of the substation, it was difficult to decide if & when to take the transformer out of service or if it can safely continue in service. To help make the decision, the utility piloted an ICM strategy to gain better assessment of the transformer condition and potential faults.
As shown in the Figure 2, the strategy solution comprised of an online DGA device, electrical relay and resistance temperature detectors (RTDs) that collect time-synchronized gas data, electrical data and thermal data from the transformer.
The scope of the study consisted of the impact of the following events on the transformer condition.
a) OLTC operations
b) Trip events associated with transmission lines
c) Operating the transformer on no-load
The list of the events recorded during the study are given in Table 1.
The following diagrams depicts the symptomatic and operational data namely; Electrical and gas parameter trends, tap changer positions as discrete levels (tap position 12 represents higher voltage level while tap position 16 represents lowest voltage level among the tap positions) along with the fault data, namely: Line faults are represented by discrete lines.
The X-Axis of the charts below show the data record and on the Y-Axis is the scale of the various measurements in Current – kA, Voltage – kV, Gas levels – PPM (parts per million) and Temperature – degrees centigrade.
The variations of key gases such as Acetylene, Hydrogen, Ethane, Methane, Ethylene, Carbon Dioxide and Carbon Monoxide in response to the tap positions and line faults are shown in Figure 3. It appears that the tap position has a significant effect on the generation of gases which are symptomatic of arcing, specifically Acetylene and Hydrogen.
The ratio of CO2 to CO is almost constant, indicating no problem with paper insulation.
As can be seen in Figure 4, the moisture variation followed a similar trend as that of the load current. The top oil temperature varied between 40°C and 60°C during the duration of the pilot. The variation of LV side voltage in response to the tap positions and LV side line faults are visible. Tap changes and LV side line faults did not result in any noticeable temperature changes.
The average top oil temperature has decreased as the transformer remained on no-load.
The gas levels were analysed after operating the transformer on no-load. It appeared that load current had no obvious impact on the arcing intensity, as can be seen by the variations in Acetylene and Hydrogen. The concentrations of these gases continued to increase even in the absence of load current, whilst in tap position 13.
Changes in Acetylene levels where low while concentrations of Hydrogen increased when the tap position was changed to 16 however not as much as when tap 13 was used. This difference in gas generation can be seen in Figure 5, indicating the presence of partial discharge. The overall levels of Acetylene were also above acceptable levels.
RESULTS OF THE ANALYSIS
Below are the key findings and recommendations of the data analysis.
- Arcing present in the main tank of the transformer
- Level of Acetylene gas was above acceptable limits
- Arcing characterized by hot metal gases manifested strong relationship with transformer OLTC tap position
- Arcing decreased at the taps with lower voltage levels
- Line faults had no noticeable impact on the arcing and the top oil temperature in the transformer tank, indicating no impact from external events
- Arcing persisted in both on-load and no-load conditions of the transformer Paper Insulation
- Paper insulation appears to be good as demonstrated by the carbon oxide ratio trend
- Moisture in the transformer oil has shown a strong relationship with the load current. Its variation followed a similar trend as that of the load current
- On no-load and on tap 16, PD (partial discharge) was present
Based on the above findings, it is inferred that the possible source of arcing is in the tapings\selector switch of the OLTC specifically the two taps with higher voltages (lower number taps). Thus, the transformer was kept in service using tap position 16 whilst voltage requirement permitted and until alternate arrangements could be made.
There is a vast amount of data and information gathered for assets, captured by an array of equipment connected to the asset as well as external sources. By analysing the data sets separately, the subject matter expert’s ability and opportunity to accurately diagnose the asset is vastly hampered.
The key premise of integrated condition monitoring (ICM) is the combination of advanced analysis of all the relevant data generated. By combining the data from monitoring and non-monitoring equipment along with external data sources, a more complete picture of the assets situation past and present can be created. As was seen in the given example, it was possible to identify the fault by assessing the correlations in the data whilst also ruling out false flags and assumptions that could have been made had the data sets been analysed separately.
Advancements in interoperable data analysis software, equipment and sensors help organisations adopt ICM as it greatly simplifies the integration process and information extraction. By utilising data science and algorithms, software applications can automatically detect correlations in the data and further streamline data analysis. Subsequently allowing the experts to focus more of their attention on the information that truly matters, freeing up more time for strategic planning.
When applied across an entire fleet of critical assets responsible for operational process delivery. For example, the Transformers, Circuit Breakers and Motors used in the day to day operation of utilities and industrial organisations. The results and information about each asset analyzed can be combined to create a more comprehensive fleet overview. This ready to use information pool provides important details on the condition and health of the entire fleet thus allowing organizations to understand their asset risk exposure, providing them with the information to preempt, plan and reduce downtime as well as deploy a successful condition based maintenance strategy reducing future risk.
2) Randy Cox, Knuist Trevor, Rajagopal Kommu, Y. V. Joshi, A. J. Chavda, R. P. Satani, H. D. Solanki. “An innovative solution to assess the Reliability of Transformers by Integrated Transformer Health Monitoring” Cigré Session 48 Paper A2 738 -2020-