The power distribution industry is fast moving towards adoption of smart grid technology. It is seen as a major step towards better monitoring of electricity networks including metering infrastructure. However one innate problem remains: monitoring has no benefit without interpretation and without obtaining conclusive information from the data flow. Modern utility businesses now face many unique challenges that were previously unknown. Issues related to quality as well as unaccounted for electricity, either through theft or transmission losses, still present the biggest challenge and financial risk, especially to utilities in Africa, Asia and Latin America.
T&D losses are a major issue with utilities in most countries. The problem seems to be both the collection of data and how to find an effective and accurate way to utilize the data. In a country such as India where there are roughly 170 million connections, detecting the losses is no easy task.
Since almost a decade ago, under the APDRP programme almost all Indian utilities have replaced their electromechanical meters with static meters. However, except for a very few utilities, most of them are still facing the problem of high T&D losses and have no information about the quality of the network and supply status. It is important to note that while one utility is able to reduce its losses drastically, another fails to reduce its losses. Utilities in the state of Delhi, India, have been able to reduce their losses substantially.
The installation of intelligent meters or smart meters (AMI) that can generate a vast quantity of information alone is not sufficient. The mantra for success is to download as much data as possible from the single phase/3 phase/AMR meters and analyze it. Industry experts can justify that data analytics is probably the best way of approaching detection of theft and quality issues in the modern electricity grid. Needless to say it is not possible to do this manually.
As of today, the data is mostly analyzed to generate reports on network health, bulk energy purchase and load patterns of grid, feeders and transformers. Utilities are utilizing data analytics at the grid level or the transformer level only. However, to obtain a true understanding of any situation, we need to extend it to beyond this point and apply it at the customer level. The analysis of data at the consumer side is crucial as it reflects the true end-to-end picture about quality of supply and can further help in planning and reduction of losses.
ANALYSIS OF CONSUMER METER DATA The prime purpose of the consumer meter is to generate consumption data for billing and tariff purposes. However, meters generate a lot of data, which, when combined with logic, can be used not only for billing but also to drive information about:
- Quality of supply at the consumer end
- Quality of installation
- Quality of meter
- Electricity theft
- Technical loss
- Quality of load. Out of all of these parameters, information about electricity theft is the most important in order to reduce losses.
THEFT ANALYSIS MATURITY LEVEL Utilities in general require extensive IT skills to trace electricity theft. The maturity level of analytics is changing with time in order to improve the strike rate and to identify newer theft. Further, it helps in generating data evidence, which can be used to prosecute culprits.
Level 01: Consumption data This maturity level of theft analysis is based on consumption data and trend and is the simplest of all. However, the strike rate is generally low and consumption trend alone cannot be treated as evidence for theft.
Level 02: Comparative consumption data Apart from meter data itself, a lot of information about customers can be obtained from other data sources such as consumption patterns of similar consumers, e.g. two petrol stations or two communication towers, or even by researching information available from the internet to find out the actual activities at any premises.
However, this requires a very high level of experience to carry out benchmarking and effort to collect secondary data. The biggest advantage in this maturity level is that it takes into consideration variations in consumption due to common factors like weather and market demand. This can be the single option for analytics if meters generate consumption data only.
Level 03: Addition of meter abnormality identifications Many newly designed meters directly identify abnormalities and log events caused by theft. This maturity level is based on using such data as well as consumption patterns to identify the theft. In general such leads have a high strike rate, but can only identify theft according to predefined criteria.
Level 04: Meter instantaneous parameters Apart from energy consumption and tamper event logging, meters also log instantaneous parameters such as voltage, current, power factor, etc. In addition to these they also log power on/off data. Theft is nothing but creation of some abnormal conditions in or around the meter, which not only effect consumption patterns but also the patterns of the instantaneous parameters. By analyzing variation of instantaneous parameters one can identify theft. In general, in order to do maturity level 04 analytics, extensive knowledge of metering and electrical engineering is required as it is necessary to carry out a detailed study to identify the theft method through its impact on the meter’s instantaneous parameters. Once such relations are established then it is easier to do analytics and identify theft.
Analytics is nothing but analysis of data to derive information. The main thing is to find relationships between data and information. Filters to sort out consumers involved in theft can be designed based on logic using data related to consumption, secondary data, tamper events and instantaneous parameters. Needless to say this maturity level has a very high strike rate, has wider acceptance by judiciary, and can even detect new methods of theft. However, as mentioned, it requires a very high level of skill to define filter logic.
Level 05: Artificial intelligence (AI) The biggest drawback with all above maturity levels is that one has to define filter logics beforehand to analyze the data and recognize consumers involved in theft. In order to improve the strike rate, one has to review the logic, develop new logic or amend the logic, which requires extensive knowledge. In brief, it is the quality of filter logics and data that decides the strike rate.
The 5th maturity level is based on two basic principles:
- For two events having a similar trend then it is expected that their outcome will also be similar
- Utilities by now hold huge banks of meter data, consumption data and secondary data information on successful and unsuccessful leads as identified through levels 1-4 of the analytics.
The key question is can we use this concept in electricity theft detection, when all we have is numbers? The answer is yes, we can. Electricity theft is born out of human behaviour, which is reflected in the consumption and electricity pattern recorded. Below is the working principal of this maturity model:
- Identify cases where confirmed theft was observed in the past
- Plot the data of various parameters and identify the trend
- Filter by consumers showing a similar trend for a given set of parameters
- Record outcome success/failure into the system to further refine the logic
- Use employee knowledge of metering to further refine the logic.
The biggest advantage of the 5th maturity level of analytics using AI is the self improving nature of the software. Plus any type of data can be used to make the filters. It is equivalent to testing neurons and genes of the human body to find out about cancer and other disease traces.
This concept of analytics can be used not only to identify theft but also to identify faulty meters and quality problems, etc. The logic can even be extended to identify consumers who may default or consumers who will increase their load requirements, etc. The major aim is to develop a system born of experience and continued analysis. Day by day, new ways of tampering with meters come into play so in such a dynamic environment it becomes imperative to develop a flexible system which learns from the past issues, predicts the likely trends for future/unforeseen issues, and keeps learning from its experience to make sense of the bigger picture.
CONCLUSION The installation of smart meters and AMI alone does not guarantee the reduction of T&D losses. As such, this infrastructure will neither generate additional data nor will it carry out additional analytics. The benefit is that it helps in collecting more data in real time.
Using this innovative approach of combining electrical engineering and artificial intelligence, the aim is to detect all the possible abnormalities in the system caused either by theft or technical faults, learn how to recognize those characteristics, and take the appropriate action.
Finally, a word of precaution for those in the process of developing analytics based on meter data: One cannot have the 5th level of maturity unless you have the systems and hands-on experience of maturity levels 1-4. There is great need for analytics, whether based on consumption data, meter data or both. It is required not only to identify theft or revenue loss, but can be used for many applications such as power quality, network reliability, breakdown prediction. etc.
ABOUT THE AUTHOR: Rajesh Bansal is working as Head of Meter Management Group in a leading utility in India. He has a 26 years of experience. Since 1994 he has been involved in the field of designing electricity meter with anti theft features. For the last seven years, he has been working in the field of power distribution, mainly in meter management and theft detection.