In a recent EEG webinar, three researchers discussed the different types of technology being used to collect data in their EEG projects, and the contribution technology is making to better energy system planning in developing countries. Programme director Simon Trace summarises the main points.

A number of EEG research projects have used different types of technology to collect data and gather evidence. In a recent webinar, we brought three researchers together to share insights from three EEG studies, as well as an Energy Insight paper, where either remote sensors, machine learning, satellite imagery or smart meters have been incorporated into the research. We explored what this kind of technology can tell us about grid outages, electricity demand, and electricity theft, and how it is helping to inform decision making around energy system planning. We also welcomed participation from our live audience, who helped to shape the discussion by asking questions and sharing their views.

Recording grid outages

Dr Noah Klugman, co-founder and CEO of nLine Inc., a company that monitors the performance of national electricity grids by developing and deploying smart and connected sensors, explained that utilities in developing countries generally don’t have enough data to inform them about what's actually happening on the grid, especially at low voltage levels within the distribution network.. When utilities are alerted about an outage, they often don’t know how major it is, when it started, how to prioritise it, what needs to be replaced or what improvements need to be made. Noah explained that data is hugely important; without a picture of what's going on, it's difficult to make the right decisions, whether related to energy policy, investments or day-to-day operations.

Noah has helped to develop, test, deploy and operate a suite of low-cost, affordable, remote sensing devices, collectively referred to as GridWatch, which count and measure outages (including duration), time stamped to the second. The technology has been used in a research project part-funded by EEG to measure power outages, voltage fluctuations and frequency instabilities in Ghana and the causal impacts of reliability on outcomes for households and firms. In one country where GridWatch has been used, it was able to record almost 150 times more low voltage outages than the utility could. In Ghana, it has also been used to measure service reliability across demographics within a district and gather information that could be used to assess the fairness and equity of the management of outages and repair times across different communities served by different parts of the distribution network.. GridWatch sensing devices have also been used in Kenya to collect information on the frequency of service outages on recent grid extensions constructed under different programmes, in order to see if the different contracting processes used had any impact on construction quality and thus system reliability.

Predicting electricity demand

June Lukyu, incoming Assistant Professor in Electrical and Computer Engineering at the University of Washington, explained how an EEG project on electricity demand forecasting in agriculture has combined satellite data and machine learning to help identify areas in Ethiopia where the use of diesel irrigation pumps is prevalent – potentially making these areas a priority for electrification (for electrically driven alternatives to diesel irrigation).

June explained that low demand for electricity challenges utilities’ cost recovery – to be sustainable, electricity connections must generate sufficient revenue to be profitable (outweighing the cost of grid infrastructure). Harnessing latent electricity demand for productive uses of electricity is therefore important – and if consumers can enjoy the benefits of, and profit from, electrification, this could create a win-win situation. Fostering electricity consumption for productive purposes (and thereby improving utilities’ revenue collection), especially in the agricultural sector (the backbone of large parts of the rural African economy), will be key for policy makers to assure long-term sustainability of grid access policies.

The team combined an innovative machine learning prototype simulation model, satellite imagery and classical ‘on-the-ground’ surveys among households, enterprises, and communities to obtain an understanding of areas with elevated latent electricity demand.

Preventing electricity theft

Dr Anant Sudarshan, South Asia Director of the Energy Policy Institute at the University of Chicago, is the co-principal investigator on an EEG project on smart metering and electricity access, being carried out in Harayana, India. The project is evaluating the ability of advanced metering infrastructure (AMI) smart meters with prepayment to break the cycle of low payment leading to restricted and low-quality electricity supply.

As well as being able to provide information about outages (enabling distribution companies to fix problems quickly), AMI smart meters offer utilities much more, including aiding the detection of theft. The technology can triangulate the location where losses are occurring at a much more detailed geographic level than is otherwise possible. Because AMI smart meters offer two-way communication, they offer the potential for non-paying consumers to be remotely disconnected or transferred to pre-payment.

Anant also discussed an EEG-funded Energy Insight he co-authored on tracking India’s COVID-19 impacts and recovery using high-frequency electricity and pollution data (which included the use of satellite data on pollution and night light sources). These variables are both strongly correlated with economic activity and are available with almost no delay (many traditional economic indicators such as employment numbers are generally unavailable at a high frequency in most parts of the world).

Anant explained the usefulness of satellite data on pollution and night light sources during lockdowns, when industries have shut down and fewer people are travelling, and that the quality of the data can be good enough for making policy decisions; understanding where economic activity is steady, recovering, or falling is critical for providing guidance on where to target government aid. The data was used to track the Indian economy during COVID-19 lockdowns and its subsequent recovery, and also provided insight into the variation in impact across different regions.

Data is only useful if it’s used

All three researchers stressed that data derived from remote sensors, machine learning, satellite imagery and smart meters can only be useful if it is being used effectively.

While there is openness and willingness to use data, energy system planners in developing countries might not have the necessary skills and capacity to access and interpret data and use it efficiently.

Investing in the installation of technology such as smart meters can be expensive and needs to be justified in terms of improvements in revenue recovery or reduction of losses. But experience shows that it is not always that case the that all the capabilities and features of a smart meter will be used when installed. This is an important issue to address. For example, installing a smart meter but still using meter readers to manually read and prepare bills, rather than the meters remote billing functionality, means administrative cost savings are unlike to be optimised, while failure to use remote disconnection features may mean the meters don’t result in improvements in reducing losses. Anant explained that, currently, the latter was the case with smart meters installed in Haryana. He also highlighted the importance of knowing what data you want to collect before spending money on technology upfront – because you might find the data you want isn’t what you’re actually collecting.

The webinar highlighted how remote sensors, machine learning, satellite imagery and smart meters are improving data collection and addressing data gaps – which will ultimately help to improve electricity access in developing countries through informing energy system planning – but stakeholders must be supported to ensure that data is accessible and can be used effectively. Continuous engagement between research teams, policy makers and utilities is needed to make full use of the insights being generated. The webinar can be watched here.