Around the world, economic activity is hampered by unreliable electricity access. If electricity is not usable in practice (e.g. due to frequent power outages or poor voltage quality) businesses often suffer from low productivity and the costs incurred for energy coping strategies (e.g. purchasing and operating a generator). Research efforts, such as the World Bank Enterprise Survey, confirm that stable electricity is consistently linked to higher business productivity. Year in and year out, government electrification agencies and investors budget for grid expansion, grid upgrades, and/or mini-grid installation to boost power quality where the grid is unreliable. A data-driven view of the power reliability status quo is crucial for identifying areas needing new or modified energy systems versus those well-served by existing infrastructure. This data is largely inadequate or fragmented (i.e. data might exist but it’s aggregated at a high-level or is not publicly available).
With support from Nigeria’s Rural Electrification Agency (REA) and in collaboration with researchers from the e-GUIDE Initiative, nLine’s sensors are gathering data on power reliability and quality in 78 markets across Nigeria. This blog shows how power quality data can quantify electricity reliability issues within and across markets to inform electrification planning. Ultimately, this data can serve government electrification agencies, power providers, and investors alike to inform and evaluate future energy infrastructure investments.
Targeting Markets in Nigeria for Reliability-Focused Electrification
Nigeria, the largest economy in Africa, has ambitious plans to reform its power sector and increase electrification. One piece of this strategy is the federal government’s Energizing Economies Initiative (EEI), which aims to provide electricity solutions to small and medium-sized enterprises (SMEs) in “economic clusters” such as markets, shopping complexes, and agricultural/industrial clusters. In many developing regions, SMEs in the informal sector play a major role in the economy. Nigeria is no exception: these enterprises employ millions of people. Providing affordable and reliable electricity to these marketplaces is important for continued economic growth.
.jpg%3Ftable%3Dblock%26id%3D4a07d996-46b2-4019-b9f9-14c3cf8605be%26cache%3Dv2&w=3840&q=75)
Through partnerships with private companies, Nigeria’s REA hopes to supply reliable electricity to 200,000 small businesses in 340 economic clusters across the country. The crucial question for electrification efforts is: which economic clusters should be targeted for energy system investments to maximize economic benefits? The Electricity Growth and Use in Developing Economies (e-GUIDE) Initiative, a Rockefeller Foundation-funded partnership across five U.S. universities, seeks to answer this question by providing data that will enable REA to improve electricity system planning under EEI. Many factors determine the economic impact of installing a new energy system for a particular “economic cluster”. One critical factor is the reliability of electricity currently available in the market, including temporal patterns in reliability (does power go off during the busiest times of market activity?) and overall reliability metrics (how many power interruptions on average does a shopkeeper experience and how long do these interruptions last?). This is where nLine enters the picture.
In this blog post, we show how nLine sensors can measure electricity quality (voltage and frequency) and reliability (number and duration of outages) at sampled shops in chosen markets across Nigeria. This high-resolution data can be used along with other available power infrastructure data, such as Nigeria SE4ALL’s web map Power Sector Analytics Explorer, to perform detailed, sophisticated analysis of the power grid’s performance. Electricity system planners are often asked to design infrastructure solutions that improve electricity without the requisite data to inform their decisions. nLine’s data begins to close the reliability data gap and helps government electrification agencies fulfill their critical mandate of providing reliable electricity for all.
High-Resolution, Real-Time Power Reliability Data in 78 Markets
Of the originally planned 340 “economic clusters” targeted under EEI, REA developed a list of 80 priority sites across the country. We co-designed a data collection strategy with e-GUIDE for these priority markets that included both the installation of nLine sensors and in-person baseline energy audit surveys with market traders. This baseline energy audit survey aimed to gather detailed information about the current energy profile in each market, including a shop’s electricity source, appliance inventory, and the owner’s/operator's overall experiences and challenges with electricity.
Beginning in March 2022, a private survey firm in Nigeria, Vista Advisory, led on-the-ground data collection in 78 markets to date.
.jpg%3Ftable%3Dblock%26id%3D24080565-d976-4cc9-a291-830b3a3b14cd%26cache%3Dv2&w=3840&q=75)
.jpg%3Ftable%3Dblock%26id%3Ddccaefff-5a43-497b-a1b1-bbbf0bdba342%26cache%3Dv2&w=3840&q=75)
.jpg%3Ftable%3Dblock%26id%3D0bfce898-b909-4180-9240-a03dc82cca6b%26cache%3Dv2&w=3840&q=75)
Our data collection approach included installing one nLine sensor in five randomly selected shops within each of the 78 visited markets. These five shops were randomly selected based on pre-qualification criteria including spatial distribution across market sections, electricity connection at the shop, consent of the shop owner, etc. The sensors remained plugged in and gathered data while energy audit surveys were conducted across the market. Once the surveys were complete, sensors were unplugged and gathered to install at the next market. Given the range in market size (from 9 to 13,691 shops), the time that sensors stayed plugged in varied significantly depending on the time required to complete the market energy audit surveys (from 2 hours to 12 days).
This rapid deployment and collection of sensors tested our ability to 1) provide reliable, real-time data from markets across varying geographies in Nigeria and 2) integrate within Vista’s energy audit survey. Supporting such a rapid sensor deployment was novel for nLine: in our previous deployments across 5 countries, sensors were plugged in at the same location for at least two months to reveal temporal trends and ensure sufficient amounts of data to derive key energy metrics such as SAIDI and SAIFI. We approached this project in Nigeria as an opportunity to push our sensor deployment strategy to its limit and determine what power reliability insights could be derived from a rapid remote-monitoring strategy.
Here, the analysis carried out with the voltage and power outage data (collected at different times between March 2022 and October 2022 across these markets) illustrates a compelling application of nLine’s data: data-driven market selection for energy supply augmentation to promote economic productivity.
A Tale of Two Markets: The State of Power Quality and Reliability in Wuye and Ngwa Road Markets
Before diving in, let’s define the two core components of what is considered “good power”:
Power reliability refers to the dependability and consistency of the electricity supply. Having access to reliable electricity entails having a stable and uninterrupted flow of electricity to meet daily needs. The higher the power reliability, the more we can rely on the availability of electricity whenever we need it, without frequent or prolonged interruptions.
Power quality is a measure of how useful the available power is. In order for electricity to be useful, it is essential that the voltage and frequency at which it is delivered remain within an appropriate range. The specific range may vary depending on the nature of the electrical load being supplied. However, maintaining voltage levels within ±10% of the nominal value (230 V in Nigeria) is a commonly accepted standard. Deviation from this recommended range can have detrimental effects on the functionality and longevity of electrical equipment and appliances. Voltage fluctuations beyond the prescribed limits can result in disruptions, overheating, damage, and a decrease in the lifespan of various devices. Reducing occurrences of damaging voltage spikes, sags, and frequency harmonics is essential for delivering high-quality electricity.
Here, we’ll highlight two markets that we found experienced very different ends of the electricity quality spectrum: Wuye Market and Ngwa Road Market. nLine’s high-resolution temporal data reveals how voltage drops to unfavorable levels during peak business hours at Ngwa Road Market and the many momentary interruptions experienced during business hours at Wuye Market. But this story also has a larger context. By aggregating high-resolution data into standard metrics of power reliability, nLine data can begin to paint a picture of the energy needs in a market. These metrics can be applied across the 78 markets to provide a longitudinal and broad view of energy needs across the country.
While reading this blog, try to use nLine data to answer this question that energy planners like REA work to answer, often with very limited input data: “Do Wuye and Ngwa Road markets require an investment in their power supplies to promote economic activity, based on the data we have?” With that motivation and this question in mind, let’s dive into the data collected from Wuye and Nwga Road markets in the context of site selection for energy supply augmentation.
It is important to acknowledge that our current data analysis pipeline does not yet distinguish between different energy sources; our approach does not filter out data originating from generators or other power sources. Therefore, in the subsequent analysis, we will be examining all voltage and outage data that we received which combines both generator and grid measurements.
Wuye Market: Good Quality Voltage with Few and Short-Duration Power Outages
Beginning in August 2022, 5 sensors were installed in 5 randomly selected shops at Wuye Market for an average duration of ~7 days. Within these 7 days, sensors collected and reported data to nLine’s back-end cloud systems for analysis.
The core data stream provided by nLine sensors is the RMS voltage measured every two minutes at the outlet where the sensor is plugged in. This single data stream captures a variety of factors we care about, such as the occurrence of over and under-voltage and the time and duration of outages. Let’s start by diving into the voltage quality analysis.
Figure 6 shows a time series of the average RMS voltage collected at Wuye Market. This plot reveals good voltage quality: when there was power, more than 98% of the voltage values recorded fell within the advised ±10% tolerance window of the nominal voltage.
.png%3Ftable%3Dblock%26id%3D92a87e14-393a-47ae-b23a-3b99f4f1d7f8%26cache%3Dv2&w=1920&q=75)
Because this market operates at only specific hours of the day (roughly 8:00 am to 6:00 pm), it is crucial to examine the nature of the voltage quality at peak market hours. This helps answer the question: “Is the voltage good enough when it is most needed by shop owners?”.
Figure 7 below is a plot of the hourly distribution of all voltage values recorded at Wuye Market over seven days. During the night (off-market hours) sensors recorded high voltage values, whereas, during market hours the voltage dropped slightly. How can we explain this trend? This trend is most likely caused by increases and decreases in load: at night, shop owners are less likely to be operating their appliances, disconnect devices as a fire prevention measure, and may even turn their power supply off. The decreased electricity demand means a decrease in current flowing along the electric lines, which in turn implies less voltage drop along the line. This can result in higher voltage throughout the market during non-business hours.
.png%3Ftable%3Dblock%26id%3D7d0c10b9-ada0-451d-b4d9-f08bfd1d396d%26cache%3Dv2&w=1920&q=75)
Overall our results indicate that the voltage quality in this market was good for the data collection period. This alone, however, does not constitute good power (i.e. being practically usable) since we also need to consider reliability (i.e. does the power shut off during market business hours?). To accurately detect power outages, we use a clustering algorithm that ensures that for a 0V voltage reading to be categorized as a power outage, at least two sensors must report this same value within the same space and time window.
Using this clustering algorithm, nLine’s sensors detected and recorded 15 outages during the 7-day deployment with an average outage duration of 47.88 minutes and a median of 4.86 minutes (note: this is inclusive of all power outages, including grid and generator). Per the IEEE's 5 minutes or less time threshold for momentary outages, this median value reveals that most of the outages were momentary interruptions. Only one out of the 15 outages lasted 9+ hours. A more in-depth analysis indicated that, though the outages were short-lived, about 53% of the outages that we measured at the five shops occurred during peak business hours.
What energy solutions are being used by these shop owners and what could be a potential cause of these outages? All five monitored shops are connected to the main grid, and four of the five shops also receive backup power supply from a personal generator (switched on manually). We can speculate that the frequent and short outages at Wuye Market might be due to a particular piece of power supply equipment malfunctioning, or nuisance circuit trips when the load increases; issues which are perhaps quickly resolved due to the relatively small size of the market and/or a more responsive local grid staff. However, more information is needed for a definitive diagnosis of outage causes.
.png%3Ftable%3Dblock%26id%3D079947e8-0a36-4a33-b702-344439b30bd7%26cache%3Dv2&w=1920&q=75)
Wuye Market - Summary of Findings
This single week of nLine sensor data from Wuye Market shows that most recorded voltage levels fall within the recommended tolerance range. The majority of outages were momentary interruptions as shown in Figure 8. While one week of data is insufficient to fully understand power quality and reliability — from this small dataset we can see that voltage quality at Wuye Market is stable, and daily power outages are short. However, only continued longitudinal data collection can give us more robust insights into the energy profile at Wuye Market. With longitudinal measurements, we are able to disentangle how factors like seasonality and increase in connections may compromise the observed voltage quality or increase the frequency of outages.
Ngwa Road Market: Low Power Quality and Reliability
Next, we will examine data from Ngwa Road Market, one of 78 markets that has a vastly different power quality and reliability profile than Wuye Market. In March 2022, Vista installed 5 sensors in 5 randomly selected shops at Ngwa Road Market for an average duration of ~7 days.
Figure 9 shows a time series plot of voltage data measured at Ngwa Road Market. A glance at this voltage plot looks concerning: we see highly variable, consistently low voltage. Such unstable and out-of-tolerance voltage levels can have serious consequences, including the malfunction of electrical appliances, which directly increases the cost of business operations, damages equipment, and has a detrimental effect on the profitability of enterprises.
.png%3Ftable%3Dblock%26id%3Da81d33c7-8782-480c-8b31-a36b7e09e87b%26cache%3Dv2&w=1920&q=75)
An in-depth analysis of voltage during peak business hours shown in Figure 10 confirms the finding that inadequate voltage persists.
Why is this different from Wuye Market, where we observe usable voltage levels during business hours? Ngwa Road Market is one of the largest markets in Abia state and it houses substantial leather and textile production and repackaging. One possible explanation is the market’s electrical infrastructure is under-sized for these large loads, leading to the observed voltage drop; as more devices come online and the load increases during the day, voltage continues to degrade.
Just as we saw with Wuye Market (Figure 7), the voltage at Ngwa Road Market follows the expected pattern of decreasing during business hours as the load increases, then slowly recovering in the evening as the load decreases. Though Wuye and Ngwa Road markets have similar voltage trends during business hours, the voltage levels observed at Ngwa Road Market are not business-friendly.
Now let’s turn to power reliability. Compared to 15 power outages measured at Wuye Market (~5 minutes median duration and ~47 minutes average duration), sensors measured 6 power outages at Ngwa Road Market with median and average durations of 12 minutes and ~5.7 hours, respectively.
.png%3Ftable%3Dblock%26id%3De08d1347-2a20-4d7b-a922-9ffcad21941f%26cache%3Dv2&w=1920&q=75)
Figure 11 shows the distribution of the durations of 6 outages recorded at Ngwa Road Market. Further analysis revealed that all the outages happened during peak business hours, with one outage resulting in a day-long blackout. While there were fewer outages at Ngwa Road Market, they lasted longer than those at Wuye Market.
Ngwa Road Market - Summary of Findings
In summary, we find that “sustained power interruptions” (according to IEEE standards) occur in Ngwa Road Market and that the quality of the power supply at the time of data collection may be detrimental to business operations. Such conditions often force business owners to resort to using diesel generators (which could mean incurring significant additional operational costs of about 3.8 times the cost of using grid electricity every month) and often result in a drop in productivity and sales (for example, some businesses in similar circumstances have observed as much as a 31% drop in sales due to erratic power supply).
Broader Power Quality Reliability Insights Across 17 Markets in Nigeria
To enable broader insights into the state of power quality and reliability in markets across Nigeria, we expand our analysis and intuition from Wuye and Ngwa Road markets to produce a standard set of power quality metrics surveyed in markets. As mentioned previously, nLine approached this project in Nigeria as an opportunity to push our sensor deployment strategy to its limit and determine what power reliability insights could be derived from a rapid remote-monitoring strategy, where some sensors collected data for only a few hours. Such short sensor deployments can still provide insights but are also at risk of underestimating or overestimating power quality and reliability (for example, if a sensor happens to be deployed during a long power outage or peak demand). To capture this uncertainty, we only provide standard power metrics for 17 of the 78 markets that had the highest quality data (i.e. a deployment of at least 4 days where at least 3 of the 5 sensors successfully transmitted data).
This section summarizes how we can study power quality and reliability trends across numerous markets. Figure 12 below captures the voltage distribution in each of the 17 selected markets. The skewness of the boxes below the green area suggests that low-voltage conditions are generally more common than over-voltage conditions.
.png%3Ftable%3Dblock%26id%3Da24ac68d-9e6c-4520-8d8f-40a0cb352ac6%26cache%3Dv2&w=1920&q=75)
From Figure 12, we can identify markets that exhibit high quality, usable voltage with rare instances of low voltage (e.g. Abubakar Gumi Central and Wuye markets) and other markets that frequently experience highly unstable voltage below standard tolerance values (e.g. Potiskum Central Market).
.png%3Ftable%3Dblock%26id%3D727408db-833c-46f5-965e-eff31034e143%26cache%3Dv2&w=1920&q=75)
.png%3Ftable%3Dblock%26id%3D23e2dd17-71e6-4b88-b81b-655b25237b0c%26cache%3Dv2&w=1920&q=75)
Put together, these KPIs (voltage quality, the number of outages, the duration of the outages) can pinpoint which markets are most impacted by power quality and reliability issues.
Conclusion
The comparative narrative analysis above aims to demonstrate how nLine’s data collection and analysis can provide granular analytics on grid performance in economic clusters. When this information is combined with other available power sector data, electrification planners such as Nigeria’s REA can better identify markets in dire need of power supply improvement to strengthen productivity and economic growth. Finally, nLine’s data can be paired with descriptive, qualitative information gathered from markets to fully understand why power outages and voltage levels are as they are, as well as their human and business implications. This can paint a more complete picture of the nature of power quality and reliability in markets and assist electrification planners in decision-making.
We would like to thank Nigeria’s REA and their Managing Director, Ahmad Salihijo, for their support provided throughout the duration of this project. We give special thanks to the core team at e-GUIDE who led this project and supported nLine’s work in Nigeria: Jay Taneja (Assistant Professor at the University of Massachusetts - Amherst and PI for this project), Deborah Braide (Nigeria Research Coordinator for this project), Civian Kiki Massa (Ph.D. student at UMass-Amherst and Research Assistant for this project), Joel Mugyenyi (Ph.D. student at Columbia University and former Research Coordinator for this project), Nathan Williams (Assistant Professor at Rochester Institute of Technology), Vijay Modi (Professor at Columbia University), and Courage Ekoh (Ph.D. student at Rochester Institute of Technology). We also wish to thank Vista Advisory for their support in installing and managing the fleet of sensors in all markets.
Appendix: A post mortem of the agile sensor deployment methodology
The rapid deployment methodology utilized for this project was unorthodox to nLine’s data collection and exploration processes. It’s therefore important to examine how this new approach influenced measurements and insights.
.png%3Ftable%3Dblock%26id%3Dd508eb99-e2d2-41e4-85f4-41274e8a0092%26cache%3Dv2&w=2048&q=75)
Figure 14 above shows the clear variability in the average deployment duration of sensors across the markets, ranging from ~2 hours at Abraham Adesanya Shopping Complex Somolu Lagos to ~12 days at Marine Modern Market. Most markets have a deployment duration of less than 1 week. The approach of moving sensors from market to market enabled fast and quick data collection across many markets in a short time. However, the difference in deployment duration across markets presented several statistical challenges when analyzing the data. Some of these are outlined below:
- Inability to Identify Data Patterns With Short Deployments
A unique challenge of this project in Nigeria was the extremely short deployment durations. Evaluating a market's power state requires analyzing voltage behavior over an extended period. In this deployment, some markets had only 3 hours of data collection. This makes deriving insights challenging. What if the power state during those 3 hours does not reflect the market's typical conditions?
Such short deployments provide insufficient data to identify patterns or draw conclusions. The limited data from short deployments prevent determining if observed conditions represent the norm or are anomalous. More data over a longer timeframe (e.g. weeks or months) is necessary for accurate assessment, meaningful analysis, and valuable recommendations.
- Non-overlapping Sensor Deployment Periods
Extreme weather conditions can have detrimental effects on power stability and reliability. The time-staggered nature of this project in Nigeria, where markets were visited one after the other over a period of many months, meant some markets were visited during the rainy season (liable to thunderstorms and lengthy outages) and others during the dry season. For example, data collection occurred in March at Ngwa Road Market and in September at Wuye Market. When analyzing power quality and reliability data, it is essential to pay close attention to trends and seasonalities. A parallel deployment, where sensor data collection happens during the same month(s) across all markets, would allow for more comparable weather conditions across markets and clarify the impact of weather on observed differences in the data.