Image: Empa/ETH

Local power grids made to measure

In our highly digitalized world, it is easy to forget that electricity is not only responsible for the glowing displays of our numerous gadgets, but also makes healthy, clean living spaces or even access to education possible in the first place in many parts of the world. Many developing countries are stuck in a vicious circle of poverty with their low electrification rates. Without lighting in the home, there is a lack of opportunities for value-added off-farm work. Children can no longer do their homework or learn to read in the evening. In addition, there are health problems often caused by smoking fireplaces in the house or sooty kerosene lamps.

Access to clean energy is generally seen as a stepping stone to generating a higher income and thus escaping poverty. This is why it has been identified as one of the 17 UN Sustainable Development Goals. Following on from this goal, Cristina Dominguez, a doctoral student at ETH Zurich’s Institute of Building Physics and Empa’s Urban Energy Systems Lab, is developing a computer model that will provide project developers in rural areas with estimates of household electricity requirements. This should enable accurate and thus sustainable planning of the power grid.

One model for many regions

This is because electrification projects in developing countries often fail because hardly any reliable data is available to determine the needs of the often widely scattered households. Data collection in particular is a major cost item that makes project developers hesitant to invest there. If a power grid is then planned too large, for example, this is passed on to the electricity prices, making electricity unaffordable for the poor population. Ultimately, electricity grids need to be tailored to benefit people in the long term, on the one hand, and to offer developers an attractive and realistic investment opportunity, on the other.

For the data collection, Dominguez chose an area in sub-Saharan Africa, the region with the lowest electrification rate in the world: “In addition to the political problems, the areas here are extremely sparsely populated and the small settlements are very widely scattered. This makes electrification much more difficult – and of course more expensive,” says Dominguez. As part of her doctoral thesis, she determined the energy use and demand of around 250 households in eastern Kenya. In order to eventually make their model applicable worldwide, they are supporting research institutes in Guatemala and Pakistan to provide her with equivalent data sets from these countries.

Thinking along developmental thrust

Dominguez’s fieldwork in Kenya collected data from households without access to electricity and those that had been connected to an electricity grid within the last six years. It was not only concerned with recording existing energy sources and their demand, but also the change in use following the completion of electrification. In addition, the Empa researcher used diaries in which the residents recorded the activities they carried out throughout the day in order to gain a better understanding of people’s daily lives and needs, and to anticipate changes that would set in after electrification and would then be reflected in the demand for electricity. In Kenya, for example, kerosene is an important source of energy to light the dark mud huts. In order to get the kerosene, often longer marches must be covered to the dealer. Time that could perhaps be invested in value-adding work at home in the future – if a power source were available.

And once the power supply is available, people begin to adjust their behaviour accordingly; they acquire electrical appliances such as televisions, and power consumption increases accordingly. But how long can the power grid continue to function if demand continues to rise? These are exactly the dynamics Dominguez wants to bring into her model: “In our on-site surveys, we asked people what appliances they would buy after the first year or second year with electricity. We then matched this with households that had already gone through this process.” Through this, Dominguez wanted to find out how people would use energy when it was available to them. Dominguez knows from her research that engineers are often unable to assess this correctly: “There are major prejudices here, which often result in power supply systems being designed too large.”

Tailor-made power grids

To make accurate predictions and identify consumption dynamics, Dominguez applies “machine learning” algorithms and “data mining” techniques. To build the models, the researcher combines global datasets from organizations such as the World Bank with data from project development companies so that she can include additional consumption patterns such as seasonal variations. These are then validated for the three focal regions using field data from Kenya, Pakistan and Guatemala. Mini-grid companies have also provided their electricity consumption data in return for the opportunity to test their model against local conditions.

Cristina Dominguez’s approach highlights the problems faced by developing countries with hardly any infrastructure: Although technical possibilities for electrification are available and have also become cheaper with solar technology, investments must be made carefully in a weak economic environment. Otherwise, there is a risk of over-indebtedness on the part of the electricity users and, in the worst case, on the part of the operating companies – all in all, a risk of exacerbating poverty and a disincentive for others to invest in these areas. In Dominguez’s computer model lies the potential to overcome at least one hurdle to electrification and thus provide the impetus for a path out of poverty.