Scheduling for PUE - a smart and maintainable tool

Hello everybody,

Today, I’m asking for valuable feedback on a smart tool which might get developed with help from EnAccess. It’s vital to hear from you if this topic could be interesting to your work ― now or in the near future.

The innovation we are discussing is related to smart planning for PUE (productive use of energy). Examples: agro-processing (grain cooling, fish drying, ice making), e-mobility, or even water purification.

These are the questions we are trying to answer:

  1. If you used a similar solution in the past, how does this proposed solution compare? How would you rank it from 1 to 5, 1 being “Much Worse” and 5 being “Much Better”?
  2. What is missing from the description of the listed contributions below to enable companies such as yours or local companies to adopt this innovation?
  3. Can you think of potential scenarios when this innovation would NOT work for local companies? (of course, PUE with flexibility of timing needs to be present, that is clear)
  4. If internet is available on locations you work with: How important is it to run such a smart service in-house, inside your organisation’s IT centre, or would you consider renting this service from a service provider who’d specialise on serving multiple organisations via a SaaS model (African or even EU/US)?
  5. Anything else we should know?

Goals and contributions

The goals of this innovation are to increase revenues by including more PUE in minigrid designs, to reduce energy costs and avoid/reduce over-sizing of grids.

The means to do that is to schedule the start & stop times of PUE processes during the day, so they match local generation and avoid peaks. This requires two innovation parts:

A) A smart scheduling algorithm which plans start and stop times, based on local forecasts and models of the PUE flexibility.

B) A software which implements continuous data pipelines (meter data, weather forecasts, prices, …) and allows operators to view and manage this scheduling.

Part A contains the “smart” work, also called “data science”. We aim to provide several re-usable models of flexible PUE and how to forecast minigrid situations. We know that some pilot projects exists where part A is being addressed, probably to test the potentials of PUE flexibility. These pilots lead to valuable information.

However, Part B is maybe more crucial in this context. It is where we involve the operators in developer companies and the users on the ground. It’s also the part where we really consider how to offer this data-driven technology in a maintainable way and also ensure the replicability of such PUE-scheduling across multiple organisations. This technology should enable microgrid developers to:

  • set up the software and data pipelining with little to no outside help (given working knowledge of processing their own meter data)
  • get the forecasting and scheduling from part A to work ― or plugging in an existing algorithm
  • maintain the system if problems occur, by monitoring problems (e.g. a data sensor changes its format, no weather forecasts available)
  • input/adapt flexibility parameters (e.g. expected arrival times of fish haul, departure times for e-vehicles, maximum temperature for cooling)
  • communicate start/stop times to users on the ground or even to IoT procedures which automate turning devices on and off


PUE and energy flexibility are being discussed among energy access consultants and investors.

For instance, at the recent workshop on “Latest developments in business models for mini-grids”, Trama TecnoAmbiental (working in Rwanda) reported on feasibility studies for 20 different sites and noted that “PUE are key to achieve attractive financial projects.”

Factor[e] ventures recently noted that “an anchor that can be served whenever power is available – which is known as a “dispatchable load,” since its demand for electricity can be served when there is surplus production – is therefore a kind of holy grail for minigrid developers.” They also discuss time-of-use tariffs, which PUE-scheduling can also support (for instance, in Uganda, operators are allowed to charge a tariff different from the national tariff).

Thanks for your feedback and have a great day!

Dynamic scheduling PUE is not a viable option, as many and at the same time the largest PUEs like workshops, mills, etc. need to perform their work and therefore consume electrcity when their customers require them to so. When a little girl with a pot of maize stands in front of your mill and asks for your service, you do not send her home because the algorithm tells you not to use electricity at a certain time.

We indeed identified deferrable loads in minigrids (not exactly PUEs) that can be controlled with an angorithm as suggested above and developed such an algorithm ourselves over the last 2 years. The algorithm has been tested and will soon (in early / mid 2022) be presented to the market.

Any algorithm of this kind must besides the loads also consider the renewable energy projection and the generation system behaviour (battery cycling, etc.).

Hi Nico, I interpreted conversations with minigrid companies and websites like these in the way that the examples I gave are counted towards “productive use”. Glad to be corrected, as I am new in this field.

Obviously, we need to look into those uses with deferrable loads (with what we call “energy flexibility”), but there seem to be a growing number.

Nice to see that there is development in this field. Will your algorithm be open source?

I agree that the algorithm needs to forecast local energy and incorporate what the system is capable of. This is what makes the replicability a challenge. We need to incorporate and maintain local circumstances, but attempt to keep this local onboarding at manageable levels, if such technology is to be used by many mini-grids.