Project Update - Flexcrop

As sensors that monitor different climate aspects and even crops become more and more common, data science becomes increasingly relevant to the greenhouse horticulture sector. One of the promising projects in this area is entitled “Energy saving in greenhouse crop production by flexible management” or “Flexcrop” for short. The goal of this project is to develop greenhouse climate management support that substantially saves on gas, by employing data streams on weather and on the energy grid. This type of management will at the same time reduce the peak loads on the electricity grid, thereby helping the transition to cleaner energy. We spoke with researcher Henry Payne (Wageningen UR) for an update on the data science side of the project.

Veggie batteries

“The goal of the Flexcrop project is to use the flexibility of the crop in the control of the greenhouse,” Henry says. “The plants in the greenhouse are like a battery, the amount of sugars they have being an equivalent to the amount of charge a battery has. Our idea is that when energy is cheap and comes from a renewable source, we charge the batteries. When energy is expensive or comes from non-renewable sources, we switch the greenhouse to passive mode and allow the battery to charge down. Continuously striking the right balance is the challenge we take on in Flexcrop: if we charge down too far, crop yield will be negatively affected. On the other hand, overcharging is not desirable either as this wastes energy."

Two hubs

There’s two hubs in the project: the one focusing on plant physiology is handled by Henry’s colleague Ana Cristina Zepeda Cabrera. Meanwhile, Henry’s focus is on how to use crop flexibility to control the greenhouse more efficiently. The initial scope of his work was to design control algorithms and test them in experimental and model-based studies. However, finding a good greenhouse model turned out to be a challenge. In horticulture, reproduction (or even validation) of a model is often hampered by incomplete documentation. Henry mentions the thesis of his colleague, dr. David Katzin, who addresses this problem and proposes one of the models that is used in Flexcrop.

The search for a suitable model led Henry to other research questions: “I realized that no one had ever researched how the quality of our models and of our data affects the quality of our predictions. For example, we use the weather forecast to predict how much energy to buy in the future, but no one has ever looked at how the quality of the weather forecast affects the quality of our energy buy-ing.” Next to the uncertainty in the input data, there is also a degree of uncertainty in the model itself, Henry explains: “We select parameters that we train the model for, but there is a lot of elements in the model that we cannot observe or measure.”

Uncertainty analysis

Realizing the potential impact of the two types of uncertainty, Henry ended up focusing on uncertainty analysis - a topic on which there is a lot of ground to cover in the horticulture sector. “Models are inherently deterministic,” he says, “but they are based on a lot of assumptions. These assumptions introduce uncertainty in the predictions one makes. I try to assess what impact an assumption can have on our predictions. This quickly gets complex, as we deal with many steps in Flexcrop: from weather prediction to temperature and light in the greenhouse, to calculating how much energy we would need to add, to what that energy would cost. Each step increases the uncertainty.

Assessing the impact of this uncertainty is highly relevant for both academic purposes and from a commercial point of view: decisions on significant amounts of power sequestration are based on these models. Quantifying the effects of the uncertainties in data and models allows for making better decisions, for example on how risk is assessed, how much margin is appropriate et cetera.”

Collaboration

Henry enjoys the collaboration with several of the project partners (Wageningen Research, Glastuinbouw Nederland, Agro Energy, Delphy, B-Mex, Blue Radix and Letsgrow.com). “The atmosphere has been very open and constructive. Some partners provide data from their horticultural installations. Others provide expertise on the practicalities of greenhouse production and on the hard mathematics. A lot of this stuff is highly technical, so we really need the expert advice of different parties.“

Next steps

“Right now, the two hubs of the project are coming together”, Henry continues. ”The plant physiology experiments have yielded data that we can use in modelling and control, and I have assessed the uncertainty of these models. Now we can apply both together. To come back to the battery metaphor: we can now assess how big the battery is and address the central problem of the project more directly.“

“With Flexcrop, we not only add models to save energy, but we include a nuanced look at how certain we are about the amount of energy we can save. This makes the outcome of our models more relevant in horticultural practice in terms of purchasing decisions and energy management. The horticultural sector is shifting gears to meet climate goals. A lot of work is being done to make greenhouses more efficient – I believe our project to be a valuable addition.”