An optimization power play: enabling a tomato grower reduce energy costs by smart allocation of energy usage and production

Not surprisingly, Vereijken Kwekerijen saw its energy bills skyrocket, which threatened their overall profitability. They were faced with the challenge of reducing costs while maintaining productivity levels. Energy costs are driven by gas and electricity consumption and revenue generated by feeding surplus electricity back into the grid. Reduction of energy costs can be achieved by finding the optimum usage of cogeneration units. This is currently a painstaking process where different scenarios are manually evaluated with no guarantee of optimality.

With a forward-looking and entrepreneurial mindset, Vereijken Kwekerijen  looking for a data-driven resolution to their problem. They asked Bright Cape if we could help them find a more sustainable solution by leveraging our advanced analytics expertise.

To tackle the challenge, Bright Cape followed a structured problem-solving approach. First, we spent significant time with the client’s subject matter experts to get a very good understanding of the needs, process and challenges. Eventually, the question was formulated as what are the minimum possible costs to operate the lighting and cogeneration units given a broad set of business requirements. The requirements included maintaining the highest levels of both quality and productivity of the TastyTom produce.

Secondly, we needed to identify a suitable analytics approach. Once the business problem was well-scoped, we decided to leverage mixed integer linear programming (MILP) to solve the problem. MILP is a class of linear optimization techniques well suited to address utility allocation problems. We thus built a mathematical model that encodes (i) the objective function of finding the minimally achievable costs and (ii) the constraints. Relevant input data sources included energy market prices (gas, electricity) and internal energy demand.

Then, in order to have computers solve the model, we codified it in a Python program using an open-source linear optimization package. The mathematical program was given the task of finding the optimal daily schedule to power the lighting and the cogeneration units to meet the business requirements at minimum costs. We employed Kedro to build a modular data science pipeline that handles the input data preprocessing, optimization model solving and business output generation.

In the first phase of the project, the model was back-tested showing 10-15% potential cost savings that could have been realized if the recommendations of the model had been implemented. In the next phase of the project, Bright Cape will update the model  to generate optimal schedules based on future predictions of gas and electricity prices.

MILP represents a promising path to achieve holistic optimization solutions to these types of problems. It can support data-driven decision-making in similar energy-related use cases in the agricultural space or different industries altogether (e.g., the orchestration of production assets in a manufacturing environment given both grid and locally generated renewable energy). The knowledge and combined expertise of Bright Cape’s data engineers, data analysts and human data interaction consultants can help you create a fitting solution to your specific matter at hand.