Energy consumption has become the main concern of any company or organisation. However, there are services where energy consumption takes a back seat to meet the requirements. This is the case of drinking water services. In order to ensure the supply to the population, until now, the strategy followed consisted of maximising storage, at any cost. This way of operating infrastructures generates energy inefficiency, as well as undermining sustainability and risking greater water losses in the event of breakdowns. Nowadays, thanks to the historical information available, it is possible to generate algorithms capable of minimising consumption while optimising water storage and ensuring water supply. In this work we propose a system through which, using prediction algorithms and a heuristic search for the optimal solution, the optimal behaviour of a drinking water production system is calculated. The system is divided into two blocks, the prediction and the control generator, so that they are able to evaluate the current state of the system and obtain a behaviour adequate to the needs with the minimum consumption. This proposal has been instantiated on a real drinking water infrastructure of a city of 5000 inhabitants in the southeast of Spain, demonstrating the viability of the proposal and the improvement compared to traditional behaviour.