Machine learning of heat pump system water usage patterns for optimized heat pump performance

A method of controlling an air-to-water heat pump system for a building having an evaporator that, in use, is exposed to ambient air from outside the building, and a tank system that stores a volume of hot water that has been heated by the heat pump, the method comprising:

• utilizing training data to train a machine learning program to generate a model that is capable of predicting periods of increased demand for hot water in a building;
• utilizing the machine learning program to predict a period of low demand for hot water, predict a period of high demand for hot water that has a greater demand for hot water than the period of low demand for hot water, and monitor a volume of hot water stored in the tank system, wherein the volume of stored hot water in the tank system is determined utilizing at least six water temperature sensors that are configured such that each water temperature sensor has a unique volume of stored hot water associated therewith such that measured water temperatures from the at least six water temperature sensors can be utilized to measure the volume of hot water stored in the tank system to determine at least six non-equal volumes of stored hot water;
• wherein the machine learning program predicts demand for hot water in the building based, at least in part, on data from the at least six water temperature sensors;
• utilizing the machine learning program to implement a load up cycle prior to a predicted period of high demand, the load up cycle including:
• 1) reducing production of hot water by the heat pump to reduce the volume of hot water stored in the tank system;
• followed by:
• 2) increasing heat pump output to increase at least one of a temperature and a volume of hot water stored in the tank system prior to a predicted period of high demand for hot water in a building.