Practical application of AI and Machine Learning

Important results from the project

The implementation of the project has increased the interest in AI within the own organization. To facilitate future AI projects, all insights and obtained knowledge regarding implementing AI has been documented and communicated to stakeholders through webinars and presentations. The project´s self-developed AI model succeeded in differentiate a number of power-consuming units within a household based on the electricity meter´s collected measurement data at a high-resolution level. The result was above expectations given the amount of work spent.

Expected long term effects

To facilitate future AI project, the project result have clarified the value of useful data and the importance of starting the collection of this as soon as possible. The expectation is that more projects can come to a result more quickly as the project´s insights and experiences help them in the right direction. The project also succeeded in identifying the energy consumption of some specific units within a household, but to see a larger area of use for the model, more units should be identified where heating is seen as an important factor.

Approach and implementation

The project was based on existing studies in the subject. The studies highlighted above all limitations but also insights into algorithm selection and data collection for de-aggregation. The project chose to use the UK-Dale dataset to train the algorithm. Our own high-resolution datasets were used to test and verify the algorithm. The project was staffed with an external AI expert who, together with internal staff, work incrementally to gradually increase the accuracy and quality of the model. Knowledge transfer regarding insights and project results were made through webinars.