Time series prediction
Quantifying the cost reduction of geothermal heating.
Geothermal heating reduces the costs of heating. When one is considering investing geothermal heating many times proper benchmarks for the cost reduction are missing and thus even the magnitude of the reduction is based on gut feeling.
This study investigates the impact of renewing a heating system in a Finnish house. In this particular case the old system was waterborne electrical heating system. The new heating system uses geothermal heating as a source of the energy and uses the existing underfloor heating system to distribute the energy.
As the graph below shows the new heating system brings electricity consumption down remarkably. What is the reduction due to the new systems and what is the impact of the outdoor temperature?
In this study machine learning is utilized to define a scenario of electricity consumption if geothermal heating was not deployed. The actual values are based on the measured outdoor temperatures and measured electricity consumptions and the predicted consumption is based on a machine learning model.
Input data consists of the measured outdoor temperatures and electricity consumption data is provided by an electricity company.
Reductions in electricity consumption
Since in a house using electricity for heating outdoor temperature plays an important role in determining the electricity consumption. After the heating system is change how one can say what the electricity consumption would have been if the system was not renewed?
To overcome the challenge machine learning was utilized and a recurrent neural network (RNN) using gated recurrent units (GRUs) was deployed. With the model electricity consumption was predicted for the period when the geothermal heating was operational. The difference of this prediction and the actual electricity consumption indicates the impact of the new heating system as shown in the graph below.
In the graph above smoothen values are shown. The green vertical dotted line shows the time when the new heating system was deployed. The blue line shows the actual electricity consumption in kWh and the red line show the predicted consumption if the new system was not deployed. The shaded area shows the difference (or savings) between the old and the new system. From the graph it is obvious that the impact has been remarkable.
Outdoor temperature is a key variable in predicting the electricity consumption. However, the underfloor heating causes a delay in heating and thus it is also essential to know the outdoor temperatures and electricity consumptions in the previous days. RNNs can take the full benefit of this information when predicting the electricity consumption for the coming days.
The reduction in electricity consumption can be expressed as a reduction in heating costs or as a dramatic drop in CO2 if renewable energy is not used fully.
This study focuses on the impact of the geothermal heating. However, machine learning could also be used to optimize electricity consumptions with the existing heating systems or stabilizing indoor temperature when outdoor temperature changes.