Matheus D. Spinelli
The goal is to automatically itemize the energy consumption down to the specific appliances and then provide specific, actionable recommendations to the consumers.
This application takes in the Green Button data that has your specific electric meter readings pattern over time and figures out automatically what each major appliance in your house consumes. The application does this by correlating specific energy signatures of specific appliances to the electric meter reading pattern over time. This technique is a new theory concept called disagreggation where an aggregate reading is decomposed to its sources through predictive machine learning algorithms. The library of appliance consumption signatures can be developed statically over time as a library. The error in the predictive algorithm can be improved by introducing more data such as outside temperature at that time or the specific time of day. E.g. it is higher likelihood that the heater was turned on during colder temperature and late at night vs. an oven.
Once the itemized measurements (i.e. when each appliance was on and off) are generated then specific recommendations can be generated. For example: 1. Recommend an alternate appliance that could lower consumption based on historical usage patterns. 2. Recommend a different pricing approach (tier vs. peak) 3. Recommend using specific appliances at different times of day based on usage pattern.
Note: The application used excel to do the data analytics and then insert the data into the webpage to generate the visualizations. We did not have time to develop a complete application but wanted to focus on illustrating the concept. Extract zipfile into a folder and open EnergyUse.htm to see the application.
This uses the Green Data and is related to a winning idea of providing real time measurement data.