Energy management and smart thermostat learning methods and control systems

1. A method of HVAC system performance monitoring using a computing device connected to at least one thermostat of an HVAC system in a building including a fan, the method comprising:

receiving thermostat data from the thermostat, the thermostat data including temperature setpoint data, measured building temperature data, and HVAC operation data for an initial time period, the HVAC operation data including usage data for the HVAC system fan;

receiving weather data from a weather service for the initial time period;

synchronizing the thermostat data with the weather data with respect to time;

determining, by interpolation, intermediate data between any data that is non-uniformly spaced with respect to time in the synchronized thermostat and weather data;

inserting the intermediate data into the synchronized thermostat and weather data to generate synchronized thermostat and weather data that is uniformly spaced;

training at least one machine learning model using the synchronized thermostat and weather data;

monitoring performance of the HVAC system over time using the trained machine learning model by repeatedly:

receiving weather data from the weather service for a future time period after the initial time period;

receiving additional thermostat data for the future time period from the thermostat, the additional thermostat data including temperature setpoint data, measured building temperature data, and HVAC operation data for the future time period;

determining, based on the received additional thermostat data, an actual amount of heating or cooling and an actual amount of fan usage for the future time period;

determining, using the trained machine learning model and based on the received weather data for the future time period, an expected amount of heating or cooling and an expected amount of fan usage for the future time period with the HVAC system set at the determined temperature setpoint;

comparing the actual amount of heating or cooling to the expected amount of heating or cooling;

comparing the actual amount of fan usage to the expected amount fan usage; and

determining that the performance of the HVAC system has decreased when the actual amount of heating or cooling differs from the expected amount of heating or cooling by more than a first threshold amount or the actual amount of fan usage differs from the expected amount of fan usage by more than a second threshold amount; and

outputting an alert when the monitored performance of the HVAC system is determined to have decreased.

2. The method of claim 1 , wherein receiving weather data from the weather service for the future time period, receiving additional thermostat data from the thermostat during the future time period, using the trained machine learning model to determine the predicted HVAC operation during the future time period based on the weather data for the future time period, and comparing the additional thermostat data for the future time period to the predicted HVAC operation for the future time period are performed once per day.

3. The method of claim 1 , wherein training at least one machine learning model using the synchronized thermostat and weather data comprises training a regression based model using the synchronized thermostat and weather data.

4. The method of claim 1 , wherein training at least one machine learning model using the synchronized thermostat and weather data comprises training a long short-term memory deep learning neural network dynamic model using the synchronized thermostat and weather data.

5. The method of claim 1 , wherein outputting an alert when the monitored performance of the HVAC system is determined to have decreased comprises outputting an alert to a service provider associated with the building.

6. The method of claim 1 , further comprising servicing the HVAC system in response to the output alert.

7. The method of claim 1 , wherein training at least one machine learning model using the synchronized thermostat and weather data comprises training an encoder-decoder long short-term memory deep learning neural network dynamic model using the synchronized thermostat and weather data.

8. The method of claim 1 , wherein training at least one machine learning model using the synchronized thermostat and weather data comprises training a distributed random forest model using the synchronized thermostat and weather data.

9. The method of claim 1 , wherein training at least one machine learning model using the synchronized thermostat and weather data comprises training a global boosting model using the synchronized thermostat and weather data.

10. A performance monitoring system comprising:

a communication interface, the communication interface operable to communicatively couple the performance monitoring system to at least one thermostat of an HVAC system in a building including a fan,

a memory; and

a processor coupled to the communication interface and the memory, the memory storing instructions that when executed by the processor cause the processor to:

receive thermostat data from the thermostat through the communication interface, the thermostat data including temperature setpoint data, measured building temperature data, and HVAC operation data for an initial time period;

receive weather data from a weather service for the initial time period;

synchronize the thermostat data with the weather data with respect to time;

determine, by interpolation, intermediate data between any data that is non-uniformly spaced with respect to time in the synchronized thermostat and weather data;

insert the intermediate data into the synchronized thermostat and weather data to generate synchronized thermostat and weather data that is uniformly spaced;

train at least one machine learning model using the synchronized thermostat and weather data;

monitor performance of the HVAC system over time using the trained machine learning model by repeatedly:

receiving weather data from the weather service for a future time period after the initial time period;

receiving additional thermostat data for the future time period from the thermostat, the additional thermostat data including temperature setpoint data, measured building temperature data, and HVAC operation data for the future time period;

determining, based on the received additional thermostat data, an actual amount of heating or cooling and an actual amount of fan usage for the future time period;

determining, using the trained machine learning model and based on the received weather data for the future time period, an expected amount of heating or cooling and an expected amount of fan usage for the future time period with the HVAC system set at the determined temperature setpoint;

comparing the actual amount of heating or cooling to the expected amount of heating or cooling;

comparing the actual amount of fan usage to the expected amount of fan usage; and

determining that the performance of the HVAC system has decreased when the actual amount of heating or cooling differs from the expected amount of heating or cooling by more than a first threshold amount or the actual amount of fan usage differs from the expected amount of fan usage by more than a second threshold amount; and

output an alert when the monitored performance of the HVAC system is determined to have decreased.

11. The performance monitoring system of claim 10 , wherein the instructions cause the processor to receive weather data from the weather service for the future time period, receive additional thermostat data from the thermostat during the future time period, use the trained machine learning model to determine the predicted HVAC operation during the future time period based on the weather data for the future time period, and compare the additional thermostat data for the future time period to the predicted HVAC operation for the future time period once per day.

12. The performance monitoring system of claim 10 , wherein the instructions cause the processor to train at least one machine learning model using the synchronized thermostat and weather data by training a regression based model using the synchronized thermostat and weather data.

13. The performance monitoring system of claim 10 , wherein the instructions cause the processor to train at least one machine learning model using the synchronized thermostat and weather data by training a long short-term memory deep learning neural network dynamic model using the synchronized thermostat and weather data.

14. The performance monitoring system of claim 10 , wherein the instructions cause the processor to output the alert when the monitored performance of the HVAC system is determined to have decreased by outputting an alert to a service provider associated with the building.

15. The performance monitoring system of claim 10 , wherein the instructions cause the processor to train at least one machine learning model using the synchronized thermostat and weather data by training an encoder-decoder long short-term memory deep learning neural network dynamic model using the synchronized thermostat and weather data.

16. The performance monitoring system of claim 10 , wherein the instructions cause the processor to train at least one machine learning model using the synchronized thermostat and weather data by training a distributed random forest model using the synchronized thermostat and weather data.

17. The performance monitoring system of claim 10 , wherein the instructions cause the processor to train at least one machine learning model using the synchronized thermostat and weather data by training a global boosting model using the synchronized thermostat and weather data.