Determining HVAC System Type Using Measured Data and Machine Learning

Abstract

Systems and methods are described for identifying and classifying airflow system types. The system may include: (1) an airflow system receiving air via a return pathway and supplying air via a supply pathway; (2) a comfort sensor to collect supply airflow data; (3) a filter sensor to collect return airflow data; and (4) one or more processors coupled to a memory storing one or more instructions that cause the processors to: (a) receive the supply airflow data; (b) receive the return airflow data; (c) analyze the supply and return airflow data to generate one or more cycle variables; and (d) analyze the cycle variables using a machine learning model, wherein the analyzing includes: (i) generating, using the machine learning model, a class prediction metric based on the one or more cycle variables, and (ii) classifying, based on the class prediction metric, an airflow system type for the airflow system.

What is claimed: 
     
         1 . A system for identifying and classifying airflow system types, the system comprising:
 an airflow system configured to receive air via a return air pathway and supply air via a supply air pathway;   a comfort sensor disposed proximate to the supply air pathway of the airflow system and configured to collect supply airflow data associated with the supply air pathway;   a filter sensor disposed proximate to the return air pathway of the airflow system and configured to collect return airflow data associated with the return air pathway; and   one or more processors communicatively coupled to a memory storing one or more instructions that, when executed, cause the one or more processors to:
 receive the supply airflow data from the comfort sensor; 
 receive the return airflow data from the filter sensor; 
 analyze the supply airflow data and the return airflow data to generate one or more cycle variables associated with the airflow system; and 
 analyze the one or more cycle variables associated with the airflow system using a machine learning model, wherein the analyzing includes:
 generating, using the machine learning model, a class prediction metric based on the one or more cycle variables associated with the airflow system, and 
 classifying, based on the class prediction metric, an airflow system type for the airflow system. 
 
   
     
     
         2 . The system of  claim 1 , wherein the supply airflow data is representative of a temperature of air flowing from the airflow system via the supply air pathway. 
     
     
         3 . The system of  claim 1 , wherein the return airflow data is representative of at least one of: (i) a temperature or (ii) a pressure of air flowing to the airflow system via the return air pathway. 
     
     
         4 . The system of  claim 1 , wherein the one or more cycle variables include at least one of: (i) an air conditioning cycle type, (ii) a heating cycle type, (iii) a cycle run time, or (iv) an airflow system heating performance. 
     
     
         5 . The system of  claim 1 , wherein the airflow system type is associated with at least an airflow system functionality, and the airflow system functionality includes at least one of (i) air conditioner functionality or (ii) heat pump functionality. 
     
     
         6 . The system of  claim 1 , wherein the airflow system type is associated with at least an airflow system speed, and the airflow system speed includes at least one of (i) dual speed, (ii) constant speed, or (iii) variable speed. 
     
     
         7 . The system of  claim 1 , wherein generating the class prediction metric includes:
 receiving, at the machine learning model, the one or more cycle variables as one or more inputs;   calculating, by one or more trees of the machine learning model, a mean probability estimate for each airflow system type class; and   determining the class prediction metric based on a highest mean probability estimate.   
     
     
         8 . The system of  claim 1 , wherein the memory further stores instructions that, when executed, cause the one or more processors to:
 train the machine learning model using stored airflow data associated with one or more airflow system types, wherein the training includes:
 receiving, at the machine learning model, the stored airflow data as one or more inputs; 
 calculating, by one or more trees of the machine learning model, a mean probability estimate for each airflow system type class; 
 determining the class prediction metric based on a highest mean probability estimate for the stored airflow data; and 
 validating the class prediction metric. 
   
     
     
         9 . The system of  claim 8 , wherein the memory further stores instructions that, when executed, cause the one or more processors to:
 preprocess the stored airflow data prior to training the machine learning model, wherein the preprocessing includes:
 identifying a subset of stored airflow data, wherein the subset includes one or more predetermined cycle variables associated with the stored airflow data; and 
 resample the subset of stored airflow data. 
   
     
     
         10 . The system of  claim 1 , wherein the memory further stores instructions that, when executed, cause the one or more processors to:
 receive ambient airflow data from a water sensor disposed proximate to the airflow system;   wherein analyzing the supply airflow data and the return airflow data to generate the one or more cycle variables associated with the airflow system includes:
 analyzing the supply airflow data, the return airflow data, and the ambient airflow data to generate the one or more cycle variables. 
   
     
     
         11 . A method for identifying and classifying airflow system types, the method comprising:
 receiving, by one or more processors, supply airflow data from a comfort sensor disposed proximate to a supply air pathway of an airflow system;   receiving, by the one or more processors, return airflow data from a filter sensor disposed proximate to a return air pathway of the airflow system;   analyzing, by the one or more processors, the supply airflow data and the return airflow data to generate one or more cycle variables associated with the airflow system; and   analyzing, by the one or more processors, the one or more cycle variables associated with the airflow system using a machine learning model, wherein the analyzing includes:
 generating, using the machine learning model, a class prediction metric based on the one or more cycle variables associated with the airflow system, and 
 classifying, based on the class prediction metric, an airflow system type for the airflow system. 
   
     
     
         12 . The method of  claim 11 , wherein the supply airflow data is representative of a temperature of air flowing from the airflow system via the supply air pathway. 
     
     
         13 . The method of  claim 11 , wherein the return airflow data is representative of at least one of: (i) a temperature or (ii) a pressure of air flowing to the airflow system via the return air pathway. 
     
     
         14 . The method of  claim 11 , wherein the one or more cycle variables include at least one of: (i) an air conditioning cycle type, (ii) a heating cycle type, (iii) a cycle run time, or (iv) an airflow system heating performance. 
     
     
         15 . The method of  claim 11 , wherein the airflow system type is associated with at least an airflow system functionality, and the airflow system functionality includes at least one of (i) air conditioner functionality or (ii) heat pump functionality. 
     
     
         16 . The method of  claim 11 , wherein the airflow system type is associated with at least an airflow system speed, and the airflow system speed includes at least one of (i) dual speed, (ii) constant speed, or (iii) variable speed. 
     
     
         17 . The method of  claim 11 , wherein generating the class prediction metric includes:
 receiving, at the machine learning model, the one or more cycle variables as one or more inputs;   calculating, by one or more trees of the machine learning model, a mean probability estimate for each airflow system type class; and   determining the class prediction metric based on a highest mean probability estimate.   
     
     
         18 . The method of  claim 11 , further comprising:
 training the machine learning model using stored airflow data associated with one or more airflow system types, wherein the training includes:
 receiving, at the machine learning model, stored airflow data as one or more inputs; 
 calculating, by one or more trees of the machine learning model, a mean probability estimate for each airflow system type class; 
 determining the class prediction metric based on a highest mean probability estimate for the stored airflow data; and 
 validating the class prediction metric. 
   
     
     
         19 . The method of  claim 18 , further comprising:
 preprocessing the stored airflow data prior to training the machine learning model, wherein the preprocessing includes:
 identifying a subset of stored airflow data, wherein the subset includes one or more predetermined cycle variables associated with the stored airflow data; and 
 resample the subset of stored airflow data. 
   
     
     
         20 . The method of  claim 11 , further comprising:
 receiving ambient airflow data from a water sensor disposed proximate to the airflow system;   wherein the analyzing the supply airflow data and the return airflow data to generate the one or more cycle variables associated with the airflow system includes:
 analyzing the supply airflow data, the return airflow data, and the ambient airflow data to generate the one or more cycle variables.