Energy program communication control system and method based on load shape analysis

A computing system, comprising:

• at least one processor connected to at least one memory;
• a data management module stored on a non-transitory computer readable medium and including instructions that when executed by the at least one processor cause the computing system to
• generate a plurality of established load shapes, wherein each established load shape represents an energy consumption pattern of a dwelling type over a time period based on observed energy consumption data;
• an analysis module stored on the non-transitory computer readable medium and including instructions that when executed by the at least one processor cause the computing system to:
• input a training set of data including data records from a plurality of energy customers, wherein a data record for a given energy customer includes at least (i) a load shape based on power consumed by the given energy customer over the time period, and (ii) participation data that indicates that the given energy customer is a known participant in an energy program or a known non-participant;
• separate, by the processor, the data records into subset records that correspond to one of the plurality of established load shapes based on the load shape from a given data record that matches one of the plurality of established load shapes;
• generate a first data structure for each given energy customer of the plurality of energy customers comprising a combination of the matching established load shape, a binary label of the participation data indicating whether or not the given energy customer is a participant in an energy efficiency program, demographic data associated with the given energy customer, and site parcel data associated with the given energy customer, wherein the site parcel data comprises at least one of:
• square footage, number of floors, dwelling type, or heat type;
• train a predictive model by machine learning to classify each of the plurality of energy customers in the training set of data into (i) a class of customers who are likely to participate in an energy efficiency program, or (ii) a class of customers who are unlikely to participate in the energy efficiency program;
• wherein the classifying is based, at least in part, on each of the first data structures associated with the plurality of energy customers in the training set of data and the binary label associated with a given energy customer;
• in response to a request to predict a propensity to participate in the energy program for a target group of energy customers who are known to be non-participants in the energy efficiency program:
• obtain, from the database, load data for each customer of the target group of energy customers who are known to be non-participants, wherein the load data for a given target energy customer is indicative of power consumed by the given target energy customer over the specified period of time;
• generate an empirical load shape for each given target energy customer from the target group based on the obtained load data for the given target energy customer, the empirical load shape representing the power consumed by the given target energy customer over the specified period of time;
• compare, by the processor, the empirical load shape for each given target energy customer to the plurality of established load shapes and select a matching established load shape;
• generate, by the processor, a data record for each given target energy customer, comprising a combination of the matching established load shape, and parameters associated with the given target energy customer including one or more of (i) the load data, (ii) demographic data, and (iii) site parcel data comprising one or more of:
• square footage, dwelling type, or heat type;
• apply the predictive model to the data records of the target group of energy customers to identify a final set of non-participants who are likely to participate in the energy efficiency program, comprising:
• input the data records of each given target energy customer into the predictive model;
• for a given data record, compare similarities between the parameters in the given data record to the training set of data based on the matching established load shape and the subset records that correspond to the matching established load shape which include the participation data;
• generate a predicted propensity to participate in the energy program for each given target energy customer based on the comparison with the subset of records from the training set of data;
• and
• filter the data records of the target group of energy customers according to the predicted propensity, wherein the data records that are likely to participate are included in the final set of non-participants who are likely to participate and data records not likely to participate are not included;
• upon determining the final set of non-participants who are likely to participate in the energy efficiency program, generate an electronic message having information about the energy efficiency program;
• and
• a transmission control module stored on the non-transitory computer readable medium and including instructions that when executed by the at least one processor cause the computing system to control, by the processor, a transmission of the electronic message based on the determined final set of non-participants, where the electronic message is transmitted to an electronic contact information associated with the target energy customers within the final set of non-participants.

A computer-implemented method, the method comprising:

• generating a plurality of established load shapes, wherein each established load shape represents an energy consumption pattern of a dwelling type over a time period based on observed energy consumption data;
• inputting a training set of data including data records from a plurality of energy customers, wherein a data record for a given energy customer includes at least (i) a load shape based on power consumed by the given energy customer over the time period, and (ii) participation data that indicates that the given energy customer is a known participant in an energy program or a known non-participant;
• separating, by the processor, the data records into subset records that correspond to one of the plurality of established load shapes based on the load shape from a given data record that matches one of the plurality of established load shapes;
• generating a first data structure for each given energy customer of the plurality of energy customers comprising a combination of the matching established load shape, a binary label of the participation data indicating whether or not the given energy customer is a participant in an energy efficiency program, demographic data associated with the given energy customer, and site parcel data associated with the given energy customer, wherein the site parcel data comprises at least one of:
• square footage, number of floors, dwelling type, or heat type;
• training a predictive model by machine learning to classify each of the plurality of energy customers in the training set of data into (i) a class of customers who are likely to participate in an energy efficiency program, or (ii) a class of customers who are unlikely to participate in the energy efficiency program;
• wherein the classifying is based, at least in part, on each of the first data structures associated with the plurality of energy customers in the training set of data and the binary label associated with a given energy customer;
• in response to a request to predict a propensity to participate in the energy program for a target group of energy customers who are known to be non-participants in the energy efficiency program:
• obtaining, from the database, load data for each customer of the target group of energy customers who are known to be non-participants, wherein the load data for a given target energy customer is indicative of power consumed by the given target energy customer over the specified period of time;
• generating an empirical load shape for each given target energy customer from the target group based on the obtained load data for the given target energy customer, the empirical load shape representing the power consumed by the given target energy customer over the specified period of time;
• comparing, by the processor, the empirical load shape for each given target energy customer to the plurality of established load shapes and select a matching established load shape;
• generating, by the processor, a data record for each given target energy customer, comprising a combination of the matching established load shape, and parameters associated with the given target energy customer including one or more of (i) the load data, (ii) demographic data, and (iii) site parcel data comprising one or more of:
• square footage, dwelling type, or heat type;
• applying the predictive model to the data records of the target group of energy customers to identify a final set of non-participants who are likely to participate in the energy efficiency program, comprising:
• inputting the data records of each given target energy customer into the predictive model;
• for a given data record, comparing similarities between the parameters in the given data record to the training set of data based on the matching established load shape and the subset records that correspond to the matching established load shape which include the participation data;
• generating a predicted propensity to participate in the energy program for each given target energy customer based on the comparison with the subset of records from the training set of data;
• and
• filtering the data records of the target group of energy customers according to the predicted propensity, wherein the data records that are likely to participate are included in the final set of non-participants who are likely to participate and data records not likely to participate are not included;
• upon determining the final set of non-participants who are likely to participate in the energy efficiency program, generating an electronic message having information about the energy efficiency program;
• and
• controlling, by the processor, a transmission of the electronic message based on the determined final set of non-participants, where the electronic message is transmitted to an electronic contact information associated with the target energy customers within the final set of non-participants.