Air conditioning control device, air conditioning control method, and computer readable medium

Abstract

An air conditioning control device ( 200 ) calculates one or more estimate values for an air conditioning zone ( 109 ) based on zone environment data and operation state data, generates one or more candidates for a setting value group pair based on current setting data, calculates predicted energy consumption and predicted comfort as a predicted balance based on the one or more estimate values for each candidate, and determines a setting value group pair to be set for an air conditioner pair ( 101 ) based on the predicted balance of each candidate.

1 . An air conditioning control device comprising
 processing circuitry to:   calculate an estimate value based on zone environment data that indicates a measurement value of each of one or more types of states of an air conditioning zone in which an air conditioner pair composed of two types of air conditioners is used and operation state data that indicates a state value of each of one or more types of states for each of the air conditioners, the estimate value being an estimate value of each of one or more types of states of the air conditioning zone different from the one or more types of states in the zone environment data;   generate one or more candidates for a setting value group pair for the air conditioner pair based on current setting data that indicates a setting value group pair that is set for the air conditioner pair;   calculate, as a predicted balance, predicted power consumption and predicted comfort in a case where the setting value group pair is set for the air conditioner pair, based on one or more estimate values that have been calculated, for each generated candidate;   determine one candidate as a setting value group pair to be set for the air conditioner pair based on the predicted balance of each candidate; and   generate training environment data equivalent to the zone environment data and training state data equivalent to the operation state data, calculate training estimate data equivalent to the one or more estimate values based on the training environment data and the training state data, generate training setting data equivalent to the current setting data, generate one or more training candidates equivalent to the one or more candidates based on the training setting data, calculate a training balance equivalent to the predicted balance in a case where the training candidate is set for the air conditioner pair, based on the training estimate data, for each training candidate, and generate an alternative model for each training candidate by learning the training candidate, the training estimate data, and the training balance,   wherein the processing circuitry calculates the predicted balance for each candidate by calculating the alternative model using as input the one or more estimate values and the candidate.   
     
     
         2 . An air conditioning control method comprising:
 calculating an estimate value based on zone environment data that indicates a measurement value of each of one or more types of states of an air conditioning zone in which an air conditioner pair composed of two types of air conditioners is used and operation state data that indicates a state value of each of one or more types of states for each of the air conditioners, the estimate value being an estimate value of each of one or more types of states of the air conditioning zone different from the one or more types of states in the zone environment data;   generating one or more candidates for a setting value group pair for the air conditioner pair based on current setting data that indicates a setting value group pair that is set for the air conditioner pair;   calculating, as a predicted balance, predicted power consumption and predicted comfort in a case where the setting value group pair is set for the air conditioner pair for each generated candidate by calculating an alternative model using as input one or more estimate values that have been calculated and the candidate; and   determining one candidate as a setting value group pair to be set for the air conditioner pair based on the predicted balance of each candidate,   wherein training environment data equivalent to the zone environment data and training state data equivalent to the operation state data are generated, training estimate data equivalent to the one or more estimate values is calculated based on the training environment data and the training state data, training setting data equivalent to the current setting data is generated, one or more training candidates equivalent to the one or more candidates are generated based on the training setting data, a training balance equivalent to the predicted balance in a case where the training candidate is set for the air conditioner pair is calculated based on the training estimate data, for each training candidate, and the alternative model is generated for each training candidate by learning the training candidate, the training estimate data, and the training balance.   
     
     
         3 . A non-transitory computer readable medium storing an air conditioning control program to cause a computer to execute:
 a state estimation process of calculating estimate value based on zone environment data that indicates a measurement value of each of one or more types of states of an air conditioning zone in which an air conditioner pair composed of two types of air conditioners is used and operation state data that indicates a state value of each of one or more types of states for each of the air conditioners, the estimate value being an estimate value of each of one or more types of states of the air conditioning zone different from the one or more types of states in the zone environment data;   a candidate generation process of generating one or more candidates for a setting value group pair for the air conditioner pair based on current setting data that indicates a setting value group pair that is set for the air conditioner pair;   a balance prediction process of calculating, as a predicted balance, predicted power consumption and predicted comfort in a case where the setting value group pair is set for the air conditioner pair for each generated candidate by calculating an alternative model using as input one or more estimate values that have been calculated and the candidate;   a setting determination process of determining one candidate as a setting value group pair to be set for the air conditioner pair based on the predicted balance of each candidate; and   a model learning process of generating training environment data equivalent to the zone environment data and training state data equivalent to the operation state data, calculating training estimate data equivalent to the one or more estimate values based on the training environment data and the training state data, generating training setting data equivalent to the current setting data, generating one or more training candidates equivalent to the one or more candidates based on the training setting data, calculating a training balance equivalent to the predicted balance in a case where the training candidate is set for the air conditioner pair, based on the training estimate data, for each training candidate, and generating the alternative model for each training candidate by learning the training candidate, the training estimate data, and the training balance.