Learning device and inference device for maintenance of air conditioner
A model generation unit converts each of a first model, a second model, and a third model into a trained model. First training data includes a first parameter representing the degree of clogging of an air filter, a second parameter pertaining to the air-conditioning power of an air-conditioning system, and a third parameter representing an increased amount of electric power cost of the air-conditioning system due to the first parameter during operation of the second parameter. Second training data includes a fourth parameter representing a first date and time and a fifth parameter pertaining to air-conditioning power of the air-conditioning system assumed on the first date and time. Third training data includes a sixth parameter representing a second date and time and a seventh parameter representing a maintenance cost of the air filter on the second date and time.
1 . A learning device that learns maintenance of an air-conditioning system including at least one air filter, the learning device comprising:
a circuitry configured to:
convert each of a first model, a second model, and a third model into a trained model by machine learning using a first training data, a second training data, and a third training data, wherein
the first training data includes a first parameter representing a degree of clogging of the at least one air filter, a second parameter pertaining to air-conditioning power of the air-conditioning system, and a third parameter representing an increased amount of electric power cost of the air-conditioning system due to the first parameter during operation of the second parameter,
the second training data includes a fourth parameter representing a first date and time and a fifth parameter pertaining to air-conditioning power of the air-conditioning system assumed on the first date and time,
the third training data includes a sixth parameter representing a second date and time and a seventh parameter representing a maintenance cost of the at least one air filter on the second date and time,
the first model estimates the third parameter from the first parameter and the second parameter,
the second model estimates the fifth parameter from the fourth parameter, and
the third model estimates the seventh parameter from the sixth parameter; and
instruct timing for maintenance on the at least one air filter based on the estimated third parameter from the first model which has been trained, the estimated fifth parameter from the second model which has been trained, and the estimated seventh parameter from the third model which has been trained.
2 . The learning device according to claim 1 , wherein
each of the third parameter, the fifth parameter, and the seventh parameter is ground truth data, and
the circuitry is configured to perform supervised learning for each of the first model, the second model, and the third model.
3 . The learning device according to claim 1 , wherein
the circuitry is configured to generate each of the first model, the second model, and the third model by supervised learning.
4 . An inference device for an air filter maintenance system that maintains an air-conditioning system including at least one air filter, the inference device comprising:
a circuitry configured to
acquire, as time passes, a first parameter that represents a degree of clogging of the at least one air filter detected by the air-conditioning system, a second parameter that pertains to air-conditioning power of the air-conditioning system indicated by the air-conditioning system, a fourth parameter that represents a first date and time, and a sixth parameter that represents a second date and time, wherein the first parameter, the second parameter, the fourth parameter, and the sixth parameter temporally change;
use a first model that has been trained on the first parameter and the second parameter, a second model that has been trained on the fourth parameter, and a third model that has been trained on the sixth parameter, the first model, the second model, and the third model being generated by a learning device, wherein the circuitry is further configured to
estimate, using the first model that has been trained, a third parameter being ground truth data representing an increased amount of electric power cost of the air-conditioning system based on the acquired first parameter during operation of the acquired second parameter, wherein the first parameter, the second parameter, and the third parameter are associated with each other, and temporally change,
estimate, using the second model that has been trained, a fifth parameter being ground truth data pertaining to air-conditioning power of the air-conditioning system assumed on the first date and time of the acquired fourth parameter, wherein the fourth parameter and the fifth parameter are associated with each other, and temporally change,
estimate, using the third model that has been trained, a seventh parameter representing a maintenance cost of the at least one air filter on the second date and time of the acquired sixth parameter, wherein the sixth parameter and the seventh parameter are associated with each other, and temporally change; and
instruct timing for maintenance on the at least one air filter based on the estimated third parameter from the first model which has been trained, the estimated fifth parameter from the second model which has been trained, and the estimated seventh parameter from the third model which has been trained.
5 . The inference device according to claim 4 , wherein
each of the first model, the second model, and the third model is generated by supervised learning.
6 . The inference device according to claim 4 , wherein
the first model, the second model, and the third model include neural networks.
7 . An inference device that infers maintenance of an air-conditioning system including at least one air filter using a first model, a second model, and a third model that have been trained by machine learning,
the first model estimating a third parameter from a first parameter and a second parameter,
the second model estimating a fifth parameter from a fourth parameter,
the third model estimating a seventh parameter from a sixth parameter,
the first parameter representing a degree of clogging of the at least one air filter,
the second parameter pertaining to air-conditioning power of the air-conditioning system,
the third parameter representing an increased amount of electric power cost of the air-conditioning system due to the first parameter during operation of the second parameter,
the fourth parameter representing a first date and time,
the fifth parameter representing air-conditioning power of the air-conditioning system assumed on the first date and time,
the sixth parameter representing a second date and time,
the seventh parameter representing a maintenance cost of the at least one air filter on the second date and time,
the inference device comprising:
a circuitry configured to:
estimate the third parameter from the first parameter and the second parameter using the first model;
estimate the fifth parameter from the fourth parameter using the second model;
estimate the seventh parameter from the sixth parameter using the third model; and
instruct timing for maintenance on the at least one air filter based on the estimated third parameter from the first model which has been trained, the estimated fifth parameter from the second model which has been trained, and the estimated seventh parameter from the third model which has been trained.
8 . The inference device according to claim 7 , wherein the machine learning includes supervised learning.
9 . The learning device according to claim 1 , wherein
the first model, the second model, and the third model include neural networks.
10 . The learning device according to claim 1 , wherein
the circuitry is configured to
acquire, as time passes, the first training data, wherein the first training data includes the first parameter representing the degree of clogging of the at least one air filter which is detected by the air-conditioning system and acquired from the air-conditioning system, the second parameter pertaining to air-conditioning power of the air-conditioning system indicated by the air-conditioning system and acquired from the air-conditioning system, and the third parameter being ground truth data representing the increased amount of electric power cost of the air-conditioning system estimated based on the first parameter during operation of the second parameter, wherein the first parameter, the second parameter, and the third parameter are associated with each other, and temporally change,
acquire, as time passes, the second training data, wherein the second training data includes the fourth parameter representing the first date and time which is input and the fifth parameter being ground truth data pertaining to air-conditioning power of the air-conditioning system assumed on the first date and time, wherein the fourth parameter and the fifth parameter are associated with each other, and temporally change, and
acquire, as time passes, the third training data, wherein the third training data includes the sixth parameter representing the second date and time which is input and the seventh parameter representing the maintenance cost of the at least one air filter on the second date and time, wherein the sixth parameter and the seventh parameter are associated with each other, and temporally change.
11 . The learning device according to claim 1 , further comprising:
a memory configured to store the first model, the second model, and the third model.
12 . The inference device according to claim 7 , wherein
the first model, the second model, and the third model include neural networks.
13 . The inference device according to claim 7 , wherein
each of the first model, the second model, and the third model is generated by supervised learning.
14 . The inference device according to claim 7 , wherein
the circuitry is configured to
acquire, as time passes, first training data, wherein the first training data includes the first parameter representing the degree of clogging of the at least one air filter which is detected by the air-conditioning system and acquired from the air-conditioning system, the second parameter pertaining to air-conditioning power of the air-conditioning system indicated by the air-conditioning system and acquired from the air-conditioning system, and the third parameter being ground truth data representing the increased amount of electric power cost of the air-conditioning system estimated based on the first parameter during operation of the second parameter, wherein the first parameter, the second parameter, and the third parameter are associated with each other, and temporally change,
acquire, as time passes, second training data, wherein the second training data includes the fourth parameter representing the first date and time which is input and the fifth parameter being ground truth data pertaining to air-conditioning power of the air-conditioning system assumed on the first date and time, wherein the fourth parameter and the fifth parameter are associated with each other, and temporally change, and
acquire, as time passes, third training data, wherein the third training data includes the sixth parameter representing the second date and time which is input and the seventh parameter representing the maintenance cost of the at least one air filter on the second date and time, wherein the sixth parameter and the seventh parameter are associated with each other, and temporally change.
15 . The inference device according to claim 7 , further comprising:
a memory configured to store the first model, the second model, and the third model.