KR-102962660-B1 - A device for generating a temperature prediction model and a method for providing a simulation environment

KR102962660B1KR 102962660 B1KR102962660 B1KR 102962660B1KR-102962660-B1

Abstract

An apparatus for generating a temperature prediction model is disclosed. An apparatus for generating a temperature prediction model according to an embodiment of the present invention comprises: a temperature prediction model that provides a simulation environment; and a processor that sets hyperparameters of the temperature prediction model, trains the temperature prediction model with the set hyperparameters to output a predicted temperature, updates the hyperparameters based on the difference between the predicted temperature output by the trained temperature prediction model and the actual temperature, and sets the final hyperparameters of the temperature prediction model by repeating the setting of the hyperparameters, the training of the temperature prediction model, and the updating of the hyperparameters based on the difference between the predicted temperature and the actual temperature a predetermined number of times or more.

Inventors

고한규
김봉상

Assignees

엘지전자 주식회사

Dates

Publication Date: 20260508
Application Date: 20190729
Priority Date: 20190604

Claims (14)

A temperature prediction model that provides a simulation environment; and The hyperparameters of the above temperature prediction model are set, the temperature prediction model with the set hyperparameters is trained to output a predicted temperature, and the hyperparameters are updated based on the difference between the predicted temperature output by the trained temperature prediction model and the actual temperature. A processor for setting the final hyperparameters of the temperature prediction model by repeating the setting of the hyperparameters, the training of the temperature prediction model, and the updating of the hyperparameters based on the difference between the predicted temperature and the actual temperature a predetermined number of times or more; The above processor is, During the initial time, temperature control operation information of the control system and the actual temperature corresponding to the temperature control operation information of the control system are input into the trained temperature prediction model to output a predicted temperature, and After the above initial time has elapsed, the temperature control operation information of the control system and the previous predicted temperature are input into the trained temperature prediction model to output the predicted temperature, and The hyperparameters are updated based on the difference between the predicted temperature and the actual temperature during a predetermined period after the above initial time has elapsed, and The above processor is, When setting the final hyperparameters of the above temperature prediction model, hyperparameters having multiple elements including the number of layers, the number of nodes per layer, the number of training iterations, the training rate, and the drop rate are obtained, and Among the hyperparameters obtained above, some elements having fixed values set by the user are excluded, and the remaining elements are selected. Check the preset search range for each of the remaining selected elements above, and Updating the remaining selected elements within the above preset search range, Setting the final hyperparameters including some elements having fixed values set by the user and the other updated elements. A device for generating a temperature prediction model.
In Article 1, The above temperature prediction model is, A recurrent neural network trained using time series data including temperature control operation information of the control system and a corresponding temperature to output the above predicted temperature. A device for generating a temperature prediction model.
◈Claim 3 was waived upon payment of the establishment registration fee.◈ In Paragraph 2, The above processor is, The above time series data is provided to a temperature prediction model with the above hyperparameters set to train the temperature prediction model with the above hyperparameters set so as to output the predicted temperature, and Information on the temperature control operation of the control system for a predetermined period is input into the above-mentioned trained temperature prediction model, and Updating the hyperparameter based on the difference between the actual temperature corresponding to the temperature control operation information of the control system during the above-mentioned predetermined period and the predicted temperature output based on the temperature control operation information of the control system during the above-mentioned predetermined period. A device for generating a temperature prediction model.
In Article 1, The above processor is, Within the search range of the hyperparameters, the hyperparameter that minimizes the difference between the predicted temperature output by the trained temperature prediction model and the actual temperature is set as the final hyperparameter. A device for generating a temperature prediction model.
In Article 1, The above processor is, Set the hyperparameter as the final hyperparameter to ensure that the difference between the predicted temperature output by the above-trained temperature prediction model and the actual temperature becomes smaller than a preset value. A device for generating a temperature prediction model.
In Article 1, The above processor is, Updating the above hyperparameters based on any one of the algorithms Bayesian Optimization, Reinforcement Learning, and Bayesian Optimization & HyperBand A device for generating a temperature prediction model.
delete
delete
In the method of providing a simulation environment, A step of setting hyperparameters of a temperature prediction model, training the temperature prediction model with the set hyperparameters to output a predicted temperature, and updating the hyperparameters based on the difference between the predicted temperature output by the trained temperature prediction model and the actual temperature; and The method includes the step of setting the hyperparameters, training the temperature prediction model, and updating the hyperparameters based on the difference between the predicted temperature and the actual temperature, repeating this process more than a predetermined number of times to set the final hyperparameters of the temperature prediction model. The step of updating the above hyperparameters is, During the initial time, temperature control operation information of the control system and the actual temperature corresponding to the temperature control operation information of the control system are input into the trained temperature prediction model to output a predicted temperature, and After the above initial time has elapsed, the temperature control operation information of the control system and the previous predicted temperature are input into the trained temperature prediction model to output the predicted temperature, and The hyperparameters are updated based on the difference between the predicted temperature and the actual temperature during a predetermined period after the above initial time has elapsed, and The step of setting the final hyperparameters of the above temperature prediction model is, Obtain hyperparameters having multiple elements including the number of layers, the number of nodes per layer, the number of training iterations, the training rate, and the drop rate, and Among the hyperparameters obtained above, some elements having fixed values set by the user are excluded, and the remaining elements are selected. Check the preset search range for each of the remaining selected elements above, and Updating the remaining selected elements within the above preset search range, Setting the final hyperparameters including some elements having fixed values set by the user and the other updated elements. Method of providing a simulation environment.
In Article 9, The above temperature prediction model is, A recurrent neural network trained using time series data including temperature control operation information of the control system and a corresponding temperature to output the above predicted temperature. Method of providing a simulation environment.
◈Claim 11 was waived upon payment of the establishment registration fee.◈ In Article 10, The step of updating the above hyperparameters is, A step of providing the above time series data to a temperature prediction model with the above hyperparameters set, and training the temperature prediction model with the above hyperparameters set so that it outputs the predicted temperature; A step of inputting temperature control operation information of the control system for a predetermined period into the above-mentioned trained temperature prediction model; and A step of updating the hyperparameter based on the difference between the actual temperature corresponding to the temperature control operation information of the control system during the aforementioned predetermined period and the predicted temperature output based on the temperature control operation information of the control system during the aforementioned predetermined period. Method of providing a simulation environment.
In Article 9, The step of setting the final hyperparameters of the above temperature prediction model is, Within the search range of the hyperparameters, the hyperparameter that minimizes the difference between the predicted temperature output by the trained temperature prediction model and the actual temperature is set as the final hyperparameter. Method of providing a simulation environment.
In Article 9, The step of setting the final hyperparameters of the above temperature prediction model is, Set the hyperparameter as the final hyperparameter to ensure that the difference between the predicted temperature output by the above-trained temperature prediction model and the actual temperature becomes smaller than a preset value. Method of providing a simulation environment.
In Article 9, A step of obtaining a predicted temperature by inputting the temperature control operation information of the above-described control system into a temperature prediction model in which the final hyperparameters are set; and The artificial intelligence device further comprises the step of updating parameters for reaching a target temperature based on a predicted temperature corresponding to the temperature control operation information of the control system, based on reinforcement learning. Method of providing a simulation environment.

Description

A device for generating a temperature prediction model and a method for providing a simulation environment The present invention relates to an apparatus for generating a temperature prediction model capable of generating a temperature prediction model by setting optimal hyperparameters, and a method for providing a simulation environment. Artificial intelligence is a field of computer science and information technology that studies methods to enable computers to perform thinking, learning, and self-development capable of human intelligence, and refers to making computers mimic intelligent human behavior. Furthermore, artificial intelligence does not exist in isolation but is closely related, directly and indirectly, to many other fields of computer science. Particularly in the modern era, there are very active attempts to introduce AI elements into various sectors of information technology and utilize them to solve problems within those fields. The temperature prediction model is a temperature prediction simulator that predicts how the temperature will change when control values are set, and can provide a simulation environment for various devices that require temperature prediction. However, since temperature is determined by a wide variety of variables such as the performance of the air conditioner, valve performance, building information (building structure, materials, number of windows, wall thickness, etc.), season, date, and time, it is not easy to create a simulator that reflects such diverse variables. Existing indoor temperature simulators are implemented by updating formulas capable of predicting building temperature changes by reflecting building information, ranging from air conditioner specifications to building materials and the number of windows. The method of expressing the relationship between variables and temperature using a formula and updating the formula in this manner had a problem in that, due to its complexity, accurate updates became impossible as the number of variables increased, leading to a decrease in prediction accuracy. In addition, since the method involves humans checking the difference between actual temperature data and simulation temperature data and updating the formula based on human intuition, there was a problem in deriving the optimal formula. FIG. 1 is a block diagram illustrating an artificial intelligence device according to an embodiment of the present invention. FIG. 2 is a drawing for explaining a method of setting a baseline according to an embodiment of the present invention. FIG. 3 is a diagram illustrating a method for performing reinforcement learning with the goal of tracking a baseline by a second line and an artificial intelligence unit, according to an embodiment of the present invention. FIG. 4 is a diagram illustrating a method for providing different compensation depending on the position of the gap according to an embodiment of the present invention. FIG. 5 is a diagram illustrating the comparison range between a baseline and an output value according to an embodiment of the present invention. FIG. 6 is a diagram illustrating a method for performing reinforcement learning with the goal of setting an additional baseline and avoiding the additional baseline, according to an embodiment of the present invention. FIG. 7 is a diagram illustrating a method for discarding parameters when an output value and a point on a second baseline coincide, according to an embodiment of the present invention. FIG. 8 is a drawing illustrating a method for resetting a baseline according to changes in environmental conditions, according to an embodiment of the present invention. FIG. 9 is a flowchart for explaining the operation method of an artificial intelligence device and a control system according to an embodiment of the present invention. FIG. 10 is a diagram illustrating a method for pre-learning a pattern of output values according to an embodiment of the present invention. FIG. 11 is a flowchart illustrating a method for obtaining a pattern of output values using a recurrent neural network and a method for performing reinforcement learning based on the pattern of output values, according to an embodiment of the present invention. FIG. 12 is a drawing illustrating an artificial intelligence device in which a control system, a collection unit, and an artificial intelligence unit are integrally configured according to an embodiment of the present invention. FIG. 13 is a block diagram illustrating an embodiment according to the present invention in which a control system and an artificial intelligence device are configured separately, and an output value is collected from the artificial intelligence device. FIG. 14 is a block diagram illustrating an embodiment in which an artificial intelligence device corresponding to each of a plurality of control systems is integrally configured in a control center according to an embodiment of the present invention. FIG. 15 is