SYSTEM AND METHOD THEREOF FOR PREDICTING PERFORMANCE OF NEW HEAT PUMP BY USING EXPERIENCE KNOWLEDGE OF EXISTING PRODUCTS AND TRANSFER LEARNING OF SMALL SCALE MEASUREMENT DATA FOR NEW PRODUCTS
The present invention relates to a system for predicting the performance of a new heat pump by using usage experience knowledge of existing products and transfer learning of small-scale measurement data for new products. For constructing a performance prediction system of a new heat pump product by...
Gespeichert in:
Hauptverfasser: | , , , |
---|---|
Format: | Patent |
Sprache: | eng ; kor |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | The present invention relates to a system for predicting the performance of a new heat pump by using usage experience knowledge of existing products and transfer learning of small-scale measurement data for new products. For constructing a performance prediction system of a new heat pump product by utilizing an artificial neural network, a task to be solved is to overcome a shortage problem of the number of learning data by utilizing usage experience knowledge of existing heat pump products as information for assisting learning data of the new heat pump product. As one example of the present invention, the system for predicting performance of a new heat pump comprises: an existing heat pump database construction unit which constructs a database having input data and output data of an artificial neural network for existing heat pump performance prediction based on driving record data collected from an existing heat pump; an existing heat pump performance prediction unit which performs training of the artificial neural network and a prediction process based on the database, generates an artificial neural network prediction sample in the prediction process, and performs an uncertainty analysis process for the new heat pump performance prediction based on the artificial neural network prediction sample to derive an average vector of the artificial neural network prediction sample; a new heat pump data collection unit which utilizes a chamber device in a standard test environment to collect measurement data of the new heat pump; and a transfer learning performing unit which constructs system initial information including the input data of the database and the average vector of the existing heat pump performance prediction unit as output data and performs transfer learning that combines the measurement data of the new heat pump to the initial information.
본 발명은 기존 제품의 사용경험지식과 신규 제품에 대한 소규모 측정데이터의 전이학습을 이용한 신규 히트펌프 성능 예측 시스템에 관한 것으로, 해결하고자 하는 과제는 인공신경망을 활용하여 신규 히트펌프 제품의 성능 예측 시스템을 구축함에 있어, 기존 히트펌프 제품의 사용경험지식을 신규 히트펌프 제품의 학습데이터를 보조하는 정보로 활용하여 학습데이터 수의 부족문제를 극복하는데 있다. 일례로, 기존 히트펌프로부터 수집된 운전기록데이터를 기초로 기존 히트펌프 성능 예측을 위한 인공신경망의 입력데이터와 출력데이터를 갖는 데이터베이스를 구축하는 기존 히트펌프 데이터베이스 구축부; 상기 데이터베이스를 기반으로 인공신경망의 훈련 및 예측 과정을 수행하되, 상기 예측 과정에서 인공신경망 예측 샘플을 생성하고, 상기 인공신경망 예측 샘플을 기초로 신규 히트펌프 성능 예측에 대한 불확실성 분석 과정을 수행하여 상기 인공신경망 예측 샘플의 평균 백터를 도출하는 기존 히트펌프 성능 예측부; 표준시험환경의 챔버장치를 활용하여 신규 히트펌프의 측정데이터를 수집하는 신규 히트펌프 데이터 수집부; 및 상기 데이터베이스의 입력데이터 및 상기 기존 히트펌프 성능 예측부의 평균 벡터를 출력데이터로 포함하는 시스템 초기정보를 |
---|