kth.sePublications
12345674 of 20
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Data-Driven Strategies for Heat Pump Systems: A journey from inadequate data towards knowledge-based services
KTH, School of Industrial Engineering and Management (ITM), Energy Technology, Applied Thermodynamics and Refrigeration.ORCID iD: 0000-0002-1187-7065
2024 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

Integrating high-efficiency heat pumps to renewable electricity will significantly accelerate decarbonization progress. Despite the advancements in smart sensors and communication technologies that enhance data generation in heat pump units, much of this data remains underutilized for performance analysis. The primary issue is that the data often exhibits incompleteness, inconsistency, and inaccuracies. Consequently, data collection and storage impose an economic burden on manufacturers and end-users, and they have limited returns. This thesis aims to unlock the potential of various data resources, delivering knowledge-based services and addressing gaps in data availability and utilization.

The dissertation introduces the Data-Information-Knowledge-Service (DIKS) framework as an adaptation of the traditional Data-Information-Knowledge-Wisdom (DIKW) hierarchy, emphasizing the practical application of turning inadequate data into knowledge-based services in heat pump technologies. The transformative process within the DIKS pyramid is illustrated, detailing how each layer converts inadequate raw heat pump data into actionable, knowledge-based services. It begins with the aggregation and integration of various data types, followed by advanced processing techniques to refine data quality and identify significant patterns. This foundation of knowledge is then applied to improve heat pump services, demonstrating the practical benefits of this structured approach.

Five different scenarios are examined utilizing different types of data, including high-quantity, low-quality in-situ measurements, high-quality, low-quantity lab data, and technical specifications. The first scenario, utilizing in-situ field measurement data, develops an artificial neural network (ANN) model to create soft sensors that compensate for the absence of costly physical sensors in heat pump systems. These soft sensors use incomplete data to accurately estimate essential heat pump parameters, supporting functions like operational monitoring, fault detection, smart energy management, and developing digital twins. The second scenario, also utilizing in-situ field measurement data, focuses on models that prioritize minimal input features, enhancing models' usability across various installations. These models estimate power consumption effectively and compensate for the lack of physical power meters, thus facilitating network planning and smart control, etc. In the third scenario, transfer learning techniques are employed to estimate heat pump performance with limited lab data, particularly for natural refrigerants in the context of the phasing out the fluorinated refrigerants. This approach utilizes knowledge from existing refrigerant data to improve model reliability and accuracy, aiding in refrigerant selection. The fourth scenario develops polynomial regression models from technical specifications to evaluate heat pump performance without dynamic measurements. These models assist in tracking the system performance. The final scenario introduces semi-empirical models that use thermodynamic and heat transfer correlations to enhance the understanding of physical meaning. These models are designed with reduced parameters, improving services such as fault diagnosis and system maintenance.

In conclusion, this thesis identifies common status and key issues related to inadequate data from heat pump systems. Furthermore, this thesis proposes solutions to transfer inadequate data to useful structured information and develops data-driven models according to the characteristics of different data types. The models are validated against measurements, demonstrating accurate results. This thesis demonstrates the application of the DIKS framework to effectively harness underutilized inadequate data from heat pump systems, which not only provides insights to alleviate the economic burdens placed on manufacturers and/or users related to data cost but also enable the services of heat pump systems.

Abstract [sv]

Integreringen av högeffektiva värmepumpar med förnybar elektricitet förväntas påskynda minskningen av koldioxidutsläppen avsevärt. Trots framsteg inom smarta sensorer och kommunikationstekniker som förbättrar dataproduktionen i värmepumpsenheter, förblir mycket av denna värdefulla data outnyttjad och används främst för enklare övervaknings- och felvarningsfunktioner snarare än för omfattande analys. Det största problemet är att datan ofta brister i fullständighet, konsekvens och noggrannhet. Detta innebär att datainsamling och lagring är en ekonomisk börda för tillverkare och slutanvändare som inte utnyttjas. Denna avhandling syftar till att fullt ut utnyttja hela potentialen hos olika dataresurser, så att kunskapsbaserade tjänster kan levereras och att hantera saknade data.

Avhandlingen introducerar ramverket ``Data - Information-Knowledge-Service" (DIKS) som en anpassning av den traditionella hierarkin ``Data-Information-Knowledge-Wisdom" (DIKW), med betoning på den praktiska tillämpningen av att omvandla inadekvat data till kunskapsbaserade tjänster hos värmepumpssystem. Den transformerande processen inom DIKS-pyramiden illustreras och visar detaljerat hur varje lager kan omvandla otillräckliga rådata från värmepumpar till användningsbara, kunskapsbaserade tjänster. Ramverket börjar med sammanställning och integration av olika datatyper, följt av avancerade bearbetningstekniker för att förbättra datakvaliteten och identifiera signifikanta mönster. Denna kunskapsgrund tillämpas sedan för att förbättra värmepumptjänster, vilket visar de praktiska fördelarna med detta strukturerade tillvägagångssätt.

Fem olika scenarier undersöks med olika typer av av dataanvändning, inklusive stora mängder av data med låg kvalitet från fältmätningar, högkvalitativa data av låg kvantitet från laboratoriedata och slutligen katalogdata. Det första scenariot, som använder data från fältet, utvecklar modeller med hjälp av artificiella neurala nätverk (ANN) för att skapa mjuka sensorer som kompenserar för avsaknaden av kostsamma fysiska sensorer i värmepumpsystem. Dessa mjuka sensorer använder ofullständiga data för att uppskatta centrala värmepumpparametrar, vilket vidare stödjer funktioner som operativ övervakning, feldetektering, smart energihantering och utveckling av digitala tvillingar. Det andra scenariot, som också använder data från fältet, fokuserar på modeller som prioriterar minimala inmatningsfunktioner, vilket förbättrar modellernas användbarhet i olika installationer. Dessa modeller uppskattar effektförbrukningen effektivt och kompenserar för bristen av fysiska elmätare, vilket underlättar nätverksplanering och smart styrning för att spara energi. I det tredje modellscenariot används Transfer Learning (TL)-tekniker för att uppskatta värmepumpprestanda med begränsade laboratoriedata, särskilt för värmepumpar med kolvätebaserade köldmedier i samband med utfasning av fluorerade köldmedier. Denna metod använder kunskap från befintliga köldmediadata för att förbättra modellernas tillförlitlighet och noggrannhet, vilket kan hjälpa till med optimering och val av köldmedium. Det fjärde scenariot utvecklar polynomregressionsmodeller från katalogdata för att utvärdera värmepumpprestanda utan sensoravläsningar. 

Dessa modeller hjälper till att välja lämpliga värmepumpsmodeller och stödjer planering av nätkapacitet. Det sista scenariot introducerar semi-empiriska modeller som använder termodynamiska och värmeöverföringskorrelationer för att förbättra tolkningsbarheten av datadrivna modeller. Dessa modeller är designade med reducerade parametrar, vilket förbättrar tjänster som feldetektering och systemunderhåll. 

Sammanfattningsvis visar denna avhandling tillämpningen av DIKS-ramverket för att effektivt utnyttja underutnyttjad otillräcklig data från värmepumpssystem, och omvandla den till praktiska kunskapsbaserade tjänster. De utvecklade modellerna minskar inte bara de ekonomiska bördorna för tillverkare och/eller användare i samband med datakostnader, utan förbättrar också avsevärt tjänsterna för värmepumpssystem.

Place, publisher, year, edition, pages
Stockholm: KTH Royal Institute of Technology, 2024. , p. 78
Series
TRITA-ITM-AVL ; 2024:18
Keywords [en]
Heat pump, Heating, Data-driven, Machine Learning, Soft sensors, ANN, Regression, Semi-empirical model, Transfer Learning, Natural refrigerants
National Category
Energy Engineering
Research subject
Energy Technology
Identifiers
URN: urn:nbn:se:kth:diva-353212ISBN: 978-91-8106-042-3 (print)OAI: oai:DiVA.org:kth-353212DiVA, id: diva2:1898186
Public defence
2024-10-11, Kollegiesalen, Brinellvägen 8, Stockholm, 10:00 (English)
Opponent
Supervisors
Available from: 2024-09-17 Created: 2024-09-16 Last updated: 2024-09-23Bibliographically approved
List of papers
1. Data-driven soft sensors targeting heat pump systems
Open this publication in new window or tab >>Data-driven soft sensors targeting heat pump systems
Show others...
2023 (English)In: Energy Conversion and Management, ISSN 0196-8904, E-ISSN 1879-2227, Vol. 279, p. 116769-, article id 116769Article in journal (Refereed) Published
Abstract [en]

The development of smart sensors, low cost communication, and computation technologies enables continuous monitoring and accumulation of tremendous amounts of data for heat pump systems. But the measurements, especially for domestic heat pump, usually suffer from incompleteness given technical and/or economic barriers, which prevents database of measurements from being exploited to its full potential. To this end, this work proposes a data-driven soft sensor approach for compensating multiple missing information. The soft sensors are developed based on an ANN model, an integrated multivariate polynomial regression model and empirical model by considering different constrains like data and information availability during model establishing process. All the three models have been validated against the data from a field test installation, and showed good performance for all the compensated variables. Of the three models, the ANN model shows the best performance for all soft sensors, but it has the highest requirement for additional resources to collect training data. While the integrated multivariate polynomial regression model demonstrates excellent accuracy for the majority of soft sensors with manufacturers' subcomponent data which needs no extra cost. Even though empirical model is not as accurate as the other two models, it still performs good accuracy with limited information from performance map. The methods developed in the present study paves the way for available measured data in thousands of installations to be fully utilized for innovative services including but not limited to: improved heat pump control strategies, fault detection and diagnosis, and communication with local energy grids.

Place, publisher, year, edition, pages
Elsevier BV, 2023
Keywords
Data driven, Heat pumps, Soft sensors, ANN, Regression, Database
National Category
Energy Engineering
Identifiers
urn:nbn:se:kth:diva-324638 (URN)10.1016/j.enconman.2023.116769 (DOI)000933059100001 ()2-s2.0-85147371380 (Scopus ID)
Note

QC 20230309

Available from: 2023-03-09 Created: 2023-03-09 Last updated: 2024-09-16Bibliographically approved
2. Estimating electric power consumption of in-situ residential heat pump systems: A data-driven approach
Open this publication in new window or tab >>Estimating electric power consumption of in-situ residential heat pump systems: A data-driven approach
Show others...
2023 (English)In: Applied Energy, ISSN 0306-2619, E-ISSN 1872-9118, Vol. 352, article id 121971Article in journal (Refereed) Published
Abstract [en]

International Energy Agency predicts that the global number of installed heat pumps (HP) will increase from 180 million in 2020 to approximately 600 million by 2030, covering 20% of buildings heating needs. Electric power consumption is one of the main key performance indicators for the heat pump systems from techno-economic perspective. However a common issue prevalent in many existing heat pumps is the lack of electric power measurement. The modern installations might be equipped with electric power measurement sensors but this comes at a higher system cost for the manufacturers and end-users. The primary objective of this work is to propose a virtual measurement for estimating power consumption, thereby eliminating the need for field measurement of power for heat pumps. To achieve the objective, a data-driven approach is proposed. Firstly, the in-situ data is preprocessed through data merging, cleaning, and normalization. Then, input features are pre-selected using Spearman correlation coefficients, and further refined by addressing multicollinearity problem. Following this, Extreme Gradient Boosting (XGBoost) models and polynomial models are developed by considering different features as inputs. All models are finally validated against the in-situ data from multi-units of ground source heat pump (GSHP) and air source heat pump (ASHP) installations. The results showed that the electric power consumption of GSHP can be estimated with high accuracy (99% for R2, 10 W for MAE, and 1% for MAPE) through generic data-driven models using only four easy-to-measure input features. Taking three input features as inputs for ASHP generic model, the accuracy can be reached to 83% for R2, 125 W for MAE, and 9% for MAPE. The method presented in this paper can be applied to estimate power consumption of millions of heat pumps and consequently add a significant value as well as provide different types of services, such as cost-saving benefits for manufacturers and end-users, flexibility services for aggregators and electricity grids.

Place, publisher, year, edition, pages
Elsevier BV, 2023
Keywords
Data driven, Electric power, Heat pump, Heating, Machine learning, Regression model
National Category
Energy Engineering
Identifiers
urn:nbn:se:kth:diva-338357 (URN)10.1016/j.apenergy.2023.121971 (DOI)001086100200001 ()2-s2.0-85172678028 (Scopus ID)
Note

QC 20231115

Available from: 2023-10-20 Created: 2023-10-20 Last updated: 2024-09-16Bibliographically approved
3. Innovative approaches to overcome inadequate measurements in heat pumps with non-fluorinated refrigerants
Open this publication in new window or tab >>Innovative approaches to overcome inadequate measurements in heat pumps with non-fluorinated refrigerants
2024 (English)In: Energy Conversion and Management, ISSN 0196-8904, E-ISSN 1879-2227, Vol. 319, article id 118970Article in journal (Refereed) Published
Abstract [en]

As the transition away from fluorinated refrigerants occurs due to F-gas and PFAS regulations, heat pumps face the challenge of adapting to new non-fluorinated refrigerants. Evaluating heat pump performance during this transition is challenging due to limited operational data on the new refrigerants. Conducting long-term tests to fully understand a heat pump’s performance with all possible refrigerants is labor-intensive and economically burdensome. This study introduces two complementary reduced-parameter models to assess heat pump performance across multiple new natural refrigerants despite limited data. A transfer learning model, leveraging knowledge from existing data-rich refrigerants, has been developed to evaluate the performance of heat pumps using new, data-scarce natural refrigerants. However, due to the lack of transparency in transfer learning models, semi-empirical models are being developed in parallel. The semi-empirical models, across multiple natural refrigerants, are capable of analyzing the thermodynamics and heat transfer processes within the heat pump system by utilizing only limited easy-to-measure variables as inputs. The transfer learning model demonstrates high accuracy for all outputs across seven refrigerants with RRMSE all below 7%. In comparison, the semi-empirical models are less accurate, with RRMSE results under 25% for all parameters except compressor power. By integrating these two models, a comprehensive framework is established for assessing heat pump performance with both high accuracy and a deeper understanding of the system.

Place, publisher, year, edition, pages
Elsevier BV, 2024
Keywords
Heat pump, Data driven, Machine learning, Transfer learning, Semi-empirical model, Reduced-parameter
National Category
Engineering and Technology
Identifiers
urn:nbn:se:kth:diva-352704 (URN)10.1016/j.enconman.2024.118970 (DOI)001309019400001 ()2-s2.0-85202174400 (Scopus ID)
Funder
Swedish Energy Agency
Note

QC 20240906

Available from: 2024-09-05 Created: 2024-09-05 Last updated: 2024-10-04Bibliographically approved
4. Development and validation of data-driven soft sensors for heat pumps
Open this publication in new window or tab >>Development and validation of data-driven soft sensors for heat pumps
Show others...
2024 (English)In: Volume 41: Energy Transitions toward Carbon Neutrality: Part IV, Applied Energy Innovation Institute (AEii) , 2024, Vol. 41, p. 10988Conference paper, Published paper (Refereed)
Abstract [en]

Modern heat pump systems often come equipped with sensors, enabling the collection of substantial operational data. However, many residential heat pumps installed in preceding decades lack pressure sensors, energy meters, or mass flow meters, primarily due to financial limitations. As a result of these incomplete measurements, the direct analysis of the heat pump system’s performance or the leveraging of the amassed data for inventive applications like prognosticating energy consumption, detecting and diagnosing faults, and implementing intelligent control becomes challenging.In existing literature, the focus of soft sensors in heat pump systems has been on estimating a single parameter. This approach, however, overlooks the reality that multiple parameters are often missing due to the lack of all-encompassing physical meters and sensors. Furthermore, current soft sensor models are typically developed using inputs such as compressor power consumption, pressures, evaporation, and condensation temperatures. These inputs, unfortunately, tend to be inaccessible within existing heat pump monitoring installations.In practice, it is a challenge to compensate for several critical measurements, encompassing mass flow rate, pressures, power consumption, and heating capacity, by using only commonly available sensors such as secondary loop temperatures and compressor frequency are available. Currently, there is a notable gap in research concerning this practical issue.To address the problems associated with inadequate measurements, this study presents the development and validation of soft sensors based on a data-driven approach, which can compensate for the parameters often unavailable with data collected from a limited number of commonly used sensors. Each component model employs a multivariate polynomial regression that calculates the evaporation temperature, condensation temperature, mass flow rate, and compressor power consumption, respectively. Subsequently, we present an integrated heat pump model that combines these component models into a comprehensive heat pump model.Finally, we validate the data-driven model against field test installations, demonstrating its accuracy with a relative root mean squared error (RRMSE) ranging from 10% to 20%.

Place, publisher, year, edition, pages
Applied Energy Innovation Institute (AEii), 2024. p. 10988
National Category
Engineering and Technology Energy Engineering
Identifiers
urn:nbn:se:kth:diva-352772 (URN)10.46855/energy-proceedings-10988 (DOI)
Conference
International Conference on Applied Energy (ICAE2024), Niigata City, Japan, Sep 1-5, 2024
Note

QC 20240906

Available from: 2024-09-05 Created: 2024-09-05 Last updated: 2024-09-16Bibliographically approved
5. Data-driven approaches in building heating and cooling
Open this publication in new window or tab >>Data-driven approaches in building heating and cooling
(English)Manuscript (preprint) (Other academic)
Keywords
Building energy system, data-driven, Algorithms, Heating, Cooling
National Category
Energy Engineering
Identifiers
urn:nbn:se:kth:diva-352774 (URN)
Note

QC 20240906

Available from: 2024-09-05 Created: 2024-09-05 Last updated: 2024-09-16Bibliographically approved

Open Access in DiVA

fulltext(81400 kB)11 downloads
File information
File name FULLTEXT01.pdfFile size 81400 kBChecksum SHA-512
2393638997766e8bfe34af1d473ee9d85d05cc723811b85b6ecf264480171a9d78a867e0c18c2f5592f7d27e8cf7222bbd543d0375c5adbab1acceba40b9818c
Type fulltextMimetype application/pdf

Authority records

Song, Yang

Search in DiVA

By author/editor
Song, Yang
By organisation
Applied Thermodynamics and Refrigeration
Energy Engineering

Search outside of DiVA

GoogleGoogle Scholar
Total: 11 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

isbn
urn-nbn

Altmetric score

isbn
urn-nbn
Total: 209 hits
12345674 of 20
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf