Huotari Matti; Machine Learning Applications for Energy Utilization of Smart Buildings / Älyrakennusten energiankäytön koneoppimismalleista. (Väitöskirja)
Tiivistelmä
Älyenergiaratkaisut tarjoavat joustavaa ja kestävää energiaa rakennuksissa ja liikenteessä.Älyenergiasovelluksia käytetäänkin enenevässä määrin rakennetussa ympäristössä.Samanaikaisesti ratkottavia ongelmia älyenergiasovelluksissa ovat energiansäästö,lämpöviihtyvyys, ja epänormaalien tilanteiden hallinta. Uusiutuvien energianlähteiden lisääntynyt käyttö ja lainsäädännön tiukentuneet vaatimukset energiankäytöstä vaativat uusia ratkaisuja, joita älyenergiasovellukset voivat tarjota. Ne soveltuvat erityisen hyvin, jos vaatimuksia on monia tai neovat jopa näennäisesti ristiriitaisia. Tällaisia vaatimuksia ovat muun muassa järjestelmien käytön optimointi, kulujen vähentäminen, uusiutuvan energian liittäminen järjestelmään ja käyttäjien lämpöviihtyvyys.
Tämä tutkimus pohtii älyenergiasovelluksia rakennuksissa. Aihetta lähestytään tutkimuskysymysten kautta: (i) Minkä algoritmien avulla voidaan tutkia akkulaitteiston ikääntymistä ja mikä on algoritmisesta mallista saatujen tulosten epävarmuus? (ii) Kuinka mallintaa, kun pohjadata on epätasapainoissa ja ei ole täydellisen kattavaa (esimerkkinä ilmanvaihtolaitteen data)? (iii) Miten ottaa huomioon ihmiset lämpöviihtyvyyden määrityksessä ja siihen liittyvässä päätöksenteossa? (iv) Minkälaisia tarpeita on olemassa älyenergiasovelluksille ja mitkä sovellukset tarjoavat ratkaisuja niille?
Jokaista tutkimuskysymystä käsiteltiin uusilla keinoilla, kehitetyt ratkaisut luokiteltiin taksonomisesti, ja kaikki tutkimustulokset julkaistiin lopulta viitenä julkaisuna. Ehdotettujen sovellusten sopivuus varmennettiin tapaustutkimusten avulla. Koneoppimista voidaan soveltaa monitahoisiin ongelmiin rakennusten energiankäyttöön liittyen. Saadut tutkimustulokset antavat viitettä siitä, että ehdotetut sovellukset voivat ratkaista ongelmia, jotka liittyvätenergiatehokkuuteen, lämpöviihtyvyyteen ja huoltoon.
Otto Räisänen. Open data in distribution network asset management. 3.11.2023 (Väitöskirja).
Tiivistelmä Climate change and the energy transition are causing major changes to the distribution network environment. The electrification of society raises expectations for clean, reliable, and affordable electricity supply. Meeting these increased expectations in the changing environment requires DSOs to incorporate large amounts of external data into their analysis procedures to properly assess the impacts of different trends, such as electrification of transportation and an increase in distributed generation. The amount of data available to DSOs has grown significantly through simultaneous advances in internal data collection, such as smart meters, and an increase in external data sources. A large amount of data with potential use cases in distribution asset management is nowadays available as open data. Furthermore, significant efforts are made by different organizations to increase the accessibility and quality of open data through standardization and creation of data portals. However, thus far, DSOs have not taken advantage of the full potential of open data.
In this doctoral dissertation, the use of open data in asset management is studied by an extensive review of open data sources and types and studies using open data in cases related to distribution asset management. The potential of open data is investigated in more detail through three case studies. In the first case study, a novel methodology for long-term energy demand forecasting based on open data is proposed consisting of a Laspeyeres decomposition analysis combined with time series analysis and forecasting. The methodology is shown to produce adequate results; however, the results could be improved with additional data. In the second case study, a geospatial analysis method is used to estimate the generation potential of large-scale PV farms built in peat production areas. Furthermore, the potential for P2X generation is assessed. The results show that peat farms are typically located close to existing electricity networks and P2X end use and transmission infrastructure, making them suitable locations for large-scale PV systems. The third case study focuses on developing a novel crown-snow load outage risk estimation method for overhead lines. The model uses a combination of aerial inspection data gathered by a DSO and a variety of open data variables, such as tree height, species, and diameter, and an advanced forest crown snow-load risk map product originally produced for forestry purposes as explanatory variables. The interactions between the variables and the model are studied with explainable AI methods. The model shows a good performance and is able to produce accurate risk estimates for up to 50 m resolution.
The outage risk model developed in this dissertation could be used by DSOs for more targeted supply security investments and improved operational outage management, leading to significant cost savings. The energy demand forecasting and methodologies for estimation of large-scale solar PV generation potential enhance the DSOs’ understanding of load development in their network areas and can be used as a basis of long-term strategic planning. This dissertation shows that there are a wide variety of open data with potential to bring value to all aspects of distribution asset management.
Kalevi Härkönen. Smart buildings in the green energy transition. 31.10.2023 (Väitöskirja).
Tiivistelmä In the interconnection between smart buildings and energy systems, the growing importance to develop flexibility in the way in which buildings consume energy has been recognized. Owing to the advances of internet and communications technology (ICT), increasingly smaller loads in buildings are to provide energy consumption flexibility, but the heterogeneity of buildings and the prospective customer base create barriers to their cost-effective implementation.
This research investigates how the utilization of energy consumption flexibility in buildings can be enhanced in the nexus between electric energy systems, smart buildings, and service providers. The objective of the study is to support the decision-making of owners of smart buildings and consumption flexibility service providers. The study comprises four individual case studies. Exploratory, qualitative, and abductive methodological choices are followed.
The results suggest that professionally managed non-residential buildings seem well suited for consumption flexibility programs that support the balancing of the electric grid. The current techno-economic flexibility potential of residential buildings is, however, more limited. Standardization is essential in achieving interoperability between the variety of systems in modern smart buildings, and it may provide significant benefits to both the industry and consumers. However, it seems likely that the evolving adoption of standards among smart buildings has not resulted in building management system vendors adjusting their strategies accordingly.
This research supports owners of smart buildings and consumption flexibility service providers by offering tools to overcome the barriers of progress, eventually contributing to the economically viable development towards sustainable energy systems. It also expands the research on the demand-side management of smart buildings, as previous studies on the topic predominantly apply to buildings in warmer climates.
Wind energy is an integral resource for the transition of the energy sector to cleaner electricity production. While the advantages, such as its fuel- and water-free production, are evident, at present the economic viability compared with traditional fossil energy sources is uncertain. As each project is site-specific through its unique topography and wind map, wind farms require a custom design.
This doctoral dissertation focuses on optimizing wind farm design and provides guidance for solving this task. The aim of the study is to investigate the extent to which the layout affects the profitability of a wind farm project and, more specifically, to identify the factors that most influence the optimal solution. A methodology using techno-economic performance metric is developed. The rationale for the selection of this metric is also explained.
The wind farm layout problem can be formulated as a mixed-integer nonlinear programming problem with nonlinear constraints, in which the location of each turbine is considered continuous, and their type and total number are considered discrete. A relatively recent metaheuristic nature-inspired algorithm is used for the optimization routine. Furthermore, for wind farm infrastructure design, the dissertation presents a novel combined road and electrical cable layout search approach based on a least cost pathfinding algorithm. This approach considers the dependence of the cable laying costs on the road network solution. For each layout produced by the main algorithm, the infrastructure design is found.
Another focus of the study is to determine whether uncertainty in input parameters affects the design of a wind farm. Using a global variance-based sensitivity analysis technique, the work takes into account the uncertainties of the input parameters and analyzes their impact on the financial viability. A project risk assessment method is presented, which can be incorporated into the optimization framework. The analysis reveals that for a given number of turbines the risk of the project cannot be mitigated through the locations of the turbines.
The proposed optimization methodology performs reliably in solving the wind farm optimization problem. It was found that the number and location of turbines are driven by the existing infrastructure, the wind direction and the fixed part of the initial investments. Yet, the viability of the project as a whole is determined by the average wind speed, the price of electricity, and the discount rate.