Akademiska Hus AB är ett statligt ägt fastighetsbolag med uppdrag att äga och förvalta fastigheter för utbildning, forskning och studentbostäder. Ett av deras energimål är att halvera mängden inköpt energi till år 2025 med startår 2000. Syftet med detta examensarbete är att undersöka förbättringsmöjligheter och åtgärder kopplade till VVS-system för att nå energimålet. Arbetet har utförts på förvaltningsområdet Campus Solna som har ett specialavtal med Stockholm Exergi för fjärrvärme och fjärrkyla. Specialavtalet är utformat med fördelaktiga tariffer vid hög nyttjandegrad, ett exempel är kompensation vid användning av fjärrkyla under vinterperioden men det föreligger tilläggsavgifter vid underutnyttjad leverans. Detta sätter begränsningar i vilka typer av besparingsåtgärder som kan utföras med positivt ekonomiskt resultat samt minskad andel inköpt energi. De besparingsåtgärder som har valts för beräkningar är två typer av värmepumpslösningar och ett byte till effektivare värmeåtervinningsbatterier. Beräkningarna har utförts med hänsyn till gällande avtal och därefter jämförts mot ett annat befintligt standardavtal inom Akademiska Hus för att undersöka om förvaltningsområdet Campus Solna skulle gynnas av en annan avtalsform. Nyttjandetiden för fjärrvärme och fjärrkyla är ett förhållande mellan energi och effekt visar resultatet att de åtgärder där den procentuella andelen mellan energi- och effektbesparing är ungefär lika stor finns största ekonomiska lönsamheten. Resultatet efter utförda beräkningar visar att de två typerna av värmepumpslösningar ger stora besparingar i inköpt energi men i förhållande till effektbesparingarna generar detta tilläggsavgifter och en ekonomisk förlust. Vid jämförelse mot standardavtalet utan tilläggsavgifter sparas en stor andel inköpt energi och den potentiella ekonomiska besparingen ökar. Dock, visade det sig vid en totalkostnadsanalys att den årliga kostnaden för värmepumpslösningen med standardavtalet blir lika stor som kostnaden med det befintliga avtalet utan någon åtgärd och detta på grund av att tarifferna för inköpt fjärrvärme och fjärrkyla är högre. Värmeåtervinningsbatterierna har ett bättre förhållande mellan energi- och effektbesparing vilket resulterar i en ekonomisk vinst även vid gällande avtal. Eftersom återvinning av värme minskar andelen förbrukad energi är det även positivt ur en miljö- och hållbarhetssynpunkt. Slutsatsen är att om det befintliga avtalet fortsätter att nyttjas måste åtgärder som skapar ett bra förhållande mellan energi- och effektbesparing hittas. Alternativet är att försöka förhandla fram ett nytt avtal när gällande specialavtal löper ut för att göra åtgärder som värmepumpar lönsamma. Utöver ovannämnda undersökning har arbetsmetoderna för energibesparingar på Akademiska Hus undersökts genom kvalitativa intervjuer där syftet var att belysa hur nuvarande samarbete ser ut mellan de olika yrkesrollerna samt om det finns en tydlig strategi för hur energimålet ska nås. En slutsats efter utförda intervjuer är att det upplevs att det finns för lite tid att genomföra ett energiarbete. Bakomliggande orsak till detta anses vara att det skulle behövas en bättre arbetsmetodik, tydligare struktur för arbetsfördelning samt mer resurser i form av en dedikerad arbetsgrupp som jobbar operativt med energifrågan. / Akademiska Hus AB is a government owned real estate company, their primary mission is to own and administrate facilities for education, research and student housing. One of their energy goals is to reduce the amount of acquired energy by half between the year 2000 and 2025. The purpose of this thesis is to find solutions and ways to make improvements regarding HVAC-systems to achieve this energy goal. The work has been carried out at the property area Campus Solna, which have an agreement with special amendments and conditions with Stockholm Exergi concerning district heating and cooling. The agreement is designed with beneficial tariffs at high utilization, one example is that during the winter season Akademiska Hus gets compensated for using the district cooling but it also exists surcharges at underutilization of delivered heating and cooling. This puts limitations in which types of measures that can be performed to reduce the amount of acquired energy and at the same time have a positive economical result. The selected solutions are two different types of heat pumps and a change to more efficient heat recovery exchangers for ventilation. The calculations have been executed with regard to the current special agreement and then compared to another existing agreement within the corporation to examine if Campus Solna would benefit from another agreement. The utilization time for the district heating and cooling consist of a relation between energy and power, results show that saving measures where the ratio between them are about the same size is the most profitable economically. The results after the performed calculations show that the two types of heating pumps generate large savings in acquired energy but in relation to the reduced power usage causes surcharges and an economical loss. When compared to the other agreement without the additional charges the potential economical outcome improves while reducing the acquired energy. However, when performing a total cost analysis, the calculations with the other agreement indicate that the total annual cost for acquired energy with the heating pump solutions will amount to the same as using the existing agreement without performing any energy saving measures. This is due to the difference in tariffs between the agreements. The heat recovery exchangers have a better relation between energy and power savings, which results in an economical profit regardless of the agreements. This measure also decreases the consumed energy, which has a positive impact on the environment and sustainability. In conclusion, if the current agreement is to be used in the future energy saving measures have to meet the requirement of the energy and power ratio. However, finding enough of such measures that helps Campus Solna to achieve their energy goal can be problematic. The alternative is to renegotiate the agreement to make solutions like heat pumps profitable. Through qualitative interviews with a selection of personnel, work methods connected to energy saving has been discussed to see if there are possibilities for improvements regarding collaboration between the different departments to achieve the energy goal. One conclusion after the interviews is that the personnel are experiencing that there is not enough time to focus on energy savings and projects in their daily work. Hence, there is a need for better working methods, a clearer structure for how the work is to be distributed and more resources in the form of a dedicated working group that works operationally with the energy issue.

In 2016 German data centers consumed 12.4 terawatt-hours of electrical energy, which accounts for about 2% of Germany’s total energy consumption in that year. In 2020 this rose to 16 terawatt-hours or 2.9% of Germany’s total energy consumption in that year. The ever-increasing energy consumption of computers consequently leads to considerations to reduce it to save energy, money and to protect the environment. This thesis aims to answer fundamental questions about the energy consumption of software, e. g. how and how precise can a measurement be taken or if CPU load and energy consumption are correlated. An overview of measurement methods and the related software tooling was created. The most promising approach using software called 'Scaphandre' was chosen as the main basis and further developed. Different sorting algorithms were benchmarked to study their behavior regarding energy consumption. The resulting dataset was also used to answer the fundamental questions stated in the beginning. A replication and reproduction package was provided to enable the reproducibility of the results. / Im Jahr 2016 verbrauchten deutsche Rechenzentren 12,4 Terawattstunden elektrische Energie, was etwa 2 % des gesamten Energieverbrauchs in Deutschland in diesem Jahr ausmacht. Im Jahr 2020 stieg dieser Wert auf 16 Terawattstunden bzw. 2,9 % des Gesamtenergieverbrauchs in Deutschland. Der stetig steigende Energieverbrauch von Computern führt folglich zu Überlegungen, diesen zu reduzieren, um Energie und Geld zu sparen und die Umwelt zu schützen. Ziel dieser Arbeit ist es, grundlegende Fragen zum Energieverbrauch von Software zu beantworten, z. B. wie und mit welcher Genauigkeit gemessen werden kann oder ob CPU-Last und Energieverbrauch korrelieren. Es wurde eine Übersicht über Messmethoden und die dazugehörigen Softwaretools erstellt. Der vielversprechendste Ansatz mit der Software 'Scaphandre' wurde als Hauptgrundlage ausgewählt und weiterentwickelt. Verschiedene Sortieralgorithmen wurden einem Benchmarking unterzogen, um ihr Verhalten hinsichtlich des Energieverbrauchs zu untersuchen. Der resultierende Datensatz wurde auch zur Beantwortung der eingangs gestellten grundlegenden Fragen verwendet. Ein Replikations- und Reproduktionspaket wurde bereitgestellt, um die Reproduzierbarkeit der Ergebnisse zu ermöglichen.

No description available.

The aim of this thesis is to optimize hydro power plants with data generated from observations and field tests at the plants. The output is optimal production tables and curves in order to operate and plan hydro power plants in an optimized way concerning power output, efficiency and distribution of water. The thesis is performed in collaboration with Vattenfall AB, which currently use an internal optimization program called SEVAP. Two alternative methods have been selected, employed and compared with the current optimization program, these are Interior-Point Method and Sequential Quadratic Programming. Three start-point strategies are created to increase the probability of finding a global optima. A heuristic rule is used for selection of strategy in order to prevent rapid changes in load distribution for small variations in dispatched water. The optimization is performed at three plants in Sweden with different size and setup. The results of this evaluation showed marginally better results for the employed methods in comparison to the currently used optimization. Further, the developed program is more flexible and compatible to integrate with future digitalization projects. / Syftet med detta examensarbete är att optimera vattenkraftverk med data som genererats från indextester vid kraftverken. Resultatet är optimala produktionstabeller och kurvor för drift och planering av vattenkraftverk. Dessa är baserade på att optimalt fördela vattnet mellan aggregaten för att maximera uteffekt och verkningsgrad. Detta arbete har utförts i samarbete med Vattenfall AB, som för närvarande använder ett internt optimeringsprogram som heter SEVAP. Två optimeringsmetoder har valts, implementerats och jämförts med det nuvarande optimeringsprogrammet. Dessa metoder är inrepunktsmetoden (IPM) och sekventiell kvadratiskt programmering (SQP). Tre startpunktsstrategier har används för att öka sannolikheten att hitta ett globalt optima. För att förhindra hastiga förändringar i lastfördelning för små variationer av avsänt vatten har en heuristisk regel används. Optimeringen har utförts på tre stationer med olika uppsättning och storlek. Resultatet av detta examensarbete visar marginellt bättre resultat för de använda metoderna i jämförelse med den nuvarande optimeringen. Det utvecklade programmet är flexibelt och kompatibelt att integrera med framtida digitaliseringsprojekt.

No description available.

No description available.

The acceleration towards achieving a low carbon society has raised many challenges in the energy sector. The existing systems, highly dependent on fossil fuels, are not sustainable and recommendations are made to accelerate the transition by using more renewable and low-carbon sources of energy. By being responsible for over 70 % of the greenhouse gas emissions, cities or districts have a major role to play in this transition and present a large potential for implementation of renewable energy systems. The optimization of those systems and a better use of energy are crucial to reduce emissions and reach carbon neutrality. This study evaluates the potential of the implementation of three storage solutions for a decentralized energy system in a residential cluster in Ludvika, Sweden, equipped with photovoltaics panels. The first solution includes stationary batteries, the second includes a hydrogen storage solution and the third offers a hybrid solution from the two previous storages. Simulations were conducted using two numerical tools with an hourly resolution.The first scenario was conducted using Spine modelling tools, and the other simulations, including the cost analysis, were conducted on Excel with the support of Visual Basic for Applications. The comparison between the three solutions showed that the stationary batteries, blessed with a higher efficiency, offers the best results. The autonomy of the site, initially at 19.1 %, was raised to 22.8 % due to the batteries, and the system allowed to save up to 9.6 MW∙h per year. The investment price of the battery system was the highest of the three solutions. However, the payback period is reached in 20 years, within the average lifespan of the batteries and 10 years shorter than with the other solutions. The results also showed that the benefits of implementing a hydrogen storage solution were not visible as the excess in the photovoltaic production was not enough to cover the energy demand in a long-term period.

The remarkable surge of artificial intelligence has placed a new emphasis on energy optimization in high-performance computing systems (HPCs). Energy consumption has become a critical consideration and limiting factor for the utility of machine learning applications, with each iteration consuming significantly more energy than the last. Understanding the distribution of energy consumption across different system components is crucial for enhancing the efficiency of systems that drive machine learning applications. This understanding allows us to identify optimal system configurations for machine learning-dedicated machines and minimize the substantial energy costs associated with training and deploying machine learning models. In this thesis, we present our findings from measuring power consumption for individual system components, including the CPU, GPU, and disk drive, using the PowerPack framework and on-chip power estimates for the FT and MLPerf Inference benchmarks. The direct power consumption measurements taken by the PowerPack framework offer a level of accuracy, granularity, and synchronization across multiple system components that is not feasible with on-chip power estimates alone. PowerPack provides fine-grain application and component-level profiling without the overhead cost of simulation. The PowerPack framework consists of software tools, including the NI-DAQmx library, and hardware components, such as the NI cDAQ-9172 CompactDAQ Chassis and three NI Analog Input Modules, designed to take physical direct current (DC) power measurements with component-level granularity. PowerPack's physical power measurements were used to assess the accuracy of two on-chip power estimates: AMDuProf, a power and performance analysis tool specifically designed for AMD processors, and NVIDIA-SMI, a software power analysis tool developed for NVIDIA GPUs. The AC power draw of our system under test was measured using a HOBO Plug Load Data Logger to determine the energy consumption distribution among the CPU, GPU, and disk drive. Our findings reveal the power consumption characteristics for the seven major functions of the MLPerf Inference benchmark. The major functions that comprise MLPerf Inference are ONNX Runtime model prediction, system check, post-processing ONNX model output, image retrieval, backend initialization, and loading annotations using the COCO API. Defining the power consumption characteristics of each function with component-level granularity provides valuable insight for function execution scheduling and identifying bottlenecks in the machine learning inference phase. / Master of Science / The popularity of services that use machine learning, such as OpenAI's ChatGPT and Dall-E, Amazon Alexa, and content curation on social media, has made minimizing the cost of using machine learning incredibly important for businesses across all industries. Machine learning is a powerful tool that enables businesses to optimize production and improve their customer experience, but it can be tremendously expensive to develop and run these services. Training and running machine learning models consume large amounts of energy, which is a major contributor to their cost. The work presented in this thesis aims to identify how much energy is consumed by individual computer components while running machine learning applications. Understanding how energy consumption is divided among a computer's components is a crucial step toward reducing the energy required to run these applications, thereby making them more affordable.

The world has an imperative need to switch to clean and renewable energy sources. Effectively harnessing the vast energy that radiates on the earth's surface from the sun is a long standing goal of clean energy proponents. This work introduces the novel concept of a solar energy belt stretched across America. The thesis provides the first steps towards showing that not only can such a belt be built, but it can effectively yield a substantial quantity of America's energy in a cost effective fashion. In order to maximize impact of the work, a general mathematical model is first derived that may be extended by any interested party. The model is then viewed from the perspective of mathematical combinatorics and is thus made subject to combinatorial optimization. Genetic Algorithms are used for this optimization scheme. A GUI and simulation is provided that demonstrates the evolution of the solar belt solution in real-time. Results are demonstrated.

O notável desenvolvimento da indústria eletrônica observado nos últimos tempos tem permitido aplicações de conjuntos integrados de sensores em ambientes sem fio, conhecidos por wireless sensor networks (WSN), que passam por sensoriamento de processos industriais, ambientes tóxicos, projetos militares de monitoração de variáveis de segurança, até observação de fenômenos físicos naturais. Uma das principais especificações de uma rede WSN, o consumo de energia afeta diretamente a capacidade e tempo de vida útil do sistema, pois, na maioria dos casos, seus módulos possuem baterias independentes e sua substituição nem sempre é tarefa simples. Motivado pela necessidade de oferecer robustez e economia de energia nas redes WSN, o protocolo difusão direcionada se baseia na centralização de dados e a identificação de seus módulos é feita por meio de pares valor-atributo. Sua estrutura permite a adição de componentes de software que podem atuar na análise e modificação dos dados recebidos com o objetivo de alterar o protocolo original. O objetivo principal deste trabalho é investigar um método para otimização do uso de energia disponível em redes WSN com intuito de prolongar seu tempo de vida útil. O metódo consiste em analisar os valores das energias atribuídas aos módulos componentes da rede por meio de incorporação de um algoritmo de rotas baseado em energia ao protocolo difusão direcionada. Comparações de desempenho da rede em relação ao seu tempo de vida e energia dos módulos são realizadas utilizando o simulador NS-2. As simulações feitas em diversos cenários indicaram melhoria de desempenho em relação ao protocolo difusão direcionada original. Os cenários onde o protocolo original foi alterado apresentaram um número maior de rotas descobertas e possibilitaram um aumento de pelo menos 22% no tempo de vida da rede, em relação ao protocolo original. / The notable development of electronic industry in the last years allows the implementation of sensor integrated circuits in wireless environments, known as wireless sensor networks (WSN), which leads to industrial process sensing, toxic environments, military security monitor projects and natural physical phenomenon. As one of the main specifications of a WSN network, the energy consumption directly affects the capacity and the system useful lifetime, since most of the time its modules have independent batteries and their substitution is not always a simple task. Motivated by the need of offering robustness and energy economy for WSN networks, the directed diffusion protocol is data-centric based and its modules identification is made by attribute-value pairs. The directed diffusion structure enables the addition of software components which can act on the analysis and modification of received data with the objective of changing the original protocol. The main objective of this work is to investigate a method for optimization of available energy on WSN networks with the intention of increasing its useful lifetime. The method consists on analyzing the energy values attributed to the component modules of the network by incorporating an energy-based routing algorithm to directed diffusion protocol. Performance comparisons of the network related to its lifetime and modules energy are developed using the NS-2 simulator. Simulations performed in several scenarios indicated a better performance in relation to the original directed diffusion protocol. The scenarios where the original protocol was changed had larger number of discovered routes and allowed a rising of at least 22% on network lifetime, in relation to the original protocol.