Denna avhandling presenterar ett tillvägagångssätt för att förutspå förflyttning inom folkmassor med hjälp av LSTM-neurala nätverk. Specifikt analyseras inflytandet som olika frekvenser av tidsserier har på både prognosen för folkmassorna och designen i arkitekturen inom transport och handel. Arkitekturen påverkas även då frekvensändringar provocerar fram en ökning eller minskning i datamängd och arkitekturen därför bör anpassas. Tidigare forskning inom prognoser relaterade till folkmassor har huvudsakligen fokuserat på att förutspå folkmassans nästa förflyttning snarare än att definiera mängden människor på en specifik plats under ett specifikt tidsspann. Dessa studier har använt olika tekniker som till exempel Random Forest eller Feed Forward neurala nätverk för att ta reda på inflytandet som de olika frekvenserna har över prognosens resultat. Denna avhandling tillämpar istället LSTM-neurala nätverk för analysering av detta inflytande och använder specifika fältrelaterade tekniker för att hitta de bästa parametrarna för att förutspå framtida välstånd i folkmassor. Resultatet visar att frekvensordningen i en tidsserie tydligt påverkar resultatet av prognoserna inom transport och handel, och att detta inflytande är positivt när frekvensordningen av tidsserierna kan fånga upp frekvensens form i prognosen. Därför, med frekvensordningen i åtanke, visar resultaten i prognoserna för de analyserade platserna en förbättring på 40% för SMAPE och 50% för RMSE jämfört med inhemska tillvägagångssätt och andra tekniker. Utöver detta visar de även att det finns ett samband mellan frekvensordningen och komponenterna i arkitekturerna. / This thesis presents an approach to predict crowd movement in defined placesusing LSTM neural networks. Specifically, it analyses the influence that different frequencies of time series have in both the crowd forecast and the design of the architecture in the domains of transportation and retail. The architecture is also affected because changes in the frequency provokes an increment or decrement in the quantity of data and, therefore, the architecture should be adapted. Previous research in the field of crowd prediction has been mainly focused on anticipating the next movement of the crowd rather than defining the amount of people during a specific range of time in a particular place. These studies have used different techniques such as Random Forest or Feed-Forward neural networks in order to find out the influence that the different frequencies have in the results of the forecast. However, this thesis applies LSTM neural networks for analysing this influence and uses specific field-related techniques in order to find the best parameters for forecasting future crowd movement. The results show that the order of the frequency of a time series clearly affects the outcomes of the predictions in the field of transportation and retail, being this influence positive when the order of the frequency of time series is able to catch the shape of the frequency of the forecast. Therefore, taking into account the order of the frequency, the results of the forecast for the analyzed places show an improvement of 40% for SMAPE and 50% for RMSE compared to the Naive approach and other techniques. Furthermore, they point out that there is a relation between the order of the frequency and the components of the architectures.

No description available.

No description available.

No description available.

No description available.

No description available.

Rogue waves are unexpected, individual ocean surface waves that are disproportionately large compared to the background sea state. They present considerable risk to mariners and offshore structures when encountered in large seas. Rogue waves have gone from seafarer’s folktales to an actively researched and debated phenomenon. In this work an easily derived spectral parameter, as an indicator of rogue wave risk, is presented, and further evidence for the generation mechanism responsible for these abnormal waves is provided. With the additional goal of providing a practical rogue wave forecast, the ability of a standard wave model to predict the rogue wave probability is assessed. Current forecasts, like those at the European Centre for Medium-Range Weather Forecasts (ECMWF), rely on the Benjamin Feir Index (BFI) as a rogue wave predictor, which reflects the nonlinear process of modulation instability as the generation mechanism for rogue waves. However, this analysis finds BFI has little predictive power in the real ocean. From the analysis of long term sea surface elevation records in nearshore areas and hourly bulk statistics from open ocean and coastal buoys in the Northeast Pacific, crest-trough correlation shows the highest correlation with rogue wave probability. These results provide evidence in support of a probabilistic prediction of rogue waves based on random linear superposition and should replace forecasts based on modulation instability. Crest-trough correlation was then forecast by a regional WAVEWATCH III ® wave model with moderate accuracy compared with the high performance of forecasting significant wave height. Results from a case study of a large fall storm October 21-22, 2021, are presented to show that the regional wave model produces accurate forecasts of significant wave height at high seas and presents a potential rogue wave probability forecast. / Graduate

The city of Stockholm has a goal of reducing traffic work within the municipality by the year 2030. To achieve this, a change in the travel pattern will be required for Stockholm County in its entirety. The population has increased in all municipalities in Stockholm County and that increase is expected to continue. In recent years, traffic work has decreased per capita but increased overall. This indicates that countermeasures are required to reduce the amount of traffic work. This study aims to investigate whether a reduction in the capacity of the Central Bridge can lead to a reduction in traffic.According to studies globally, a reduction in capacity leads to reduced traffic work rather than the system breaking down. Important lessons in avoiding system breakdown are that the road network is robust and that there are alternative means of transport. In this study, the Swedish Transport Administration's base forecast for the year 2040 has been used together with a compilation of similar projects globally and lessons learned from them. According to the base forecast, the traffic work is estimated to increase until 2040. If the Central Bridge has a reduced capacity, it leads to a one percentage point reduction in traffic work in the municipality of Stockholm per reduced lane. This further leads to better air quality in the inner city as traffic work moves out into the rest of the municipality and county.The study has used Sampers, the demand model that handles the Swedish Transport Administration's basic forecasts, to analyse various aspects of the transport system. Despite its usefulness, the model has some limitations such as the separation of the public transport road network and the car road network, where congestion is only considered for car traffic. This shortcoming means that the effects of a change in congestion do not affect public transport journey times. The model is good at identifying the impact of a change in the road network on vehicle traffic but has more difficulty in identifying transfers between transport modes. Nevertheless, the model gives some indication of a shift to more sustainable modes of transport such as public transport, walking and cycling. It also gives an indication of more local car travel as transport costs increase for trips to and from Stockholm city centre for motorists. Over both the Customs section and the Saltsjö-Mälar section, the share of public transport increases.

I en tid då samhället alltmer elektrifieras, står elnätsföretagen inför en växande utmaning att garantera att elsystemet kan hantera de ökande kraven. I detta sammanhang kan långtidsprognosverktyg spela en avgörande roll för att optimera det befintliga samt det framtida elnätet. Genom att använda sådana verktyg kan elnätsföretag förbättra sina planer för nätutveckling, vilket är viktigt för att främja en effektiv och hållbar utveckling av elnätet. Genom att följa välgrundade riktlinjer och bästa praxis kan man säkerställa tillgänglighet, pålitlighet och kostnadseffektivitet i elförsörjningen. Syftet med examensarbetet är att utforska vad som krävs för att hantera ett effektivt prognosverktyg och uppfylla föreskrift 7 § enligt EiFS 2024:1 från Energimarknadsinspektionen. Uppdraget kom som ett resultat av detta krav, där det anges att distributionsnätsföretaget ska rapportera sin prognos för överföringskapacitet i megawatt för att hantera kundernas energiförbrukning och produktion. Examensarbetet fokuserar på att utveckla metoder och använda programvaran AFS för att uppnå dessa mål samt att analysera nuvarande och framtida belastningar och parametrar som påverkar prognoserna En litteraturstudie har utförts kring de olika prognosverktygen och metoderna som finns ute på marknaden. Parallellt har fem olika elnätsföretag intervjuats för att undersöka vad för prognosverktyg de har utvecklat och hur de arbetar för att uppnå nätutvecklingsplanen. Utöver litteraturstudien och intervjuerna har prognosverktyget Aiolos Forecast Studio (AFS) analyserats för att undersöka dess potential vid långtidsprognostisering. Resultatet från litteraturstudien och intervjuerna visade att det krävs mer än bara prognosverktyg för att utföra långtidsprognoser. Litteraturstudien visade att det finns olika verktyg samt metoder för att utföra dessa prognoser, men att det beror på företagets specifika krav eller behov för vad som passar dem.  Intervjuerna avslöjade att många använder eller har använt Excel för prognosarbete, men att man behövde utveckla andra verktyg för att kunna hantera större mängder data då det kan överstiga Excels kapacitet. Med huvudverktyget som beräknar prognoserna, framkom det att det krävs en hel del förarbete. Samt att ett kompletterande verktyg som kan applicera prognoserna ute på elnätet är att föredra. Analysen av Aiolos Forecast Studio gav blandade resultat. Då uppdraget kom från Jämtkraft Elnät AB, är den bearbetade datan från deras lokalnät. AFS är ett givande verktyg för korttidsprognoser men går även att använda inför framtidsscenarion. De olika framtidsscenarion som arbetades med var om solelproduktionen skulle ha en ökad toppeffekt från år 2023 med 18 MW till 62 MW för 2030 och 150 MW för 2035. Det utfördes skalningar i programmet för att få de nya önskade toppeffekterna. Resultatet visade att de prognostiserade toppeffekterna för år 2030 samt 2035 låg på 14% och 20% ifrån det önskade toppvärdet. Detta beror på att när man arbetar med en regressionsmodell så kommer prognosen att skala relativt till det historiska utfallsdata, det vill säga 18 MW. Därefter analyserades skalningarna för ett årsperiod, inklusive en vecka med låg förbrukning och en vecka med hög förbrukning. Detta gav en tydlig inblick i hur solelproduktionen påverkade belastningen under olika förhållanden och med olika skalningar. Under lågförbrukningsveckan, som inträffade i juni, visade resultaten tydligt att vid en skalning på 150 MW skiftade belastningen från förbrukning till produktion under vissa timmar, med tydliga tecken på den så kallade ankkurvan. Under högförbrukningsveckan, som ägde rum i december, var resultaten knappt märkbara. Detta var förväntat, med tanke på att den globala strålningen ligger på lägre nivåer under vintermånaderna i Jämtlands län, där Jämtkraft AB är placerad. Sammanfattningsvis tyder resultaten på att den mest effektiva metoden verkar vara att skapa ett anpassat verktyg, där AFS kan vara till nytta för potentiella elnätsföretag genom att analysera historiska data och upptäcka avvikelser. / Electricity grid companies are facing a growing challenge to ensure that the electrical system can handle increasing demands in a time when society is becoming more electrified. In this context, long-term forecasting tools can play a crucial role in optimizing both the existing and future electricity grid. By using such tools, electricity grid companies can improve their network development plans, which is crucial for promoting an efficient and sustainable development of the electrical grid. By following well-established guidelines and best practices, it is possible to ensure accessibility, reliability, and cost-effectiveness in electricity supply. The purpose of the thesis is to explore what is required to manage an effective forecasting tool and meet the requirements of Regulation 7 § according to EiFS 2024:1 from the Swedish Energimarknadsinspektionen. The assignment arose as a result of this requirement, which states that distribution network companies must report their forecast for transmission capacity in megawatts to manage customers' energy consumption and production. The thesis focuses on developing methods and using the AFS software to achieve these goals, as well as analyzing current and future loads and parameters that affect the forecasts. A literature review has been conducted on various forecasting tools and methods available on the market. Simultaneously, five different electricity grid companies have been interviewed to investigate what forecasting tools they have developed and how they work to achieve the nätutvecklingsplanen. In addition to the literature review and interviews, the forecasting tool Aiolos Forecast Studio (AFS) has been analyzed to examine its potential for long-term forecasting. The findings from the literature review and interviews indicated that it takes more than just a single forecasting tool to conduct long-term forecasts. The literature review highlighted the presence of various tools and methods for performing these forecasts, but their suitability depends on the specific requirements or needs of the company. The interviews revealed that many individuals use or have used Excel for forecasting work, but additional tools needed to be developed to handle larger amounts of data as it might exceed Excel's capacity. Regarding the primary forecasting tool, it became evident that substantial preparatory work is required. Additionally, having a supplementary tool capable of applying the forecasts directly to the electricity grid is preferable. The analysis of AFS yielded mixed results. Since the assignment came from Jämtkraft Elnät AB, the data worked on was from their local grid. AFS is a valuable tool for short-term forecasts but can also be used for future scenarios. The different future scenarios worked on involved whether solar power production would have an increased peak capacity from 18 MW in 2023 to 62 MW in 2030 and 150 MW in 2035. Scalings were performed in the program to achieve the new desired peak capacities. The results showed that the forecasted peak capacities for 2030 and 2035 were 14% and 20% off from the desired peak value. This is because when working with a regression model, the forecast will scale relative to the historical outcome data, 18 MW. Following that, the scalings were analyzed for a year-long period, including a week of low consumption and a week of high consumption. This provided a clear insight into how solar power production influenced the load under different conditions and with different scalings. During the low-consumption week, occurring in June, the results clearly showed that at a scaling of 150 MW, the load shifted from consumption to production during certain hours, with clear signs of the so-called duck curve. During the high-consumption week, which took place in December, the results were barely noticeable. This was expected, considering that global radiation levels are lower during the winter months in Jämtlands län, where Jämtkraft AB is located. In summary, the results suggest that the most effective approach appears to be creating a customized tool, where AFS can be beneficial for potential grid companies by analyzing historical data and detecting anomalies.

We study the benefits of resource flexibility, considering two different flexibility structures. First, we want to understand the impact of the firm's pricing strategy on its resource investment decision, considering a partially flexible resource. Secondly, we study the benefits of a flexible resource strategic approach, considering a resource flexibility structure that has not been studied in the previous literature. First, we study the capacity investment decision faced by a firm that offers two products/services and that is a price-setter for both products/services. The products offered by the firm are of varying levels (complexities), such that the resources that can be used to produce the higher level product can also be used to produce the lower level one. Although the firm needs to make its capacity investment decision under high demand uncertainty, it can utilize this limited (downward) resource flexibility, in addition to pricing, to more effectively match its supply with demand. Sample applications include a service company, whose technicians are of different capabilities, such that a higher level technician can perform all tasks performed by a lower level technician; a firm that owns a main plant, satisfying both end-product and intermediate-product demand, and a subsidiary, satisfying the intermediate-product demand only. We formulate this decision problem as a two-stage stochastic programming problem with recourse, and characterize the structural properties of the firm's optimal resource investment strategy when resource flexibility and pricing flexibility are considered in the investment decision. We show that the firm's optimal resource investment strategy follows a threshold policy. This structure allows us to understand the impact of coordinated decision-making, when the resource flexibility is taken into account in the investment decision, on the firm's optimal investment strategy, and establish the conditions under which the firm invests in the flexible resource. We also study the impact of demand correlation on the firm's optimal resource investment strategy, and show that it may be optimal for the firm to invest in both flexible and dedicated resources when product demand patterns are perfectly positively correlated. Our results offer managerial principles and insights on the firm's optimal resource investment strategy as well as extend the newsvendor problem with pricing, by allowing for multiple resources (suppliers), multiple products, and resource pooling. Secondly, we study the benefits of a delayed decision making strategy under demand uncertainty, considering a system that satisfies two demand streams with two capacitated and flexible resources. Resource flexibility allows the firm to delay its resource allocation decision to a time when partial information on demands is obtained and demand uncertainty is reduced. We characterize the structure of the firm's optimal delayed resource allocation strategy. This characterization allows us to study how the revenue benefits of the delayed resource allocation strategy depend on demand and capacity parameters, and the length of the selling season. Our study shows that the revenue benefits of this strategy can be significant, especially when demand rates of the different types are close, while resource capacities are much different. Based on our analysis, we provide guidelines on the utilization of such strategies. Finally, we incorporate the uncertainty in demand parameters into our models and study the effectiveness of several delayed capacity allocation mechanisms that utilize the resource flexibility. In particular, we consider that demand forecasts are uncertain at the start of the selling season and are updated using a Bayesian framework as early demand figures are observed. We propose several heuristic capacity allocation policies that are easy to implement as well as a heuristic procedure that relies on a stochastic dynamic programming formulation and perform a numerical study. Our study determines the conditions under which each policy is effective. / Ph. D.