Artificial neural networks (ANN): prediction of sensory measurements from instrumental data

The objective of this study was to predict by means of Artificial Neural Network (ANN), multilayer perceptrons, the texture attributes of light cheesecurds perceived by trained judges based on instrumental texture measurements. Inputs to the network were the instrumental texture measurements of light cheesecurd (imitative and fundamental parameters). Output variables were the sensory attributes consistency and spreadability. Nine light cheesecurd formulations composed of different combinations of fat and water were evaluated. The measurements obtained by the instrumental and sensory analyses of these formulations constituted the data set used for training and validation of the network. Network training was performed using a back-propagation algorithm. The network architecture selected was composed of 8-3-9-2 neurons in its layers, which quickly and accurately predicted the sensory texture attributes studied, showing a high correlation between the predicted and experimental values for the validation data set and excellent generalization ability, with a validation RMSE of 0.0506.

Energy Management System for LVDC Island Networks

Recent developments in power electronics technology have made it possible to develop competitive and reliable low-voltage DC (LVDC) distribution networks. Further, islanded microgrids—isolated small-scale localized distribution networks— have been proposed to reliably supply power using distributed generations. However, islanded operations face many issues such as power quality, voltage regulation, network stability, and protection. In this thesis, an energy management system (EMS) that ensures efficient energy and power balancing and voltage regulation has been proposed for an LVDC island network utilizing solar panels for electricity production and lead-acid batteries for energy storage. The EMS uses the master/slave method with robust communication infrastructure to control the production, storage, and loads. The logical basis for the EMS operations has been established by proposing functionalities of the network components as well as by defining appropriate operation modes that encompass all situations. During loss-of-powersupply periods, load prioritizations and disconnections are employed to maintain the power supply to at least some loads. The proposed EMS ensures optimal energy balance in the network. A sizing method based on discrete-event simulations has also been proposed to obtain reliable capacities of the photovoltaic array and battery. In addition, an algorithm to determine the number of hours of electric power supply that can be guaranteed to the customers at any given location has been developed. The successful performances of all the proposed algorithms have been demonstrated by simulations.

A Study of Multicast Performance in IEEE 802.11 Wireless Local Area Networks

Langattomat lähiverkot ovat yleistyneet ja monin paikoin niillä pyritään täysin korvaamaan perinteiset kaapeloidut verkot. Samalla verkoissa välitettävät palvelut monipuolistuvat. Etenkin interaktiivisten sisältöjen ja viiveriippuvaisten palvelujen välittämisessä korostuvat verkolle asettavat vaatimukset suorituskyvystä ja palveluntasosta. Radiotaajuuksilla tapahtuva tiedonsiirto on ongelmallista siirtomedian jaetun luonteen vuoksi. Käytettäessä ympärisäteileviä antenneja jokainen kantomatkan päässä oleva asema kuulee lähetyksen, vaikkei se olisi sille suunnattu. Tämä luo turvallisuusongelman, mutta lisäksi heikentää tehokkuutta, koska nykyisellä antenniteknologialla vain yksi asema kerrallaan voi lähettää häiriöttä kantoalueellaan. Täsmälähetykset yhden lähteen ja useiden kohteiden välillä luovat erillisiä siirtolinkkejä. Olisi luontevampaa luoda yksi lähetysvirta, johon useat vastaanottajat voivat kytkeytyä. Kaistan säästön vastapainona on tarve luoda ja ylläpitää näitä yhteyksiä. Käyttäjien liikkuvuuden ja siirtotien vaihtelevien ominaisuuksien vuoksi langattoman verkon rakenne ja rajat ovat epäselviä. Näihin haasteisiin on vastattava sekä protokolla- että sovellustasolla. Tässä tutkielmassa käydään ensin läpi IEEE 802.11 -standardin mukaisen langattoman verkon perusteet ja sen ominaisuuksiin liittyvät tyypilliset ongelmakohdat. Yleis- ja ryhmälähetyksiä tarkastellaan ensin perinteisessä IEEE 802.3 standardin mukaisessa langallisessa lähiverkossa ja selvitetään, miten vastaavat ominaisuudet toteutuvat IEEE 802.11 standardien mukaisissa langattomissa verkoissa. Käyttötapaustutkimuksissa keskitytään rajatussa ympäristössä tapahtuviin ryhmälähetyksiin. Erityisesti tutkitaan langattomien ryhmälähetysten käyttökelpoisuutta sekä langattoman ympäristön asettamia erityisvaatimuksia.

Helsinki hub – new traffic flows and business networks

This thesis is the Logistics Development Forum's assignment and the work dealing with the development of the Port of Helsinki as part of Helsinki hub. The Forum aims to develop logistics efficiency through public-private co-operation and development of the port is clearly dependent on both factors. Freight volumes in the Port of Helsinki are the biggest single factor in hub and, therefore, the role of the port of the entire hub development is strong. The aim is to look at how the port will develop as a result of changes in the foreign trade of Finland and the Northern European logistics trends in 25 years time period. Work includes the current state analysis and scenario work. The analyses are intended to find out, which trends are the most important in the port volume development. The change and effect of trends is examined through scenarios based on current state. Based on the work, the structure of Finnish export industry and international demand are in the key role in the port volume development. There is significant difference between demands of Finnish exporting products in different export markets and the development between the markets has different impacts on the port volumes by mass and cargo type. On the other hand, the Finnish economy is stuck in a prolonged recession and competition between ports has become a significant factor in the individual port's volume development. Ecological valuesand regulations have changed the competitive landscape and maritime transport emissions reductions has become an important competitive factor for short routes in the Baltic Sea, such as in the link between Helsinki and Tallinn.

Relay protection in active distribution networks

Increasingly growing share of distributed generation in the whole electrical power system’s generating system is currently a worldwide tendency, driven by several factors, encircling mainly difficulties in refinement of megalopolises’ distribution networks and its maintenance; widening environmental concerns adding to both energy efficiency approaches and installation of renewable sources based generation, inherently distributed; increased power quality and reliability needs; progress in IT field, making implementable harmonization of needs and interests of different-energy-type generators and consumers. At this stage, the volume, formed by system-interconnected distributed generation facilities, have reached the level of causing broad impact toward system operation under emergency and post-emergency conditions in several EU countries, thus previously implementable approach of their preliminary tripping in case of a fault, preventing generating equipment damage and disoperation of relay protection and automation, is not applicable any more. Adding to the preceding, withstand capability and transient electromechanical stability of generating technologies, interconnecting in proximity of load nodes, enhanced significantly since the moment Low Voltage Ride-Through regulations, followed by techniques, were introduced in Grid Codes. Both aspects leads to relay protection and auto-reclosing operation in presence of distributed generation generally connected after grid planning and construction phases. This paper proposes solutions to the emerging need to ensure correct operation of the equipment in question with least possible grid refinements, distinctively for every type of distributed generation technology achieved its technical maturity to date and network’s protection. New generating technologies are equivalented from the perspective of representation in calculation of initial steady-state short-circuit current used to dimension current-sensing relay protection, and widely adopted short-circuit calculation practices, as IEC 60909 and VDE 0102. The phenomenon of unintentional islanding, influencing auto-reclosing, is addressed, and protection schemes used to eliminate an sustained island are listed and characterized by reliability and implementation related factors, whereas also forming a crucial aspect of realization of the proposed protection operation relieving measures.

Benefits of a low-cost one-phase current distortion detection device in appliance supply in low-voltage residential networks

In this bachelor’s thesis are examined the benefits of current distortion detection device application in customer premises low voltage networks. The purpose of this study was to find out if there are benefits for measuring current distortion in low-voltage residential networks. Concluding into who can benefit from measuring the power quality. The research focuses on benefits based on the standardization in Europe and United States of America. In this research, were also given examples of appliances in which current distortion detection device could be used. Along with possible illustration of user interface for the device. The research was conducted as an analysis of the benefits of current distortion detection device in residential low voltage networks. The research was based on literature review. The study was divided to three sections. The first explain the reasons for benefitting from usage of the device and the second portrays the low-cost device, which could detect one-phase current distortion, in theory. The last section discuss of the benefits of usage of current distortion detection device while focusing on the beneficiaries. Based on the result of this research, there are benefits from usage to the current distortion detection device. The main benefitting party of the current distortion detection device was found to be manufactures, as they are held responsible of limiting the current distortion on behalf of consumers. Manufactures could adjust equipment to respond better to the distortion by having access to on-going current distortion in network. The other benefitting party are system operators, who would better locate distortion issues in low-voltage residential network to start prevention of long-term problems caused by current distortion early on.

Novel cancer-related regulatory targets for the signalling sphingolipids sphingosine-1-phosphate and sphingosylphosphorylcholine

The signalling sphingolipid sphingosine-1-phosphate (S1P) is necessary for development of the immune system and vasculature and on a cellular level regulates migration, proliferation and survival. Due to these traits S1P has an important role in cancer biology. It is considered a primarily cancer-promoting factor and the enzyme which produces it, sphingosine kinase (SphK), is often over-expressed in tumours. S1P is naturally present in the blood, lymph, tissue fluids and cell cytoplasm and functions through its cell surface receptors (S1P1-5) and as an intracellular second messenger. Sphingosylphosphorylcholine (SPC) is closely related to S1P and has similar regulatory functions but has not been extensively studied. Both S1P and SPC are able to evoke either stimulatory or inhibitory effects on cancer cells depending on the context. The aim of this thesis work was to study novel regulatory targets of S1P and SPC, which mediate the effects of S1P/SPC signalling on cancer cell behaviour. The investigated targets are the transcription factor hypoxia-inducible factor 1 (HIF-1), the intermediate filament protein vimentin and components of the Hippo signalling pathway. HIF-1 has a central role in cancer biology, as it regulates a multitude of cancer-related genes and is potently activated by intratumoural hypoxia through stabilization of the regulatory subunit HIF-1α. Tumours typically harbour high HIF-1α levels and HIF-1, in turn, facilitates tumour angiogenesis and metastasis and regulates cancer cell metabolism. We found S1P to induce follicular thyroid cancer cell migration in normal oxygen conditions by increasing HIF-1α synthesis and stability and subsequently HIF-1 activity. Vimentin is a central regulator of cell motility and is also commonly over-expressed in cancers. Vimentin filaments form a cytoskeletal network in mesenchymal cells as well as epithelial cancer cells which have gone through epithelial-mesenchymal transition (EMT). Vimentin is heavily involved in cancer cell invasion and gives tumours metastatic potential. We saw both S1P and SPC induce phosphorylation of vimentin monomers and reorganization of the vimentin filament network in breast and anaplastic thyroid cancer cells. We also found vimentin to mediate the anti-migratory effect of S1P/SPC on these cells. The Hippo pathway is a novel signalling cascade which controls cancer-related processes such as cellular proliferation and survival in response to various extracellular signals. The core of the pathway consists of the transcriptional regulators YAP and TAZ, which activate predominantly cancer-promoting genes, and the tumour suppressive kinases Lats1 and Lats2 which inhibit YAP/TAZ. Increased YAP expression and activity has been reported for a wide variety of cancers. We found SPC to regulate Hippo signalling in breast cancer cells in a two-fold manner through effects on phosphorylation status, activity and/or expression of YAP and Lats2. In conclusion, this thesis reveals new details of the signalling function of S1P and SPC and regulation of the central oncogenic factors HIF-1 and vimentin as well as the novel cancer-related pathway Hippo.

Pretraining Convolutional Neural Networks for Visual Recognition

Convolutional Neural Networks (CNN) have become the state-of-the-art methods on many large scale visual recognition tasks. For a lot of practical applications, CNN architectures have a restrictive requirement: A huge amount of labeled data are needed for training. The idea of generative pretraining is to obtain initial weights of the network by training the network in a completely unsupervised way and then fine-tune the weights for the task at hand using supervised learning. In this thesis, a general introduction to Deep Neural Networks and algorithms are given and these methods are applied to classification tasks of handwritten digits and natural images for developing unsupervised feature learning. The goal of this thesis is to find out if the effect of pretraining is damped by recent practical advances in optimization and regularization of CNN. The experimental results show that pretraining is still a substantial regularizer, however, not a necessary step in training Convolutional Neural Networks with rectified activations. On handwritten digits, the proposed pretraining model achieved a classification accuracy comparable to the state-of-the-art methods.