This work has been funded by the European project CASBliP (STRP. FP6-2004-IST-4. Proposal/Contract no: 027063 FP6).
The power industry restructuring is undergoing in many countries around the world and asa result many electricity markets have been established. In the electricity market environment,profits of generation companies depend, to a large extent, on bidding strategies employed. Hence,how to develop optimal bidding strategies has become a major concern of generation companies. Inthis paper, a fuzzy set theory based method for building optimal bidding strategies is presented forgeneration companies participating in a recently established electricity market in which the widely-used step-wise bidding protocol is utilized and the available historical data is not sufficient. Takeninto account of the insufficient history data especially bidding data, bidding behaviors of rivalgeneration companies are modeled as fuzzy sets and a bidding strategy optimization model is thendeveloped. The well-known genetic algorithm is next employed to solve the bidding strategyoptimization problem. Finally, a simple numerical example with five generation suppliersparticipating in an electricity market is served for illustrating the essential features of the presentedmethod.
In this paper, we have studied the effect of fatigue on walking gait during normal walking. Acceleration of COG in lateral, vertical and anterior/posterior directions were recorded and analyzed to study the effect of the fatigue. Increase in cadence and shortening of step length were observed after the fatigue. The results also showed an increase in the RMS value of acceleration in lateral direction affecting the lateral stability during walking. The variability analysis showed that the amplitude variability is increased in lateral, vertical and anterior/posterior directions after fatigue which can be related to the weakness of the lower extremity muscles. Frequency analysis revealed that higher frequency components in the acceleration increased due to the fatigue. An increase in the wavelet entropy after fatigue showed the increase in the disorder in the vertical and anterior/posterior accelerations.
Answer Set Programming (ASP) is a powerful paradigm based on logic programming for non-monotonic reasoning. Current ASP implementations are restricted to “grounded range-restricted function-free normal programs” and use an evaluation strategy that is “bottom-up” (i.e., not goal-driven). Recent introduction of coinductive Logic Programming (co-LP) has allowed the development of top-down goal evaluation strategies for ASP. In this paper we present this novel goal-directed, top-down approach to executing predicate answer set programs with co-LP. Our method eliminates the need for grounding, allows functions, and effectively handles a large class of predicate answer set programs including possibly infinite ones.
Most current machine learning systems for medical decision support do not produce any indication of how reliable each of their predictions is. However, an indication of this kind is highly desirable especially in the medical field. This paper deals with this problem by applying a recently developed technique for assigning confidence measures to predictions, called conformal prediction, to the problem of acute abdominal pain diagnosis. The data used consist of a large number of hospital records of patients who suffered acute abdominal pain. Each record is described by 33 symptoms and is assigned to one of nine diagnostic groups. The proposed method is based on Neural Networks and for each patient it can produce either the most likely diagnosis together with an associated confidence measure, or the set of all possible diagnoses needed to satisfy a given level of confidence.
It is well established that quasi-brittle materials experience visco-elastic creep strain under sustained loads. The creep strain represents the non-instantaneous strain that occurs with time when the stress is sustained. Most of the existing creep prediction models could achieve relatively low accuracy level because of the creep dependency oil large number of parameters (e.g. relative humidity, stress level, age of loading). In addition, creep strain behavior is considered as a time-dependent visco-elastic property of masonry structures. This manuscript investigates the potential use of recurrent neural networks (RNN) for predicting creep of structural masonry. The main merit of using RNN is that RNN paradigm assembles the time-dependent process within its architecture during training. Thus, RNN becomes more capable of capturing time-dependent nonlinear relationships than the existing creep prediction model. Several network architectures are examined to enhance models' performance. The results showed that RNN architectures can reduce the creep prediction error by 30% when compared to feed-forward neural network models.
One solution to the problem of one-step-ahead prediction of the electricity power consumption in suburban areas is proposed. It is based on implementation of artificial neural networks (ANN) that are properly structured for prediction. A structure named Extended Feed Forward Accommodated for Prediction (EFFAP) ANN is introduced. A proper arrangement of training data is advised. After implementation, promising results were obtained giving, in most instances, predictions with an error less than 1%. In the worst situation observed, the discrepancy between the target and predicted values is 9% the most which we consider as acceptable. The method proposed implements ANNs that are generally widely known but creates new structure which is fully original. In comparison to existing solutions implemented on similar concepts our method is based on much smaller amount of measured data and/or exhibits incomparably simpler ANNs. That allows for its application in dynamic (on-line) forecasting system since unchanged network structure is used during time. In addition, it simplifies the initial solution creation for the training process that, again, enables automatization of the forecast. The method was integrated into a remote-power-reading, billing, control, and planning system. It is our opinion that the method proposed may be implemented with equal success to one-step-ahead prediction of broader class of time series exhibiting inherent quasi-periodical properties.
This paper presents a new approach for fault detection in power transmission line using support vector machine (SVM) technique. This method uses fault current samples for half cycle from the inception of fault. The line currents are applied as inputs to SVM for fault detection. The SVM is trained with linear, polynomial and Radial Basis Function (RBF) Kernel. The feasibility of the proposed method has been tested on a 110-kV, 21.687-km transmission line for all the ten types of fault using MATLAB Simulink. Upon testing on 640 fault cases with varying fault resistance, fault inception angle and fault distance the performance of the proposed method is quite promising. The proposed method is also tested with real time data recorded by digital fault recorder (DFR). The results encourage the use of this proposed method for detection of various faults in transmission line accurately.
The current statistical, machine learning, and data mining (data analysis) algorithms available are only suited for analyzing single datasets (studies) in isolation. The findings of the analysis and the knowledge produced (e.g,. important predicting variables and risk-factors, rules for prediction, etc.) are published and this knowledge is then manually synthesized within the human brain. In this paper we argue that this process can be further automated: it is possible to develop algorithms that integratively analyze (co-analyze) data from different studies with different characteristics and measured quantities (variables). To develop such algorithms, we further argue that modeling and inducing causal relations is necessary. We call this process Integrative Causal Analysis or INCA. We present three cases and corresponding existing or newly developed algorithms and techniques that illustrate the feasibility of integrative causal analysis and the key enabling ideas. Specifically, we discuss algorithms for learning causal relations from (a) data obtained over different experimental conditions, (b) data over different variable sets, and (c) data over semantically similar variables that nevertheless cannot be pulled together for various technical reasons. The latter case particularly, often occurs in the setting of analyzing multiple gene-expression datasets. The grand vision of Integrative Causal Analysis is to enable the automated or semi-automated, large-scale integration of a large part of the available data to construct causal models involving a significant part of human concepts.
In this paper, determine the fault type of failed power transformers with a few key gases with artficial neural network (ANN) using Levenberg-Marquardt algorithm is presented. Three Dissolved Gas in oil Analysis (DGA) criteria commonly used in industry was trained and tested with neural network Levenberg-Marquardt algorithm. Three key gases Methane (CH(4)), Ethylene (C(2)H(4)) and Acetylene (C(2)H(2)) were chosen for this study. Percentage of each gas used as inputs of ANN. The output is one of the fault types PD, D1, D2, T1, T2, T3. The results of this study are useful in development of a reliable transformer automated diagnostic system using artificial neural network. Multiple layer feedforward ANN is trained with Levenberg-Marquardt learning algorithm. This algorithm appears to be the fastest method for training moderate-sized feedforward neural networks. We determined best neural network topology and reached 100% diagnostic success.
Early failure detection in motor pumps is an important issue in prediction maintenance. An efficient condition-monitoring scheme is capable of providing warnings and predicting the faults at early stages. Usually, this task is executed by humans, but the logical progression of the condition-monitoring technologies is the automating the diagnosis process. To this end, intelligent diagnosis systems are used. Many researchers have explored artificial intelligence techniques to diagnose failures in general. However, all papers found in literature are related to a specific problem that can appear in many different machines. In real applications, when the expert analyzes a machine, not only one problem appears, but more than one problem may appear together. So, it is necessary to propose new methods to assist diagnosis, looking for a set of occurring faults. In this work, we describe methods to support motor pump failure diagnoses based on parametric net model and ANNs committees, and we propose methods to combine them. We describe a case study realized with a real dataset. The results obtained with these methods are encouraging.
A hybrid algorithm that incorporates two biologically inspired computational intelligence methods was used for the assessment of abdominal pain. Namely, Genetic Algorithms (GA) where used in search for the optimal subset of clinical diagnostic factors that can be given as inputs for Probabilistic Neural Networks (PNN) to perform medical diagnosis based on the clinical data. Thus, the implemented GA was a two-objective one. The first objective was to minimize the number of diagnostic factors that were considered for medical diagnosis. The second objective was to minimize the Mean Square Error of the constructed PNN at the testing phase. The obtained results of the proposed hybrid algorithm are related favorably to the corresponding ones derived by applying Receiver Operating Characteristic analysis. Eventually, it was found that a number up to 60% of the diagnostic factors that are recorded in patient's history may be omitted without any loss in clinical assessment validity, while, at the same time, the performance of the genetically pruned PNN is improved in terms of execution speed and prediction accuracy.
Social Tagging is the process by which many users add metadata in the form of keywords, to annotate and categorize items (posts, songs, pictures, web links, products etc.). Political blogs can recommend posts to users, based on tags they have in common with other similar users. However, a post in politics may be interpreted in a number of ways by different users. This is because terms, especially in politics, carry an ideological burden and therefore it is very likely for posts to present a semantic ambiguity. The significance of this study is that in contrast to current recommendation algorithms, we apply Higher Order Singular Value Decomposition (HOSVD) on a 3-dimensional tensor to find latent semantic relationships between the three types of entities that exist in a social blogging system: users, posts, and tags. We perform experimental comparison of the proposed method against state-of-the-art recommendation algorithms with two real data sets (Wordpress and Technorati). Our results show significant improvements in terms of effectiveness measured through recall/precision.
