Search

Solar power plants are growing tremendously to manage the ever-growing demand for power production and supply in a sustainable manner. Solar irradiance forecasting, being a time series problem, can aid in the planning and design of solar power plants. This work have developed a deep learning forecast model for accurately predicting future values of solar irradiance for given location. The locations have been selected with the help of visual information, which indicates the intensity of solar irradiance. This visual information has been provided by NASA’s Prediction of Worldwide Energy Resources (POWER) project. The proposed model utilizes convolutional layers for extraction of internal representations of solar irradiance time-series data along with the attention-based LSTM network for the identification of temporal dependencies. The study is conducted on solar datasets from two locations for a period of 36 years taken from NASA’s Prediction of Worldwide Energy Resource (POWER) project archive of Renewable Energy. Experiments were conducted with comparisons against various deep learning models. From the study conducted, we analyze the effect of different factors, such as model, optimizer, horizon, to decide upon the final forecast result. It has been observed that the proposed model performed with an R2 score of more than 50 percent which indicate excellent forecast performance. Experimental analysis is also conducted for bias and variance as performance evaluators along with other indicators. The bias and variance components also show superior performance of the proposed model. The convolutional and the attention components enhance the performance of the LSTM model with the bias variance mechanism providing improved generalizability.

Recent wind speed forecasting works have applied various time series decomposition methods to segregate a complex wind time series into linear and nonlinear components. Then, existing statistical t...

The wind resources have been estimated by using physical models, statistical models, and artificial intelligence models. Wind power calculation helps us measure the annual energy that will sustain the balance between electricity generation and electricity consumption. Wind speed plays a significant role in calculating wind power, due to which here we focus on wind speed prediction. In this paper, hybrid models for wind speed forecasting have been proposed. The hybrid models are formed by combining the time series decomposition technique, that is, discrete wavelet transform (DWT), with statistical models, that is, autoregressive integrated moving average (ARIMA) and generalized autoregressive score (GAS), respectively. These hybrid models are referred to as DWT-ARIMA and DWT-GAS. DWT decomposes the original series into sub-series. After that, statistical models are applied to each sub-series for prediction. In the end, aggregate the prediction results of each sub-series to get the final forecasted series. For experimentation purposes, statistical and hybrid models are applied to various datasets that are taken from the NREL repository. In our studies, the hybrid version demonstrates better results in terms of accuracy and complexity, which indicates superior performance in most cases compared to the existing statistical models.

This article presents the Recurrent Neural Network (RNN) and its Attention mechanism to develop forecasting models for renewable energy applications. In this study, wind speed and solar irradiance forecasting models have been developed as these two factors play a significant role in renewable energy production. The irregular nature of wind poses the challenge of accurate wind speed prediction, while solar irradiance forecasting can aid in the planning and deployment of solar power plants. In this paper, six RNN techniques, namely RNN, GRU, LSTM, Content-based Attention, Luong Attention, and Self-Attention based RNN are considered for forecasting the future values of wind speed and solar irradiance in particular geographical locations. The aim is the identification of the advantages, comparison, and importance of different recurrent neural network methods for forecasting models. All models are developed on the datasets of the National Renewable Energy Laboratory (NREL) and NASA’s Prediction of Worldwide Energy Resource (POWER).

The access to sufficient amount of data has always challenged researchers to productively effectuate their solutions. One of the promising solutions can be generation of data using generative adversarial networks (GAN). This paper is focused on generating realistic data for different time series applications using GAN. The approach adopted here uses GRU-based GAN with conditional input for data generation. The data generated using GAN can contribute in the formation of larger datasets. The time component plays a major role in forecasting in various domains so it is crucial to target data related to time series. The competence of the data generated has been judged by using it to train the most prominent time series forecasting models and then testing it using real data. The linear regression model, ARIMA model, and GRU-based forecasting model are chosen for carrying out the experiment. The similarity between the actual data and generated data is also demonstrated using Wilcoxon signed-rank test as the datasets used here are nonparametric. The experimentation has been executed on three real world datasets from different domains.

Solar power is among one of the major renewable energy sources, and there are various applications depending on solar energy. These applications require continuous time series radiation data to access its usefulness, so missing values in solar dataset can have a significant effect on the performance of solar energy systems. This paper is focused on investigating a machine learning-based data imputation method using generative adversarial networks (GANs). Inspired by the success of GANs in other domains, it is being used for data imputation. The incomplete data adversely effects any time series study and forms hindrance in the path of advancements. The solar insolation missing values in the solar dataset are imputed in this work. The paper uses GAN with the concept of mask and hint matrix to ease the process of imputation. The work is demonstrated on the real-world solar dataset from three different regions. The competence of the model is judged by using the new imputed dataset for training different time series forecasting models. The performance of the model is also compared against frequently used imputation methods.

Optimization is a crucial part of every field that requires computation and plays an important role in modern optimization. Metaheuristics algorithms are stochastic algorithms designed in such a way that they don’t require the calculation of gradient for optimization. Hence they can be used for optimizing non-linear functions too. Since most real-world optimization problems are non-linear in nature, algorithms provide a sufficiently good solution (close to actual solution) for such optimization problems. This paper is aimed to explain the Genetic Algorithm(GA), Simulated Annealing(SA), and Particle Swarm Optimization(PSO) and compare their performance on benchmark functions. Benchmark functions provides a baseline measurement of the performance of algorithms and allows the algorithm to be compared to some other algorithms in the same manner. The algorithm runs on these functions with an aim to locate an optimal position. The performance and the results obtained can be affected by the parameter setting or operations inside each method. Based on the results, PSO performed better compared to GA and SA in obtaining best minimum fitness.

This paper analyses the recently proposed article “A feature reduced intrusion detection system using ANN classifier” by Akashdeep, Ishfaq Manzoor & Neeraj Kumar, (Expert systems with Applications, 2017) which has a limitation in its experimental setup. The work of Akashdeep et.al has crafted the test dataset to attain improved accuracy. They have utilized 5 fractional test datasets for performance evaluation. The reduced list of features obtained in their work does not give the asserted performance for the original test dataset. Table 18 of the above article by Akashdeep et.al gives the performance comparison of their work with existing works which isn’t appropriate as these works have different test dataset composition. Another issue with the work of Akashdeep et.al is the utilization of partial training dataset for determining the reduced list of features. Their work reduces the training dataset by random undersampling of the majority class instances and random replication of the minority class instances. The reduced list of features obtained by Akashdeep et.al comprises 25 features. This work applies the feature selection algorithm proposed by Akashdeep et.al on the original training dataset leading to a feature subset having 29 features. It has been observed experimentally that the reduced feature subset (29 features) obtained in later outperforms the former reduced feature set (25 features). This work uses the classification algorithms c4.5, Naive Bayes and Random Forest for performance comparison of these reduced feature sets. Oversampling one class may deteriorate the performance of another class. This work also evaluates random undersampling/oversampling of a specific class to design an optimal training dataset. The results show that the classification models developed using this training dataset have a better detection rate for the minority classes.

Time series Forecasting has attracted attention over the last decade with the boost in processing power, the amount of data available and the development of more advanced algorithms. It is now widely used in a range of different fields including Medical Diagnostics, Weather Forecasting, Financial time series etc. In this paper, we propose a model of attention mechanism that allows for attended input to be fed to the model instead of the actual input. The motivation for the model is to show a new way to view the input so that the model can make more accurate predictions. The proposed LSTM model with the attention mechanism is then evaluated on common evaluation metrics and the results are compared with state of art models like CNN-LSTM and Stacked LSTM to show its benefits.

This work was developed for forecasting wind speed data by using various statistical models, which can further be utilized for the estimation of the annual energy production of commercial wind farms. The pattern of historical data available for modeling may be linear or nonlinear. For linear time series pattern, Autoregressive Integrated Moving Average (ARIMA), ARIMAX model by using exogenous variable X and Vector Autoregressive (VAR) models have been developed. In order to achieve good performance on the nonlinear pattern, Generalized Autoregressive Score (GAS), GAS with exogenous variable (GASX) model has been developed. Wind time series data taken from different site locations of the National Renewable Energy Laboratory (NREL) repository have been utilized to develop the models. In our investigation, it has been found that the VAR model performs the best in most of the cases since it includes more than one variable for the development of the model.

Data preprocessing is an important step for designing classification model. Normalization is one of the preprocessing techniques used to handle the out-of-bounds attributes. This work develops 14 classification models using different learning algorithms for dynamic selection of normalization technique. This work extracts 12 data complexity measures for 48 datasets drawn from the KEEL dataset repository. Each of these datasets is normalized using min–max and z-score normalization technique. G-mean index is estimated for these normalized datasets using Gaussian Kernel Extreme Learning Machine (KELM) in order to determine the best-suited normalization technique. The data complexity measures along with the best-suited normalization technique are used as an input for developing the aforementioned dynamic models. These models predict the best suitable normalization technique based on the estimated data complexity measures of the dataset. The result shows that the model developed using Gaussian Kernel ELM (KELM) and Support Vector Machine (SVM) give promising results for most of the evaluated classification problems.

Estimation of wind power generation for grid interface helps in calculation of the annual energy production, which maintains the balance between electricity production and its consumption. For this purpose, accurate wind speed forecasting plays an important role. In this paper, linear statistical predictive models such as autoregressive integrated moving average (ARIMA), generalized autoregressive score (GAS) model and a GAS model with exogenous variable x (GASX) have been applied for accurate wind speed forecasting. Along with this, a non-linear statistical predictive modelling technique called non-linear GASX (NLGASX) has been proposed and applied to model non-linear time-series data. Furthermore, the proposed NLGASX model is optimized using modelling techniques based on neural networks, namely Sigmoid, TANH, Softmax and RELU. The proposed optimized NLGASX model performs far better as compared with other models. Wind speed is also used as an input to wind power curve model for predicting the wind power. According to the predicted wind power the annual energy has been calculated.

The ever-growing interest in and requirement for green energy have led to an increased focus on research related to forecasting solar irradiance recently. This study aims to develop forecast models based on deep learning (DL) methodologies and multiple-site data to predict the daily solar irradiance in two locations of India based on the daily solar radiation data obtained from NASA’s POWER project repository over 36 years (1983–2019). The forecast modeling of solar irradiance data is performed for extracting and learning the symmetry latent in data patterns and relationships by the machine learning models and utilizing it to predict future solar data. The goodness of fit and model performance are compared with rolling window evaluation using mean squared error, root-mean-square error and coefficient of determination (R2) for evaluation. The contributions of this study can be summarized as follows: (i) time series models based on deep learning methodologies were implemented to forecast the daily solar irradiance of two locations in India in consideration of the historical data collected by NASA; (ii) the models were developed on the basis of single-location univariate data as well as multiple-location data; (iii) the accuracy, performance and reliability of the model were investigated on the basis of standard performance evaluation metrics and rolling window evaluation; (iv) the feature importance of the nearby locations with respect to forecasting target location solar irradiance was analyzed and compared based on the solar irradiance data obtained from NASA over 36 years. The results indicate that the bidirectional long short-term memory (LSTM) and attention-based LSTM models can be used for forecasting daily solar irradiance data. According to the findings, the multiple-site data with solar irradiance historical data improve upon the forecast performance of single-location univariate solar data.

This study uses the kernalised version of Least Mean Square (LMS) algorithm which offers a technique of taking sample-by-sample update for reproducing kernel Hilbert spaces (RKHS), named as KLMS. This study uses online learning as it works in a dynamic mode where dependencies of variables change with time. Generally all statistical models build a model with historical data and make future predictions on the basis of developed model which is time-invariant. Also for a non-stationary process, it is not possible to capture the characteristics of relationship between input and output through a single curve. This paper presents an online method that builds a curve which evolves over the time. After the proposed approach is applied to the real world wind power system, the non-stationary behaviour of wind has been observed. The results of the simulation shows that the proposed technique can dynamically trace the time-varying nature of wind power system.

Technology in the music players is developing rapidly, especially in smart phones. Nowadays users have access to millions of songs available online. Selecting favorite music among these large archives is one of the biggest problem. Every user has his own taste of music selection. Selecting music depends on the surroundings and the mood of the user. New users and new items emerge every day, and the system has to react to them promptly. The problem of personalized music recommendation that takes different kinds of auxiliary information into consideration is resource constraint due to large amount of data involvement but these models provide much accurate results so more of these are being used for commercial purpose. The main aim of the recommendation system is to recommend songs such that it is closed to the user’s choice. As a comparative study, we will be analyzing the Graph-based Novelty Research On The Music Recommendation, Music Recommendation System Based on the Continuous Combination of Contextual Information, Smart-DJ: Context-aware Personalizing for Music Recommendation on Smart phones. These models are outlined to assist the users to find out the new music that is personalized. For the analysis purpose, we will be using data set provided by Douban Music.

The evolution of electromagnetic devices is occurring at an exponential rate and our dependency upon the electronic devices like satellites is increasing at an unprecedented rate as well. This means that if these devices fail, we have no idea of how we will restore them and the services they provide. One such force capable of destroying them is Coronal Mass Ejection. The cost of losing even one satellite is unthinkable and if a powerful CME hits directly, it will take down most of them. Thus NASA constantly monitors the sun's activities and tries to predict such occurrences so that orbits of satellites can be changed accordingly. We are using same data of the sun provided by NASA and presenting an online probabilistic model based on time series forecasting techniques to predict such event based on sun's current state. This paper discusses and performs a comparative study of DAR, VAR, Gaussian, Non-Gaussian and Beta models to present a single or group of models for this type of problem.

Adaptive filters play an important role in prediction. This ability of adaptive filters have been successfully used in prediction of wind-power generation. This paper focuses on the comparison between adaptive filtering algorithms in order to determine which filter produces least error for predicting wind-power generation. Algorithms such as Standard least mean square (SLMS), Normalized least mean square (NL-MS), Weighted least mean square (WLMS), Stochastic Gradient Descent least mean square (SGDLMS), Recursive least Square (RLS) are implemented. The performance of the filters is evaluated using actual operational power data of a wind farm in America. Four performance criteria are used in the study of these algorithms: Mean Absolute Error, R-squared value, Computational Complexity, and Stability of the system.

In the recent past it has been statistically proven that heart diseases are the largest contributing factor to the increase in mortality rate. Diagnosis of heart disease depends on a number of factors such as heart-rate, cholesterol levels, blood pressure, age and many other physiological factors. In most of the cases the diagnosis is done based on the experience and intuition of the doctor. This requires the doctor to be highly skilled and experienced in order to diagnose properly. It would be very helpful if we developed methods that could extract useful information from tonnes of clinical data. This paper reviews the L1/2 Logistic Regularization, Lasso, Elastic Net and Group Lasso regularization techniques, in order to select a subset of important features having the highest informative value in the prediction of heart diseases. Different regularization techniques select different subsets of features. This paper compares these techniques and finds out which is the most suitable method in the prediction of heart diseases.

The uncertainty in the time series data like wind speed, network traffic, stock price etc. makes the prediction of these data a very tedious task. In order to improve the performance of prediction, several models have been invented. In this paper, some of the models like autoregressive models and Holt-Winters have been discussed. Further, the various steps involved in obtaining the results and comparing the performance of above model have been examined.

this paper, a novel image retrieval technique based on the combination of Haar wavelet transformation and CIECAM02 color histogram (CH) have been proposed. In color based image retrieval, color histogram is one of the most repeatedly used image features and it is used at a great extent in content-based image retrieval (CBIR) systems as a significant color feature. The color histogram unchanged by translation and rotation. The local characteristics and texture features of an image are extracted by wavelet transformation. On conflation of wavelet transformation and color histogram new algorithm has been proposed. One may select a query image perceived to be similar to the visualized target image. A set of images similar to the query is then returned from the database. The final experimental results show that the proposed technique gives better performance than the other schemes, in terms of retrieval time.

With large number of documents on the web, there is a increasing need to be able to retrieve the best relevant document. There are different techniques through which we can retrieve most relevant document from the large corpus. Similarity between words, sentences, paragraphs and documents is an important component in various tasks such as information retrieval, document clustering, word-sense disambiguation, automatic essay scoring, short answer grading, machine translation and text summarization. Text similarity means user’s query text is matched with the document text and on the basis on this matching user retrieves the most relevant documents. Text similarity also plays an important role in the categorization of text as well as document. We can measure the similarity between sentences, words, paragraphs and documents to categorize them in an efficient way. On the basis of this categorization, we can retrieve the best relevant document corresponding to user’s query. This paper describes different types of similarity like lexical similarity, semantic similarity etc.

Enormous online information is available due to the World Wide Web. This needed efficient and accurate summarization systems to extract significant information. Text summarization system automatically generates a summary of a given document and helps people to make effective decisions in less time. In this paper two methods have been proposed for query-focused multi-document summarization that uses k-mean clustering, term-frequency and inversesentence-frequency method for sentence weighting to rank the sentences of the documents with respect to a given query. The proposed method finds the proximity of documents and query, and later uses this proximity to rank sentences of each document. It is assumed that the document which is nearer to a query might contain more meaning full sentences with respect to the information need expressed by user’s query further if a sentence contains rare query term than it is more informative than the sentences that contains frequent query term. Both methods first gives weights to documents according to their proximity and use these document weights to rank each of their sentences with tf-idf ranking function. A relative study for proposed methods has been done and experimental results shows that both methods are comparable because of a slight difference in performance. DUC 2007 test dataset and ROUGH-1.5.5 summarization evaluation package is used for evaluation purpose.

The traditional anti-spam techniques like Black and White List is not up to the mark in current scenario. The goal of Spam Classification is to distinguish between spam and legitimate mail message. But with the popularization of the Internet, it is challenging to develop spam filters that can effectively eliminate the increasing volumes of unwanted mails automatically before they enter a user's mailbox. Many researchers have been trying to separate spam from legitimate emails using machine learning algorithms based on statistical learning methods. In this paper, we evaluate the performance of Non Linear SVM based classifiers with various kernel functions over Enron Dataset.

The web log data embed much of the user’s browsing behavior and the operational data generated through Internet end user interaction may contain noise. Which affect the knowledge based decision. Handling these noisy data is a major challenge. Null value handling is an important noise handling technique in relational data base system. In this work the issues related to null value are discussed and null value handling concept based on train data set is applied to real MANIT web server log. A prototype system based on Fuzzy C-means clustering techniques with trained data set is also proposed in this work. The proposed method integrates advantages of fuzzy system and also introduces a new criterion, which enhances the estimated accuracy of the approximation. The comparisons between different methods for handling null values are depicted. The result shows the effectiveness of the methods empirically on realistic web logs and explores the accuracy, coverage and performance of the proposed Models.