Feedforward Multilayer Perceptron Models Research Articles

Application of Deep Neural Networks to Distribution System State Estimation and Forecasting

Classical neural networks such as feedforward multi-layer perceptron models (MLPs) are well established as universal approximators and as such, show promise in applications such as static state estimation in power transmission systems. The dynamic nature of distributed generation (i.e. solar and wind), vehicle to grid technology (V2G) and false data injection attacks (FDIAs), may pose significant challenges to the application of classical MLPs to state estimation (SE) and state forecasting (SF) in power distribution systems. This paper investigates the application of conventional neural networks (MLPs) and deep learning based models such as convolutional neural networks (CNNs) and long-short term networks (LSTMs) to mitigate the aforementioned challenges in power distribution systems. The ability of MLPs to perform regression to perform power system state estimation will be investigated. MLPs are considered based upon their promise to learn complex functional mapping between datasets with many features. CNNs and LSTMs are considered based upon their promise to perform time-series forecasting by learning the correlation of the dataset being predicted. The performance of MLPS, CNNs, and LSTMs to perform state estimation and state forecasting will be presented in terms of average root-mean square error (RMSE) and training execution time. An IEEE standard 34-bus test system is used to illustrate the proposed conventional neural network and deep learning methods and their effectiveness to perform power system state estimation and power system state forecasting.

ANN Modeling of Thermal Conductivity and Viscosity of MXene-Based Aqueous IoNanofluid

Research shows that due to enhanced properties IoNanofluids have the potential of being used as heat transfer fluids (HTFs). A significant amount of experimental work has been done to determine the thermophysical and rheological properties of IoNanofluids; however, the number of intelligent models is still limited. In this work, we have experimentally determined the thermal conductivity and viscosity of MXene-doped [MMIM][DMP] ionic liquid. The size of the MXene nanoflakes was determined to be less than 100 nm. The concentration was varied from 0.05 mass% to 0.2 mass%, whereas the temperature varied from 19 °C to 60 °C. The maximum thermal conductivity enhancement of 1.48 was achieved at 0.2 mass% and 30 °C temperature. For viscosity, the maximum relative viscosity of 1.145 was obtained at 0.2 mass% and 23 °C temperature. After the experimental data for thermal conductivity and viscosity were obtained, two multiple linear regression (MLR) models were developed. The MLR models’ performances were found to be poor, which further called for the development of more accurate models. Then two feedforward multilayer perceptron models were developed. The Levenberg–Marquardt algorithm was used to train the models. The optimum models had 4 and 10 neurons for thermal conductivity and viscosity model, respectively. The values of statistical indices showed the models to be well-fit models. Further, relative deviations values were also accessed for training data and testing data, which further showed the models to be well fit.

Traffic Forecasting and Planning of WiMAX under Multiple Priority Using Fuzzy Time Series Analysis

Network traffic prediction plays a fundamental role in characterizing the network performance and it is of significant interests in many network applications, such as admission control or network management. Therefore, The main idea behind this work, is the development of a WIMAX Traffic Forecasting System for predicting traffic time series based on the daily and monthly traffic data recorded (TRD) with association of feed forward multi-layer perceptron (FFMLP). The quality of forecasting WIMAX Traffic obtained by comparing different configurations of the FFMLP that consist of investigating different FFMLP model architectures and different Learning Algorithms. The decision of changing the FFMLP architecture is essentially based on prediction results to obtain the FFMLP model for flow traffic prediction model. The different configurations were tested using daily and monthly real traffic data recorded at each of the two base stations (A and B) that belongs to a Libyan WiMAX Network. We evaluate our approach with statistical measurement and a good statistic measure of FMLP indicates the performance of selected neural network configuration. The developed system indicates promising results in which it successfully network traffic prediction through daily and monthly traffic data recorded (TRD) association with artificial neural network.

Development of artificial neural network models based on experimental data of response surface methodology to establish the nutritional requirements of digestible lysine, methionine, and threonine in broiler chicks

An artificial neural network (ANN) approach was used to develop feed-forward multilayer perceptron models to estimate the nutritional requirements of digestible lysine (dLys), methionine (dMet), and threonine (dThr) in broiler chicks. Sixty data lines representing response of the broiler chicks during 3 to 16 d of age to dietary levels of dLys (0.88–1.32%), dMet (0.42–0.58%), and dThr (0.53–0.87%) were obtained from literature and used to train the networks. The prediction values of ANN were compared with those of response surface methodology to evaluate the fitness of these 2 methods. The models were tested using R2, mean absolute deviation, mean absolute percentage error, and absolute average deviation. The random search algorithm was used to optimize the developed ANN models to estimate the optimal values of dietary dLys, dMet, and dThr. The ANN models were used to assess the relative importance of each dietary input on the bird performance using sensitivity analysis. The statistical evaluations revealed the higher accuracy of ANN to predict the bird performance compared with response surface methodology models. The optimization results showed that the maximum BW gain may be obtained with dietary levels of 1.11, 0.51, and 0.78% of dLys, dMet, and dThr, respectively. Minimum feed conversion ratio may be achieved with dietary levels of 1.13, 0.54, 0.78% of dLys, dMet, and dThr, respectively. The sensitivity analysis on the models indicated that dietary Lys is the most important variable in the growth performance of the broiler chicks, followed by dietary Thr and Met. The results of this research revealed that the experimental data of a response-surface-methodology design could be successfully used to develop the well-designed ANN for pattern recognition of bird growth and optimization of nutritional requirements. The comparison between the 2 methods also showed that the statistical methods may have little effect on the ideal ratios of dMet and dThr to dLys in broiler chicks using multivariate optimization.

The use of Artificial Neural Networks in the modeling of socioeconomic category of Integrated Water Resources Management (Case study: Saf-Saf River Basin, North East of Algeria)

This research work has an objective to determine the effective variables in socioeconomic category of Integrated Water Resources Management for Saf-Saf river basin characterized by fast growing demand of urban and rural populations and the demand of economic sectors including industry and agriculture. In this paper, the artificial neural network models were used to model and predict the relationship between water resources mobilization and socioeconomic variables in the Saf-Saf river basin. The study area chosen is Saf-Saf river basin and real data were collected from 30 municipalities for reference year 2010. The results indicate that the feed-forward multilayer perceptron models with back-propagation are useful tools to define and prioritize the most effective variable on water resources mobilization and use. The model evaluation shows that the correlation coefficients are more than 94 % for training, verification, and testing data. The model aims to link the water resources mobilization and driving forces variables with the objective to strengthen the Integrated Water Resources Management approach.

Prediction of Inelastic Response Spectra Using Artificial Neural Networks

Several studies have been oriented to develop methodologies for estimating inelastic response of structures; however, the estimation of inelastic seismic response spectra requires complex analyses, in such a way that traditional methods can hardly get an acceptable error. In this paper, an Artificial Neural Network (ANN) model is presented as an alternative to estimate inelastic response spectra for earthquake ground motion records. The moment magnitude (MW), fault mechanism (FM), Joyner‐Boore distance (dJB), shear‐wave velocity (Vs30), fundamental period of the structure (T1), and the maximum ductility (μu) were selected as inputs of the ANN model. Fifty earthquake ground motions taken from the NGA database and recorded at sites with different types of soils are used during the training phase of the Feedforward Multilayer Perceptron model. The Backpropagation algorithm was selected to train the network. The ANN results present an acceptable concordance with the real seismic response spectra preserving the spectral shape between the actual and the estimated spectra.