Saleh M. Al-Alawi

An ANN-based approach for predicting global radiation in locations with no direct measurement instrumentation

In this work, a novel approach using an artificial neural network was used to develop a model for analyzing the relationship between the Global Radiation (GR) and climatological variables, and to predict GR for locations not covered by the models training data. The predicted global radiation values for the different locations (for different months) were then compared with the actual values. Results indicate that the model predicted the Global Radiation values with a good accuracy of approximately 93% and a mean absolute percentage error of 7.30. In addition, the model was also tested to predict GR values for the Seeb location over a 12 month period. The monthly predicted values of the ANN model compared to the actual GR values for Seeb produced an accuracy of 95% and a mean absolute percentage error of 5.43. Data for these locations were not included as part of ANN training data. Hence, these results demonstrate the generalization capability of this novel approach over unseen data and its ability to produce accurate estimates. Finally, this ANN-based model was also used to predict the global radiation values for Majees, a new location in north Oman.

Assessment and prediction of tropospheric ozone concentration levels using artificial neural networks

Abstract This work deals specifically with the use of a neural network for ozone modelling in the lower atmosphere. The development of a neural network model is presented to predict the tropospheric (surface or ground) ozone concentrations as a function of meteorological conditions and various air quality parameters. The development of the model was based on the realization that the prediction of ozone from a theoretical basis (i.e. detailed atmospheric diffusion model) is difficult. In contrast, neural networks are useful for modelling because of their ability to be trained using historical data and because of their capability for modelling highly non-linear relationships. The network was trained using summer meteorological and air quality data when the ozone concentrations are the highest. The data were collected from an urban atmosphere. The site was selected to represent a typical residential area with high traffic influences. Three neural network models were developed. The main emphasis of the first model has been placed on studying the factors that control the ozone concentrations during a 24-hour period (daylight and night hours were included). The second model was developed to study the factors that regulate the ozone concentrations during daylight hours at which higher concentrations of ozone were recorded. The third model was developed to predict daily maximum ozone levels. The predictions of the models were found to be consistent with observations. A partitioning method of the connection weights of the network was used to study the relative percent contribution of each of the input variables. The contribution of meteorology on the ozone concentration variation was found to fall within the range 33.15–40.64%. It was also found that nitrogen oxide, sulfur dioxide, relative humidity, non-methane hydrocarbon and nitrogen dioxide have the most effect on the predicted ozone concentrations. In addition, temperature played an important role while solar radiation had a lower effect than expected. The results of this study indicate that the artificial neural network (ANN) is a promising method for air pollution modelling.

Forecasting monthly electric load and energy for a fast growing utility using an artificial neural network

Abstract In this paper, novel artificial neural network (ANN) based weather-load and weather-energy models have been developed to forecast electric load and energy for 24 months ahead. A set of weather and other variables which have been identified for both models together with their correlations and contribution to the forecasted variable is reported. The proposed ANN models have been applied to historical energy, load, and weather data available for the Muscat power system from 1986 to 1990. Forecast results, when compared with the actual data for 1991–1992, show that monthly electric energy and load can be predicted within a maximum error of 6% and 10%, respectively, even with forecasted weather. The proposed ANN models provide better accuracy than previously developed models.

Patterns of SO2 emissions: a refinery case study

Abstract Air quality modelling is an essential tool for most air pollution studies and the introduction of SO 2 standards creates a need for modelling the dispersion of SO 2 . This work deals specifically with the use of the Industrial Source Complex Short Term (ISCST) model at a refinery. The study is performed over a period of 21 days. The first objective of this study was to measure the atmospheric levels of SO 2 and then to compare their values with the international standard limits. The second objective was to evaluate the ISCST model by comparing the calculated and measured concentrations. The third objective was to demonstrate the effect of wind regimes on the dispersion of SO 2 and to determine the spatial distribution of SO 2 over the modelled area. The results showed that the levels of SO 2 were well below the ambient air quality standard. Based on isopleths for SO 2 distribution in the study area (as output from the ISCST model), it can be stated that no health risk is present in areas adjacent to the refinery.

Tuning a static var compensator controller over a wide range of load models using an artificial neural network

Abstract A novel approach using an artificial neural network (ANN) for tuning a static var compensator (SVC) controller over a wide range of load models is presented in this paper. To enhance power system damping over a wide range of load models, it is desirable to adapt the SVC controller gain in real time based on load models. To do this, online measurements of load parameters which are representative of load models are chosen as the input signals to the neural network. The output of the neural network is the desired gain of the SVC controller. The neural network, once trained by a set of input-output patterns in the training set, can yield a proper SVC controller gain under any load model. Simulation results show that the tuning gain of a SVC controller using the ANN approach can provide better damping of the power system over a wide range of load models than the fixed-gain controller.

An advanced ANN model for predicting the rotational behaviour of semi-rigid composite joints in fire using the back-propagation paradigm

This paper describes an artificial neural networking (ANN) model developed to predict the behaviour of semi-rigid composite joints at elevated temperature. Three different semi-rigid composite joints were selected, two flexible end-plates and one flush end-plate. Seventeen different parameters were selected as input parameters representing the geometrical and mechanical properties of the joints as well as the joint’s temperature and the applied loading, and used to model the rotational capacity of the joints with increasing temperatures. Data from experimental fire tests were used for training and testing the ANN model. Results from nine experimental fire tests were evaluated with a total of 280 experimental cases. The results showed that the R2 value for the training and testing sets were 0.998 and 0.97, respectively. This indicates that results from the ANN model compared well with the experimental results demonstrating the capability of the ANN simulation techniques in predicting the behaviour of semi-rigid composite joints in fire. The described model can be modified to study other important parameters that can have considerable effect on the behaviour of joints at elevated temperatures such as temperature gradient, axial restraints, etc.

ESTABLISHING PVT CORRELATIONS FOR OMANI OILS

An adequate knowledge of any reservoir fluid PVT properties is essential for most types of petroleum calculations. These calculations include amount of oil in the reservoir, production capacity, variations in produced gas-oil ratio during the reservoirs production life, calculation of recovery efficiency, reservoir performance, production operations and the design of production facilities. PVT properties can be measured experimentally by using collected bottom-hole or surface samples of crude oils. But, the experimental determination of PVT is time consuming and very costly. In addition, even with the availability of PVT analyses, it is often necessary to extrapolate the data to field and/or surface conditions through the use of empirical correlations. Furthermore, geological and geographical conditions are considered very critical in the development of any correlation. But, universal correlations are difficult to develop. That is why correlations for local regions, where crude properties are expected to be uniform, is a reasonable alternative. In this study, experimental PVT data for North and South Oman crudes, statistical and artificial neural network (ANN) analyses are used to develop reliable PVT correlations. Comparisons with previously published correlations are presented.

An Intelligent Transformer Load Estimation Model Using Artificial Neural Networks

Abstract This paper discusses the development of an intelligent system to estimate the transformer peak load from the asset management point of view in a large electric utility. This utility has over 50,000 distribution transformers, some of these transformers are approaching their design life and therefore it is important that their loading levels are continuously monitored. From the resource point of view, however, it is not cost effective to monitor individual transformer peak loading. It is therefore very much desirable to develop a system to estimate the peak loads on individual transformers from available data. The utility recently conducted a series of power quality tests on a number of transformers and from these data, a database was developed containing hourly average kVA demand and other related information. The measurements spanned over a variety of load centres including residential, commercial, industrial, and combinations thereof. An intelligent model is developed based on Artificial Neural Networks (ANNs) to estimate hourly average loads on distribution transformers from these data. In this paper, results from the development of a predominantly residential load model is reported. The ANN model was developed based on 5818 cases of residential transformer loads taken from the available database. The model was constructed to predict the load using 12 input variables that deemed important in determining the transformer hourly loading. The peak load was extracted from the hourly average loads. Once the ANN model’s training was completed, the resulting ANN was evaluated on 318 transformers selected randomly that was not used in the training set. Results indicate that the ANN model showed good agreement with actual loads. The R2 value for the training set was 0.855 and for the testing set was 0.854 respectively. Using contribution analysis, it was found that the highest contributions were from kVA ratings (13.8%), LV length (13.2%), HV rating (10.1%) and hour of the day (10.1%). While the lowest contributions were number of transformers connected (6.7%) on a feeder, temperature (5.8%), humidity (5.5%) and wind speed (2.2%).

MATRIX AND CEMENT EFFECTS ON RESIDUAL OIL SATURATION IN SANDSTONE FORMATIONS: A NEURAL NETWORK APPROACH

Abstract Residual oil in sandstone is affected by mineral composition, clay matrix and cementing material. Matrix minerals affect the affinity of a fluid to spread on a rock surface significantly and in turn controls the fluid distribution within the pore spaces. At the interface between the rock surface and the contacting fluid, electrical charges are in the origin of the extent of phase wetness. Available framework grains, a dominant component of rock matrix, affect porosity and, hence, amounts of rock preferential wetness. Cement, clay matrix and quartz overgrowth, which make up for the rest of the grain population in a rock, influence wetness and, therefore, amounts of residual oil. In this paper, spectro-electromicroscopy (SEM) point-count technique in conjunction with neural network analysis were used to determine the effect of certains rock parameters on the amounts of residual oil following waterflooding operations. Using artifcial neural network, the intent was then to predict the extent of resid...