Congestion in power systems primarily caused by deregulation of the power system and rising electricity consumption. Grid congestion occurrences are possible due to increased levels of distributed energy resources (DER) penetration and result in serious harm to power networks, including potential blackouts. Thus, one of the main objectives for assuring network dependability and security is the development of creative operation planning methodologies and control tools for grid congestion management. In order to reduce congestion and support Independent System Operators (ISOs), this work presents a Congestion Management (CM) technique utilizing Convolution Neural Network (CNN). Using a generation rescheduling-based strategy, the primary goal of CM is to ease transmission line congestion while addressing all limitations with least amount of congestion expense. The control of congestion in public transmission and distribution systems continues to be a difficult problem. Traditionally, load shedding and generator rescheduling have been employed to accomplish this. The suggested work uses Improved Artificial Bee Colony Algorithm (IABCA)-optimized CNN, which helps to deliver considerable results for congestion management with lower error. The suggested approach improves a number of current optimization techniques in terms of congestion cost and losses by producing more accurate prediction outputs. For the IEEE 30 Bus system, simulation outcomes based on IABCA are presented using MATLAB simulation software.
A real-time student attendance monitoring system using machine learning techniques is being developed to automate the manual attendance monitoring process in schools. The system will use image processing and computer vision algorithms to detect and recognize student's faces and compare them with a database of registered students. By providing real-time data and automating the monitoring process, the proposed system is expected to significantly reduce the workload for teachers and staff and improve the accuracy and consistency of attendance data. The process involves several steps, including the acquisition of images, improving their quality through pre-processing, identifying important features for recognizing students, and finally using machine learning techniques like MultiTask Cascaded Convolutional Neural Networks (MTCNNs) to accurately classify the students. Additionally, the system will provide comprehensive reporting to help identify patterns and trends in attendance data, which will provide valuable insights for school administrators.
In recent years, numerous swarm intelligence optimization algorithms have emerged for tackling diverse optimization problems. However, the determination of several hyperparameters remains a critical challenge. Take Particle Swarm Optimization (PSO) as an example: while it has demonstrated superior optimization performance in various applications, setting the weights for inertial velocity, the particle's best-known position, and the swarm's best-known position poses a significant hurdle. Hence, this study introduces an analytical framework aimed at examining the optimized mean fitness function value through mathematical models encompassing a range of fitness functions. Moreover, this framework facilitates the determination of optimized hyperparameter values associated with lower mean fitness function values. Experimental findings indicate that employing the proposed method for hyperparameter determination leads to enhanced convergence efficiency and decreased mean fitness function values.
Crop yield forecasting is a crucial component of agricultural research and development since it has a direct impact on food production and security. Due to the quick changes in soil and climatic conditions, the old techniques of estimating crop yields based on farmers' observations and experience have grown more difficult. In recent years, machine-learning approaches have been used to solve this problem. This study focuses on the use of several machine learning models for agricultural production prediction, including closest neighbor regression, nearest polynomial regression, random forest regression, gradient-boosted tree regression, and support vector regression. Effective feature selection methods were used to convert the raw data into a machine learning-friendly format. The results of this study demonstrated that, when compared to other strategies, the use of an ensemble approach in feature selection can produce more precise predictions. The ensemble technique creates a thorough and accurate forecast by fusing many datasources and machine learning algorithms. This study emphasizes the value of employing machine learning methods for predicting agricultural yields as well as the advantages of applying an ensemble approach to feature selection.
Sign language plays a critical role in facilitating communication among individuals with hearing or speaking disabilities. Until September 2021, signers from Kerala had to rely on Indian Sign Language (ISL), American Sign Language (ASL), or other regional sign languages to communicate. The introduction of Malayalam Sign Language (MSL) by National Institute of Speech Hearing, Thiruvananthapuram in September 2021 was a significant step towards addressing this issue. This paper presents a character recognition system specifically designed for MSL. The system focuses on recognizing nine Malayalam characters used in MSL communication. The proposed system utilises a deep learning approach, employing a modified Inception V4 model, for accurate image classification and character identification. The recognition system aims to address the communication barriers faced by individuals with hearing or speaking disabilities who rely on MSL for effective communication. The model utilises a modified Inception V4 model for image classification and letter identification, ensuring accurate recognition of Malayalam Sign Language symbols. This advancement significantly improves the communication experience for individuals who rely on sign language. Compared to other state-of-the-art methods, the proposed Inception V4 model performs better, achieving a Top-1 error rate of 17.7% and a Top-5 error rate of 3.8%.
