Journal of Applied Sciences and Emerging Technologies

Optimization of Piezoelectric Energy Harvester with Line-Regulation

2024-10-01T07:55:18+00:00

Stair based Piezoelectric energy harvester (SBPH) is proposed to be placed in the footstep of stairs to harvest clean energy using the applied force. A system, independent of external power supply is developed which has the capability to charge a mobile phone. To reduce the loading effect, 90 Piezoelectric pressure sensors (PPS) in the SBPH are innovatively connected in parallel combination using Schottky diodes. A wooden art that consists of a network of 86 springs and 4 fasteners is used to concentrate the applied force on the anode of each PPS. A battery of 3.7 V at 3700 mAh can be charged through the SBPH. Moreover, the current system has the ability to efficiently utilize the harvested energy stored in the battery, using DC-DC converter with line regulation. Line regulation is achieved with TL494 Integrated chip (IC) as well as an Arduino UNO. The line regulation circuit based on Arduino has also a distinctive feature to disconnect the battery from the electronic circuit, when walking activity stops for a longer period of time. This system is cost effective and can be used in crowded places such as metro stations, airports, and so forth.

A Simulation of Free Space Propagation Model for Humidity Based Mediums

2025-02-11T07:17:11+00:00

The Digital Signal Quality and Strength are the typical parameters of concern when the communication medium of signals is changed from air to humidity based mediums. Signal attenuation has been observed in the cities where rain fall is common or the cities which are nearest to sea coastal areas. Signal Quality and Signal Strength have been affected when digital signals are propagated from the communication mediums where humidity factor is raised above than the normal figures. Optimization of temperature and humidity parameters from Free Space Propagation model will be beneficial for smart agriculture based monitoring system. We observed that there exist nonlinear relationship between Received Signals Power and Distance between sites whereas the Signal Strength decreases in nonlinear manner when the separation between sites increases.

SHORT TERM LOAD FORECASTING USING RECURRENT AND SPATIAL DEEP LEARNING MODEL FORECASTING

2025-03-18T19:53:09+00:00

Short-term load forecasting is critical to optimizing energy use and maintaining the stability of the power network. This study proposes an LSTM-based deep learning model for electricity demand forecasting, incorporating 21 input features, including time of day, day of week, and holiday factors within a 24-hour sequence to capture temporal consumption patterns. The model consists of two LSTM layers with 50 units each, applying dropout regularization and the Adam optimizer to improve training efficiency. When evaluated on the AEP dataset, the model achieves a MAPE of 0.65 percent, demonstrating its accuracy in handling nonstationary and seasonal variations. Comparative analysis with GRU, BiGRU, BiLSTM, and 1D CNN confirms that LSTM consistently achieves lower error rates. These findings highlight LSTM’s effectiveness in improving STLF accuracy, contributing to better energy planning and power network stability.

Hybrid CNN-LSTM Model for the Enhancement of Short-Term Load Forecasting

2025-03-21T01:06:16+00:00

Short-term load forecasting (STLF) is most important for the better management and operation of power systems. As energy consumption behaviors become increasingly complex and the incorporation of renewable energy sources grows, existing
techniques may not always provide higher accuracy in short-term load forecasting. To overcome this challenge, this paper proposes a novel approach. The proposed method is based on the convolutional neural network (CNN) and long short-term memory (LSTM) hybrid model. The CNN is used for local feature extraction, while LSTM manages the temporal dependencies within the time-series data. The model is evaluated using the American Electric Power (AEP) dataset, and its performance is contrasted with traditional methodologies. The CNN-LSTM-based hybrid model gives better performance
compared to single models, achieving a Mean Absolute Percentage Error (MAPE) of 0.61%, a Root Mean Squared Error (RMSE) of 120.91, and a Mean Absolute Error (MAE) of 90.16. These findings illustrate the model’s enhanced ability to forecast shortterm electrical load, surpassing numerous established techniques precisely.

A Review of Generative Adversarial Networks and its Applications

2025-02-11T07:21:58+00:00

The Generative Adversarial Networks are specific type of Artificial Intelligence Algorithms which are designed to solve problems related with Generative Modelling. Generative Models which are based upon Deep Learning are usually used when designing applications based on these models but Generative Adversarial Networks are very popular among the scientists who work on Computer Vision based algorithms because of their capability to produce high resolution videos and images. The Conventional Artificial Intelligence has been emerged into the new field Generative Artificial Intelligence and there is an indispensable need to investigate upon the novel algorithms which belong to the Generative Artificial Intelligence field, therefore Generative Adversarial Networks are one of these algorithms. The Generator and Discriminator are the principal components of Generative Adversarial Networks, these components are Neural Networks and both work in conjunction with each other. The output of the Generator is connected with the input of the Discriminator. The Discriminator plays an important role in distinguishing between the real data instances and fake data instances produced by the Generator. The Generator and Discriminator functions asynchronously with each other. When the Discriminator is in the phase of training, the Generator is not trained and the weights associated with data remain static during the training phase of the Discriminator. The Generator provides the training data to the Discriminator and completes its training on the acquired data from the Generator. The indispensable need of Generative Adversarial Networks today make it significant for researchers to study and apply them in the field of Computer Vision, Information Security Cyber Crimes Detection. From the review process it has been observed that Generative Adversarial Networks are very important and exciting innovative processes and algorithms which are now widely used in the field of Machine Learning and its applications.