A Novel Neural Network Structure for Electricity Load Prediction

Published: Aug. 1, 2023

Electricity load prediction is a critical task in the energy industry, as it helps to ensure stable and reliable power supply. Accurate forecasting enables companies optimize their operations, reduce costs, improve overall efficiency. However, traditional methods for prediction, such statistical models machine learning techniques, are often limited by accuracy computational This article proposes Temporal Fusion Transformer Modified structure electricity using Australian Energy Market Operator (AEMO) dataset. The proposed model outperforms LSTM techniques achieves improved experimental results demonstrate effectiveness of approach its potential practical applications industry.

Language: Английский

---

Radian Scaling: A Novel Approach to Preventing Concept Drift in Electricity Load Prediction

2022 IEEE International Conference on Consumer Electronics-Asia (ICCE-Asia), Journal Year: 2023, Volume and Issue: unknown

Published: Oct. 23, 2023

Artificial neural networks often suffer from concept drift when predicting the electricity load in smart cities due to seasonal trends. The most commonly used methods solve this issue are automatically updating model, using feature engineering capture seasonality trends, and ensemble methods, making model complex for someone replicate. Furthermore, as is mostly shifted minimum maximum values, we need a method normalize without being bound by training data's scale, mean, median. We proposed radian scaling angles between two consecutive values artificial process. Our prediction result shows high accuracy with root-mean-square error score of 0.0021 0.0036 many-to-one many-to-many forecasting scenarios, surpassing existing modification.

Language: Английский

---

An Epileptic Seizure Detection Method Based on TCN-LSTM

Published: Oct. 20, 2023

Due to individual variability of electroencephalogram (EEG) data and a severe imbalance in the proportion acquired detection samples, development automatic epilepsy technology has been constrained. This paper presents an method based on Temporal Convolutional Network-Long Short Term Memory (TCN-LSTM). First, raw EEG signals are subjected initial filtering balancing pre-processing. Subsequently, TCN-LSTM network, known for its excellent handling time-series data, is used feature extraction classification signals. In addition, Squeeze-and-Excitation Networks (SENet) channel attention mechanism incorporated, which assigns different weights extracted features leads, emphasizing channels with higher weights, thereby enhancing efficiency. approach avoids complex process manual extraction. Experimental testing dataset from Boston Children's Hospital yielded results 84.18% sensitivity, 92.84% specificity, 92.86% accuracy. The proposed model suitable automated system.

Language: Английский

---

A Design of an Energy Regulation Network Framework of Central Air Conditioning System Based on LSTM Prediction Model

Published: Aug. 18, 2023

With the development of economy, more buildings are established for people to live in. To save power consumed by central air condition systems (CACS) in and make CACS control easily, network technology is employed this field which greatly improve level system CACS. Nevertheless, when complex such as Cloud–Edge–Device architecture used, it difficult be implemented practical engineering. solve problem, a novel Network framework proposed article, an energy management adjustment strategy based on Short Term Memory networks (LSTM) model used predict load adjust system. The specific work follows: firstly, Then, framework, LSTM prediction value obtained model. Finally, prove that accuracy above affected its structure parameters, two models with different parameters respectively same group samples. experiment results may Therefore, promote application

Language: Английский

---

A Hybrid Power Load Forecasting Framework with Attention-Based Network and Multi-Scale Decomposition

Published: Jan. 1, 2024

Language: Английский

---