Epilepsy remains one of the most common chronic neurological disorders; hence, there is a need to further investigate various models for automatic detection of seizure activity. An effective detection model can be achieved by minimizing the complexity of the model in terms of trainable parameters while still maintaining high accuracy. One way to achieve this is to select the minimum possible number of features. In this paper, we propose a long short-term memory (LSTM) network for the classification of epileptic EEG signals. Discrete wavelet transform (DWT) is employed to remove noise and extract 20 eigenvalue features. The optimal features were then identified using correlation and P value analysis. The proposed method significantly reduces the number of trainable LSTM parameters required to attain high accuracy. Finally, our model outperforms other proposed frameworks, including popular classifiers such as logistic regression (LR), support vector machine (SVM), K-nearest neighbor (K-NN) and decision tree (DT).

With the progress of renewable energy generation and energy storage technologies, more and more renewable sources and devices are integrated into the power system. Due to the complexity of the power system, single and multiple power quality disturbances (PQDs) occur more frequently. Hence, real‐time detection of PQDs is the primary issue to mitigate the risk of distortions. This study presents the real‐time PQDs classification using fused convolutional neural networks (CNN) combined with long short‐term memory (fused CNN‐LSTM) architecture based on time and frequency domain features. The frequency‐domain features were obtained from time‐series data using fast Fourier transform. The original time‐domain and frequency‐domain features are extracted by respective CNN‐LSTM structures. The extracted time and frequency domain features are concatenated to classify the PQD through fully connected layers. Our proposed method was trained and tested using 16 types of synthetic noise PQDs data generated by mathematical models, in accordance with the standard IEEE‐1159. Moreover, to further verify the performance of our approach, a simulation distributed power system is carried out to detect various PQDs. We compared three advanced neural network approaches: Deep CNN, CNN‐LSTM, and multifusion CNN (MFCNN). The fused CNN‐LSTM model takes only 0.64 ms to classify each PQDs signal and achieves an accuracy of 98.95% and 98.89% in synthetic data and simulated power systems which indicates our proposed method outperformed compared methods.

The internet-of-Vehicle (IoV) can facilitate seamless connectivity between connected vehicles (CV), autonomous vehicles (AV), and other IoV entities. Intrusion Detection Systems (IDSs) for IoV networks can rely on machine learning (ML) to protect the in-vehicle network from cyber-attacks. Blockchain-based Federated Forests (BFFs) could be used to train ML models based on data from IoV entities while protecting the confidentiality of the data and reducing the risks of tampering with the data. However, ML models are still vulnerable to evasion, poisoning and exploratory attacks by adversarial examples. The BFF-IDS offers partial defence against poisoning but has no measure for evasion attacks, the most common attack/threat faced by ML models. Besides, the impact of adversarial examples transferability in CAN IDS has largely remained untested. This paper investigates the impact of various possible adversarial examples on the BFF-IDS. We also investigated the statistical adversarial detector’s effectiveness and resilience in detecting the attacks and subsequent countermeasures by augmenting the model with detected samples. Our investigation results established that BFF-IDS is very vulnerable to adversarial examples attacks. The statistical adversarial detector and the subsequent BFF-IDS augmentation (BFF-IDS(AUG)) provide an effective mechanism against the adversarial examples. Consequently, integrating the statistical adversarial detector and the subsequent BFF-IDS augmentation with the detected adversarial samples provides a sustainable security framework against adversarial examples and other unknown attacks.

Energy consumption in buildings contributes to over a third of global energy consumption and 28% of greenhouse gas emissions. With urbanization and population growth, rising building energy demand can lead to environmental degradation. While significant renewable resources are used to generate electricity to mitigate environmental problems, demand-side management remains crucial for achieving net-zero emissions and enhancing energy efficiency. Accurate building load forecasting is pivotal in devising optimal demand response schemes to shift or reduce the demand on power grids. Recent studies have achieved progressive breakthroughs in building energy forecasting through machine learning algorithms. However, most studies focused on building-level energy forecasting rather than individual load forecasting, which cannot support controlled demand response programs. In this study, we propose a multi-task learning model incorporating Patch, Temporal Convolutional Network and Time-Series Transformer (PatchTCN-TST) based on the channel-independent strategy for floor-level multiple electricity loads and indoor environmental forecasting. The PatchTCN-TST model is implemented to predict future data ranging from one-step ahead to three-step ahead on a real-world office building in Bangkok, Thailand. The experiment results indicate that the prediction performance of our model outperforms the prevalent methods, including LSTM, GRU, TCN, Transformer, Informer and Autoformer. The PatchTCN-TST model demonstrates superior accuracy in three forecasting scenarios, significantly reducing MAE, MSE, RMSE, and aSMAPE by 34%, 23%, 12%, and 36.4%, respectively, compared to the best baseline model.

Energy demand has grown explosively in recent years, leading to increased attention of energy efficiency (EE) research. Demand response (DR) programs were designed to help power management entities meet energy balance and change end-user electricity usage. Advanced real-time meters (RTM) collect a large amount of fine-granular electric consumption data, which contain valuable information. Understanding the energy consumption patterns for different end users can support demand side management (DSM). This study proposed clustering algorithms to segment consumers and obtain the representative load patterns based on diurnal load profiles. First, the proposed method uses discrete wavelet transform (DWT) to extract features from daily electricity consumption data. Second, the extracted features are reconstructed using a statistical method, combined with Pearson’s correlation coefficient and principal component analysis (PCA) for dimensionality reduction. Lastly, three clustering algorithms are employed to segment daily load curves and select the most appropriate algorithm. We experimented our method on the Manhattan dataset and the results indicated that clustering algorithms, combined with discrete wavelet transform, improve the clustering performance. Additionally, we discussed the clustering result and load pattern analysis of the dataset with respect to the electricity pattern.

The accurate prediction of photovoltaic (PV) power generation is crucial for improving virtual power plant (VPP) efficiency and power system stability. However, short-term PV power forecasting remains highly challenging due to the significant impact of weather changes, especially the complexity of cloud motion. To this end, this paper proposes an end-to-end innovative deep learning framework for data fusion based on multimodal learning, which utilizes a new auto cross modal correlation attention (ACMCA) mechanism designed in this paper for feature extraction and fusion by combining historical PV power generation time-series data and sky image data, thereby enhancing the model’s prediction performance under complex weather conditions. In this paper, the effectiveness of the proposed model was verified through a large number of experiments, and the experimental results showed that the model’s forecast skill (FS) reached 24.2% under all weather conditions 15 min in advance, and 24.32% under cloudy conditions with the largest fluctuations. This paper also compared the model with a variety of existing unimodal and multimodal models, respectively. The experimental results showed that the model in this paper outperformed other benchmark methods in all indices under different weather conditions, demonstrating stronger adaptability and robustness.