Aiming to address the problems of interest conflict between charging stations and electric vehicle (EV) owners, as well as severe load fluctuations caused by disorderly EV charging, this paper proposes a multi-objective optimal scheduling model based on an improved NSGA-III algorithm (TSM-NSGA-III). The model utilizes dynamic electricity price as a decision variable instead of a fixed time-of-use price, with optimization objectives set to maximize charging station profit, maximize EV owner satisfaction, and minimize the load peak-valley difference rate. The TSM-NSGA-III algorithm enhances the original NSGA-III through three key improvements: (1) chaotic reverse learning to improve initial population quality, (2) the sparrow search algorithm to avoid local optima, and (3) Manhattan distance to preserve population diversity and discover potential optimal solutions. Experimental results demonstrate that the proposed method achieves a 26% faster convergence and a 9.9% higher average solution quality compared to NSGA-III. Furthermore, it obtains superior Pareto frontiers with significantly better performance in both charging station revenue and user satisfaction, effectively overcoming the algorithm’s tendencies toward premature convergence and neglect of diverse optimal solutions.

In light of the issues related to the low detection accuracy and inefficiency of existing anomaly detection methods when processing data from photovoltaic power plants, this paper presents an enhanced anomaly detection method for photovoltaic power plants using an improved isolation Forest algorithm. Firstly, the pearson feature selector is employed to choose high-correlation features as model inputs. Secondly, data is segmented through non-axis parallel cutting, and a cutting standard function is designed to select the optimal hyperplane. Subsequently, parallel training is conducted to construct multiple isolated trees, forming an isolated Forest. Finally, the isolated density method is utilized for data anomaly assessment, addressing the problem of overlooking local anomalies in anomaly detection. The improved isolation Forest algorithm is compared with eight other anomaly detection methods across seven datasets. After experimental analysis, this approach effectively enhances the accuracy and efficiency of photovoltaic data anomaly detection. The detection accuracy value in the photovoltaic dataset is found to be 10.36

To address the issues of missing activity transition interval features and neglected temporal correlations between actions in existing WiFi-CSI continuous activity recognition, this paper proposes the Wi-ContiAct model. First, Noise Immunity Ratio (NIR) and Activity Specificity Response (ASR) are utilized to select highly sensitive subcarriers, followed by effective activity interval extraction through a non-activity baseline. Second, the detection granularity is extended from isolated transition points to continuous transition intervals by integrating the Dynamic Collaborative Response Entropy Rate (DCRER) with a three-state Hidden Markov Model (HMM). Finally, a multimodal temporal fusion recognition model is established, integrating activity interval features, transition interval features, and temporal correlations, with recognition results corrected by HMM state transition priors. Experimental results demonstrate that the model achieves a segmentation F1-score of 96.7% and recognition accuracy of 96.53% in scenarios involving five or more consecutive actions.

Aiming at the problems of the Informer model, where information is easily lost during the distillation process and training convergence efficiency is low, and the KDE algorithm, where prediction intervals are inaccurate due to fixed bandwidth and the tendency to ignore the temporal correlation of adjacent days’ output, this paper proposes a photovoltaic power probability interval prediction model based on PInformer-HKDE. Firstly, sparse entropy attention is introduced to improve the sparse attention mechanism in the distillation process of the Informer model, solving the issue of important information loss. Secondly, a novel decay weight loss function considering time series patterns is adopted to better capture the dependency relationships between sequences in the Informer model and improve convergence efficiency. Finally, an improved probability density function is employed in KDE, and a dynamic time window mechanism is added to accurately quantify the uncertainty prediction intervals. Experimental results show that the uncertainty prediction intervals generated by this method at different confidence levels exhibit high reliability and good sharpness.

Traditional e-commerce review methods face difficulties in identifying implicit evaluation targets and extracting features from both local details and overall semantic levels. To address this, this paper proposes an aspect term extraction model integrating auxiliary sentence construction and a hybrid neural network (EBERT-CBAC). This model first constructs auxiliary sentences to mine potential aspect term information and incorporates part-of-speech (POS) tags, enabling the traditional BERT model to capture multi-dimensional POS word structures. After obtaining contextual embeddings, it uses a CNN layer to extract local contextual features, followed by a BiGRU layer to capture global contextual features within the sentence. Considering that a single attention mechanism cannot capture important sentence features from different perspectives, a multi-head attention mechanism is introduced to process sentence-level information. Finally, sequence labeling is completed in the CRF layer to accurately identify both explicitly expressed and semantically implicit aspect terms in reviews. Experimental results show that, compared to the baseline model BERT-BiLSTM-CRF, the proposed model achieves an average F1-score improvement of 2.48

To address the challenges of feature extraction difficulty and limited diagnostic accuracy caused by the non-stationary characteristics of gearbox signals under variable-speed conditions, this paper constructs a fault diagnosis model based on Adaptive Wavelet Packet Transform and Bidirectional Gated Mamba (AWPT-BGMNet). The model replaces the max-pooling layer in QCNN with an adaptive wavelet packet transform pooling module, which dynamically decomposes high-frequency and low-frequency components of the signal, achieves decoupling of rotational speed interference and fault features, suppresses noise, and effectively overcomes the failure of fixed frequency-band decomposition under variable-speed conditions. Meanwhile, to address the problem that the conventional Mamba cannot utilize future temporal information, resulting in incomplete feature extraction, a bidirectional gated Mamba module is designed. By performing forward and backward temporal modeling, it realizes global contextual perception of the entire fault evolution process, significantly enhancing fault feature extraction and improving diagnostic accuracy. Experimental results show that the proposed method outperforms comparative models in multiple tasks, achieving average diagnostic accuracies of 98.62% and 96.98%, respectively, verifying its high accuracy and good generalization ability in complex variable-speed scenarios.

Existing intrusion detection methods based on amplitude statistics are susceptible to sudden environmental noise interference, while traditional spectral analysis approaches suffer from insufficient resolution for detecting subtle movements in single-antenna links. To this end, this paper proposes Wi-PID, a real-time lightweight detection framework based on time-frequency feature fusion. This is achieved by constructing a heterogeneous time-frequency feature space: introducing the scale-invariant energy fluctuation rate (TEFR) in the time domain to suppress non-stationary environmental noise interference by leveraging the relative variation characteristics of signal amplitude; In the frequency domain, it designs the spectral centroid change rate (SCCR), capturing the dynamic drift of Doppler shifts to significantly enhance the system’s perception accuracy for low-speed intrusions. Furthermore, a dynamic thresholding strategy driven by a hybrid statistical-extremal boundary is proposed, combined with an environment-adaptive weighting mechanism to achieve a dynamic equilibrium between detection sensitivity and false alarm rate. Experiments conducted on low-cost commercial ESP32 hardware demonstrate that Wi-PID exhibits exceptional robustness in complex scenarios such as corridors and conference rooms, achieving an overall accuracy exceeding 95%. This effectively resolves the challenge of high-precision intrusion detection within resource-constrained environments.

To address the limitations of traditional Fresnel zone-based indoor localization methods in terms of hardware offset sensitivity and multipath interference. This paper proposes an indoor localization model based on Fresnel clustering fusion, an indoor localization model based on Fresnel clustering fusion is proposed, which consists of two key stages: Firstly, A phase calibration model based on Differential Semidefinite Programming (SDP) has been developed to effectively avoid local optima during nonlinear offset correction and achieve globally optimal results. Additionally, a residual attention autoencoder was integrated to enhance the reflection characteristics of high-frequency targets and reduce multipath interference. Secondly, a localization method that integrates Fresnel zone analysis based on clustering fusion is proposed, which optimizes the feature space and addresses the issue of boundary ambiguity through a KL divergence constraint to achieve accurate target localization. Furthermore, the clustering centers are mapped to the intersection points of the Fresnel ellipse for precise localization targets. Experimental results demonstrate that the proposed method achieves a localization accuracy of up to 0.7 m.

Conventional methods for photovoltaic power uncertainty quantification often produce unreliable prediction intervals in complex scenarios like extreme weather. This limitation stems from their insufficient consideration of the underlying data physics and meteorological uncertainties. To address this issue, we propose the VAE-WSDE-Net model. First, a variational autoencoder (VAE) learns a meteorological-power manifold from the data, establishing a foundation for physically consistent modeling. Then, we construct a WSDE-Net on this manifold. A key innovation is our tangential anisotropic noise injection mechanism, which aligns stochastic perturbations with the sensitivity of different physical features. This design provides more reliable prediction intervals. Additionally, we introduce a Wasserstein loss to resolve the gradient vanishing problem of the Negative Log-Likelihood (NLL) loss. Experimental results demonstrate the superiority of our model in both prediction accuracy and uncertainty quantification. Case study verified that its prediction intervals offer higher reliability and sharpness across diverse meteorological conditions.

This paper proposes a data cleaning method combining the Local Outlier Factor with Image Threshold Segmentation (LOF-ITSM), in order to address the problem that traditional image threshold segmentation methods cannot effectively detect outlier data in wind power curves, and global thresholds are difficult to adapt to curve fuzzy boundaries. Firstly, the LOF algorithm is used to detect wind power data and identify outlier types of abnormal data; Then, by introducing a local adaptive threshold optimization mechanism, the image threshold segmentation method is improved to better adapt to the fuzzy boundaries in the curve, while alleviating the sensitivity of the LOF algorithm to stacked abnormal data. The experimental results show that this method can better detect abnormal data in the wind power curve, laying a data foundation for promoting efficient utilization of wind energy and green transformation of energy structure.

Aiming at the problem that Anomaly Transformer model is difficult to detect multiple types of anomaly photovoltaic power data at the same time, this paper proposes an anomaly detection method for photovoltaic power generation data based on improved Anomaly Transformer (IAT). Firstly, the highly relevant features were filtered by Spearman feature selector. Secondly, the Wasserstein distance was introduced to improve the calculation method of prior association and sequence association to solve the problem of overlapping distribution between them. Finally, an anomaly evaluation strategy was proposed to smooth the abnormal reconstruction loss, so that it was easier to identify abnormal data. The experimental results show that the improved Anomaly Transformer improves the recall rate by 1.06

To address the problem that static knowledge graphs cannot evolve over time, which leads to the conflict between entities and relations in the process of knowledge representation, this paper combines the temporal hyperplane with the translation model in knowledge representation, and proposes a knowledge representation method based on the temporal hyperplane for power news texts. First, multiple temporal hyperplanes are established and the temporal factor is added to the scoring function of the translation model; then, the entities and relation of the power news are projected onto the temporal hyperplanes, and the optimal knowledge representation is determined according to the loss function. Taking the power news text as an example, this algorithm well resolves the time-related conflicts in the power news text, and the comprehensive indexes are significantly improved on the time-related triplets.

This study proposes a Cross Self-Calibrated Constant False Alarm Rate (CSC-CFAR) algorithm for millimeter-wave radar target detection, which innovatively integrates a cross-axial Training Cells selection mechanism with the Range-Doppler (RD) matrix to enable multi-directional signal characterization. A probabilistic assessment framework based on RD matrix intensity distribution is developed to enhance adaptability to diverse signal conditions and the threshold generation method, integrated with a ripple estimation approach utilizing Guard Cells, is designed to effectively mitigate the radial effects surrounding target points. Experimental results demonstrate that the proposed algorithm outperforms existing CFAR algorithms across diverse signal environments.

Addressing the challenge that news texts in the power field often contain numerous professional terms and many new terms are generated every year, which are difficult to accurately identify using general named entity recognition methods, this paper proposes an entity recognition model for power texts based on dependency syntactic analysis (SYN-BiLSTM-CRF). This model first generates power text word vectors and inputs them into a forward LSTM for feature extraction. Simultaneously, dependency syntactic parsing is performed on the power text, and the syntactic information vectors are fused with the output of the forward LSTM before being input into a backward LSTM. This enhances the model's ability to learn inter-word dependency relations by incorporating additional syntactic features. Finally, CRF is employed to obtain the predicted NER labels. The experiments demonstrate that the proposed SYN-BiLSTM-CRF model achieves an F1-score of 85.36% on power-related texts, representing a 2.78% improvement over the baseline BiLSTM-CRF model (82.58%). Additionally, it attains a recall of 89.06%, outperforming the BERT model's recall (87.59%). These results prove that the proposed method significantly enhances entity recognition accuracy in this specialized domain.

Aiming at the problem that current scenario analysis methods fail to fully capture complex time series correlations during scenario generation and do not consider time series similarities during scenario reduction, a wind power day-ahead scenario analysis method based on ICGAN and IDTW-Kmedoids is proposed. First, introducing a multi-time scale convolution layer into the CGAN scenario generation model(ICGAN) comprehensively extracts wind power time series correlation information, thereby improving scenario set generation quality. Secondly, the Kmedoids clustering algorithm (IDTW-Kmedoids) is used for scenario reduction. This algorithm uses an improved DTW algorithm to calculate the distance between clusters, which can better calculate the similarity of time series data and improve the effect of scenario reduction. The calculation results show that compared with traditional scenario analysis methods, this method can better capture the correlation and similarity of complex time series and can derive more representative typical scenarios.

As an important task of fine-grained sentiment analysis, aspect-based sentiment classification faces many challenges. In order to improve the classification accuracy, this paper focuses on the problems of long-distance semantic feature capture, emotional noise filtering, opinion word and aspect matching. This paper proposes an Interval-enhanced Talking-heads Attention Network (ITAN) for aspect-based sentiment analysis. Firstly, the limited semantic interval enhancement module was introduced to limit the semantic interval according to the aspect relative distance threshold, and a variety of word combination relations were generated in the whole sentence. Then, the gated filtering operation is used to integrate and extract the information enhancement representation of the limited interval. At the same time, the talking-heads attention mechanism module is combined to capture the emotional information from different perspectives to ensure that the opinion words and aspects are effectively matched. Finally, the sentiment classifier is used to integrate the information to generate the sentiment representation. The proposed model achieves 86.47

In response to the current low efficiency of CSI-based crowd-counting technology and the methods overly focus on debugging neural networks or mining more significant feature extraction methods, making the models complex and unstable in changing environments, the novel crowd-counting method named PhiC-Field is proposed in this paper, which is based on a cross-phase difference image generation technique. Firstly, the multi-antenna CSI subcarrier data is compressed and projected onto an orthogonal sampling field plane in pixel form. Secondly, by performing multi-point calculations, a blurred image reflecting the relationship between the number of people and the CSI subcarrier is obtained while preserving all subcarrier information without complex data preprocessing. This article proposes a neural network adapted to Phic-Field, and experiments in various experimental environments have shown that the accuracy of the model is higher than other methods.

The focus of this research lies in wireless gesture recognition, a prominent model of human-computer interaction that has garnered significant attention in recent years. This study aims to achieve efficient gesture recognition on smartphones using ultrasonic signals via speakers and microphones. Firstly, the original one-dimensional audio sequence is transformed into a two-dimensional spectrogram through the data preprocessing module, which facilitates capturing gesture patterns and changes while unifying the data format to enhance subsequent processing operability. Secondly, an algorithm for contour extraction is devised to mitigate signal interference caused by multipath effects. This algorithm enhances feature representation while reducing feature dimensionality and improving model robustness in adapting to interference and changes in various environments. By incorporating the spatial attention mechanism into the CNN model, it allows the model to focus on key areas, eliminate distracting information more effectively, better understand gesture features and shapes, and achieve more accurate recognition results. In testing basic number gestures, our method achieved an accuracy rate of 94%-96%, demonstrating its effectiveness. This would further enrich the spectrum of human-computer interaction, enhance user experiences, and drive the advancement of technological innovation.

Ultra-short-term prediction of photovoltaic power generation is the prerequisite for real-time dispatch of the power system, in order to further improve the accuracy of ultra-short-term prediction of photovoltaic power, this paper proposes an ultra-short-term prediction method of photovoltaic power based on an improved Generalized Regression Neural Network(GRNN)-Long Short-Term Memory (LSTM) combination model. Firstly, in the GRNN model, an improved whale optimization algorithm based on grid constraints, quasi-reverse learning and local perturbations is proposed to solve the optimal smoothing factor of the model and improve the prediction accuracy. Secondly, in the LSTM model, a CNN convolutional layer and attention mechanism are introduced to better exploit temporal information and enhance the predictive capabilities for complex weather changes. Finally, a combination model threshold setting method based on an improved Dynamic Time Warping (DTW) algorithm is proposed. This method dynamically selects the best model's prediction results based on thresholds during different time periods, fully leveraging the advantages of each model to meet the diverse weather conditions in photovoltaic power prediction. Experimental results demonstrate that the improved GRNN model excels in predicting sunny weather power, the improved LSTM model excels in predicting complex weather power, and the combination model performs well across various meteorological conditions, providing a robust solution for photovoltaic power ultra-short-term prediction problems.

Photovoltaic power is an important part of sustainable development. Accurate prediction of photovoltaic power can improve energy utilization and prevent resource waste. However, the volatility and uncertainty of photovoltaic power make power prediction difficult. Although Informer has achieved good prediction results in the field of time series prediction, it does not put forward a good solution for the volatility of series and the leakage of future information when stacking. Therefore, this paper proposes a photovoltaic power prediction model based on VMD-Informer-DCC. Firstly, Spearman's feature selector was used to screen the sequence features. Then, the VMD layer was added to the encoder of Informer to decompose the feature sequence to reduce the volatility of the feature sequence. Finally, the dilated causal convolutional layer was used to replace the Self-attention distilling of Informer, which expanded the receptive field of Informer information extraction and ensured the causality of time series prediction. To verify the effectiveness of the model, this paper uses the dataset of a photovoltaic power plant in Jilin Province in 2021 to conduct a large number of experiments. The results show that the VMD-Informer-DCC model has high prediction accuracy and wide applicability.