Photovoltaic power has great volatility and intermittency due to environmental factors. Forecasting photovoltaic is of significance ensure the safe economical operation distribution network. This paper proposes a novel approach forecast short-term based on gated recurrent unit (GRU) Firstly, Pearson coefficient used extract main features that affect output at next moment, qualitatively analyze relationship between historical future output. Secondly, K-means method utilized divide training...

Small and cluttered objects are common in real-world which challenging for detection. The difficulty is further pronounced when the rotated, as traditional detectors often routinely locate horizontal bounding box such that region of interest contaminated with background or nearby interleaved objects. In this paper, we first innovatively introduce idea denoising to object Instance-level on feature map performed enhance detection small To handle rotation variation, also add a novel IoU...

The fluctuation and intermittence of wind power bring great challenges to the operation control distribution network. Accurate short-term prediction for is helpful avoid risk caused by uncertainties powers. To improve accuracy power, temporal convolutional network (TCN) proposed in this paper. method solves problem long-term dependencies performance degradation deep model sequence dilated causal convolutions residual connections. simulation results show that training process TCN very stable...

Existing methods for transformer fault diagnosis either train a classifier to fit the relationship between dissolved gas and type or find some similar cases with unknown samples by calculating similarity metrics.Their accuracy is limited, since they are hard learn from other algorithms improve their own performance.To of diagnosis, novel method based on graph convolutional network (GCN) proposed.The proposed has advantages two kinds existing methods.Specifically, adjacency matrix GCN...

Accurate short-term solar and wind power predictions play an important role in the planning operation of systems. However, prediction renewable energy has always been considered a complex regression problem, owing to fluctuation intermittence output powers law dynamic change with time due local weather conditions, i.e. spatio-temporal correlation. To capture features simultaneously, this paper proposes new graph neural network-based forecasting approach, which combines convolutional network...

In this article, distinctive fault characteristics of converter-interfaced renewable energy sources (CIRESs) will result in the misoperation traditional distance relays. To handle issue, a new coordination control method is proposed to ensure correct operation Combined with sequence boundary conditions, relationship between apparent reactance and CIRES positive-sequence current angle revealed. Based on this, reasonable generated by adjusting references controller make equal actual line...

Reliable and accurate ultra-short-term prediction of wind power is vital for the operation optimization systems. However, volatility intermittence pose uncertainties to traditional point prediction, resulting in an increased risk system operation. To represent uncertainty power, this paper proposes a new method interval based on graph neural network (GNN) improved bootstrap technique. Specifically, adjacent farms local meteorological factors are modeled as form from graph-theoretic...

The volatility of renewable energy and the time variability load bring challenges to forecasting. To improve accuracy prediction, paper use gated recurrent unit (GRU) network forecast short-term considering impact electricity price. First, methods are proposed group historical data based on main features input. Second, rules classification tree established judge which cluster new belongs to. is trained by input power from selected group. Finally, real-world used analyses price demonstrate...

The daily load profiles modeling is of great significance for the economic operation and stability analysis distribution network. In this paper, a flow-based generative network proposed to model Firstly, real samples are used train series reversible functions that map probability prior distribution. Then, new generated by taking random number obeying Gaussian as input data these functions. Compared with existing methods such explicit density models, approach does not need assume samples, can...

The widespread penetration of advanced metering infrastructure brings an opportunity to detect electricity theft by analyzing the consumption data collected from smart meters. However, existing models have poor performance in detection, since most them fail capture time dependence, periodicity, and latent feature complex data. To address above concerns, a graph convolutional neural network (GCN) Euclidean (CNN) are combined form novel model for detection this paper. On one hand,...

Machine learning techniques have been extensively developed in the field of electricity theft detection. However, almost all typical models primarily rely on consumption data to identify fraudulent users, often neglecting other pertinent household information such as gas data. This article aims explore untapped potential data, a critical yet overlooked factor In particular, we perform theoretical, qualitative, and quantitative correlation analyses between Then, propose two model-agnostic...

Existing substation equipment image data augmentation models face challenges such as high dataset size requirements, difficult training processes, and insufficient condition control. This paper proposes a transformer method based on Stable Diffusion model. The proposed incorporates the Low-Rank Adaptation (LoRA) concept to fine-tune pre-trained model weights, significantly reducing requirements while effectively integrating essential features of data. To minimize interference from complex...

Although the penetration of electric vehicles (EVs) in distribution networks can improve energy saving and emission reduction effects, its random uncertain nature limits ability to accept load EVs. To this end, establishing a profile model EV charging stations accurately reasonably is great significance planning, operation scheduling power system. Traditional generation methods for profiles rely too much on experience, need set up probability advance. In paper, we propose data-driven...

Due to the strong concealment of electricity theft and limitation inspection resources, number power samples mastered by department is insufficient, which limits accuracy detection. Therefore, a data augmentation method for detection based on conditional variational auto-encoder (CVAE) proposed. Firstly, stealing curves are mapped into low dimensional latent variables using encoder composed convolutional layers, new reconstructed decoder deconvolutional layers. Then, five typical attack...

High-quality datasets are of paramount importance for the operation and planning wind farms. However, collected by supervisory control data acquisition (SCADA) system may contain missing due to various factors such as sensor failure communication congestion. In this paper, a data-driven approach is proposed fill farms based on context encoder (CE), which consists an encoder, decoder, discriminator. Through deep convolutional neural networks, method able automatically explore complex...

Unique fault behaviors of renewable energy sources (RESs) may lead to the misoperation traditional pilot protection. To cope with this issue, article proposes a new protection method using improved Euclidean distance. For normal operation or external faults, currents on both ends are completely opposite, so distance current absolute values is equal 0. However, it will be much larger than 0 for internal faults because transient have big difference at time. Therefore, and can detected...

Supervised machine learning models are receiving increasing attention in electricity theft detection due to their high accuracy. However, performance depends on a massive amount of labeled training data, which comes from time-consuming and resource-intensive annotations. To maximize model within limited annotation budget, this paper aims reduce the effort through optimal sample selection. In particular, general framework three new strategies proposed select most valuable representative...

Small and cluttered objects are common in real-world which challenging for detection. The difficulty is further pronounced when the rotated, as traditional detectors often routinely locate horizontal bounding box such that region of interest contaminated with background or nearby interleaved objects. In this paper, we first innovatively introduce idea denoising to object Instance-level on feature map performed enhance detection small To handle rotation variation, also add a novel IoU...