Search

Commercial load is an essential demand-side resource. Monitoring commercial loads helps not only commercial customers understand their energy usage to improve energy efficiency but also helps electric utilities develop demand-side management strategies to ensure stable operation of the power system. However, existing non-intrusive methods cannot monitor multiple commercial loads simultaneously and do not consider the high correlation and severe imbalance among commercial loads. Therefore, this paper proposes a deep learning-based non-intrusive commercial load monitoring method to solve these problems. The method takes the total power signal of the commercial building as input and directly determines the state and power consumption of several specific appliances. The key elements of the method are a new neural network structure called TTRNet and a new loss function called MLFL. TTRNet is a multi-label classification model that can autonomously learn correlation information through its unique network structure. MLFL is a loss function specifically designed for multi-label classification tasks, which solves the imbalance problem and improves the monitoring accuracy for challenging loads. To validate the proposed method, experiments are performed separately in seen and unseen scenarios using a public dataset. In the seen scenario, the method achieves an average F1 score of 0.957, which is 7.77% better than existing multi-label classification methods. In the unseen scenario, the average F1 score is 0.904, which is 1.92% better than existing methods. The experimental results show that the method proposed in this paper is both effective and practical.

• A multispectral imaging-based conveyor belt wear assessment approach is proposed. • A novel deep learning model for assessing wear state is constructed. • The proposed approach is effective, achieving an assessment accuracy of 98.78 %. • The model has a certain anti-interference ability to noise. Accurate assessment of the conveyor belt wear state is a crucial part of measuring belt conveyor safety and reliability. Therefore, this paper proposes an accurate detection approach for conveyor belt wear based on multispectral imaging(MSI), and designs a lightweight network model, named depthwise shuffle coordinate attention network (DSCANet) to assess and classify conveyor belts in three wear states. The multispectral images of the conveyor belt in the Huainan mining area were collected by the MSI system, with a wavelength range of 675–975 nm. The multispectral data at the wavelength with the largest imaging differences was screened as the input to the assessment model DSCANet. Compared with other widely used neural network models, the proposed DSCANet demonstrated the best performance, achieving a classification accuracy of 98.78 %, with floating point operations(FLOPs) of only 136.53M. The findings indicate the great efficacy of the MSI and DSCANet combination in assessing the conveyor belt wear, holding importance in reducing the risk of sudden failures and enhancing production efficiency. [ABSTRACT FROM AUTHOR]

[Display omitted] • The experimental system of heat load data acquisition for building air conditioning is constructed to obtain real and reliable data. • The Latin hypercube sampling is introduced to improve the initialization population of the sparrow search algorithm, which improves the performance of the algorithm. • A total of 12 models were implemented to forecast the heat load of building air conditioning, and ISSA-LSTM model shows the highest forecasting accuracy. • Deep learning and machine learning models were developed to forecast the heat load of building air conditioning and the differences between different models in forecasting the heat load of air conditioning have been identified. • The accuracy differences of heat load forecasting by optimizing LSTM model with different population algorithms have been studied. Building air conditioning heat load is a critical demand response resource, and its forecasting serves as a fundamental basis for optimizing building energy consumption control. The existing air conditioning load forecasting model suffers from reduced accuracy and stability due to the significant influence of temperature, humidity, and other factors. This paper proposes a method for building air conditioning heat load forecasting using Long Short-Term Memory neural network (LSTM) optimized by the Improved Sparrow Search Algorithm (ISSA). First, an experimental system for data collection on building air conditioning thermal load is constructed to form a dataset. Subsequently, the Latin Hypercube Sampling (LHS) is introduced to improve the SSA and iteratively optimize the hyperparameters of the LSTM model using ISSA. Finally, different optimization algorithms including Particle Swarm Optimizer (PSO), Crow Search Algorithm (CSA) and Spider Wasp Optimizer (SWO) were developed to optimize the LSTM model to achieve similar purposes, using the coefficient of determination as an indicator for evaluating model accuracy. The results show that the proposed data-driven new method is the most accurate model for forecasting building air conditioning heat load in this study. The R2 of the ISSA-LSTM forecasting model is as high as 0.9971, and compared with the RNN, LSTM, GRU, Bi-LSTM, SWO-LSTM, and SSA-LSTM models, the RMSE of this model in air conditioning heat load forecasting is reduced by 80.9670%, 71.7390%, 87.8040%, 88.3027%, 86.1179%, and 47.3089%, respectively. Employing the ISSA-LSTM method significantly enhances precision in building air conditioning load forecasting and holds promising applications in optimizing building energy consumption control. [ABSTRACT FROM AUTHOR]