Search

With the development of multi-source AC/DC distribution networks, DC fault detection technology has become the key to DC protection. Aiming at the characteristics of large fault current with a rapid rise and difficulty to extract fault features when distribution line faults occur in the DC part, a fault detection technology of AC/DC distribution network based on deep belief network (DBN) is proposed combining feature extraction in the time domain and frequency domain. By analyzing the characteristics of the fault equivalent circuit, Fourier transform and phase-mode transform are used to extract the frequency domain and time domain characteristics of the fault current and voltage signals as the input of DBN, and the Softmax classifier is used to output the fault pole selection and fault area identification results. An AC/DC distribution network model is built on PSCAD to test the algorithm. The simulation result shows that the proposed detection method still has high accuracy and the strong ability to tolerate noise under the influence of line-distributed capacitance and control strategy. The further algorithm comparison experiment shows that the necessary combination of fault feature extraction and effective deep learning model training can complete the complex fault detection.

IntroductionImmunity is involved in a variety of bone metabolic processes, especially osteoporosis. The aim of this study is to explore new bone immune-related markers by bioinformatics method and evaluate their ability to predict osteoporosis.MethodsThe mRNA expression profiles were obtained from GSE7158 in Gene expression Omnibus (GEO), and immune-related genes were obtained from ImmPort database (https://www.immport.org/shared/). immune genes related to bone mineral density(BMD) were screened out for differential analysis. protein-protein interaction (PPIs) networks were used to analyze the interrelationships between different immune-related genes (DIRGs). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses of DIRGs function were performed. A least absolute shrinkage and selection operation (LASSO) regression model and multiple Support Vector Machine-Recursive Feature Elimination (mSVM-RFE) model were constructed to identify the candidate genes for osteoporosis prediction The receiver operator characteristic (ROC) curves were used to validate the performances of predictive models and candidate genes in GEO database (GSE7158,GSE13850).Through the RT - qPCR verify the key genes differentially expressed in peripheral blood mononuclear cells Finally, we constructed a nomogram model for predicting osteoporosis based on five immune-related genes. CIBERSORT algorithm was used to calculate the relative proportion of 22 immune cells.ResultsA total of 1158 DEGs and 66 DIRGs were identified between high-BMD and low-BMD women. These DIRGs were mainly enriched in cytokine−mediated signaling pathway, positive regulation of response to external stimulus and the cellular components of genes are mostly localized to external side of plasma membrane. And the KEGG enrichment analysis were mainly involved in Cytokine−cytokine receptor interaction, PI3K−Akt signaling pathway, Neuroactive ligand−receptor interaction,Natural killer cell mediated cytotoxicity. Then five key genes (CCR5, IAPP, IFNA4, IGHV3-73 and PTGER1) were identified and used as features to construct a predictive prognostic model for osteoporosis using the GSE7158 dataset.ConclusionImmunity plays an important role in the development of osteoporosis.CCR5, IAPP, IFNA4, IGHV3-73 and PTGER1were play an important role in the occurrences and diagnosis of OP.

Sensor-based human activity recognition has been widely used in various fields; however, there are still challenges involving recognition of daily complex human activities using sensors. In order to solve the problem of timeliness and homogeneity of recognition functions in human activity recognition models, we propose a human activity recognition model called 'BSTCA-HAR' based on a long short-term memory (LSTM) network. The approach proposed in this paper combines an attention mechanism and a temporal convolutional network (TCN). The learning and prediction units in the model can efficiently learn important action data while capturing long time-dependent information as well as features at different time scales. Our series of experiments on three public datasets (WISDM, UCI-HAR, and ISLD) with different data features confirm the feasibility of the proposed method. This method excels in dynamically capturing action features while maintaining a low number of parameters and achieving a remarkable average accuracy of 93 % , proving that the model has good recognition performance. [ABSTRACT FROM AUTHOR]