33 results about "Mix network" patented technology

This invention discloses a method, equipment, storage medium, and program product for precise detection of slope sections. Through preprocessing and feature extraction using a CNN-LSTM hybrid network, interference such as noise, spikes, and numerical drift is effectively suppressed, enabling accurate identification of slope sections even in low signal-to-noise ratio environments. A two-level strategy of "AI coarse positioning + boundary statistical refinement" controls the positioning error of the slope's start and end points within 1-2 sampling points, achieving micron-level boundary positioning accuracy. This invention has advantages such as strong versatility, good engineering applicability, and convenient operation.

This invention discloses a signal modulation recognition method based on adaptive Fourier neural operators and gated attention, comprising the following steps: S1, acquiring a standardized signal X input S2, for the standardized signal X input S3. Perform feature embedding to compress the long sequence of dual-channel time-series signals into a high-dimensional feature vector sequence; S4. Construct a staged deep hybrid network architecture, including a shallow stage and a deep stage; S5. In the shallow stage, restore it back to the time-domain feature sequence space through a one-dimensional inverse fast Fourier transform; S6. In the deep attention stage, obtain a high-discriminative global context feature sequence; S7. Perform global average pooling on the high-discriminative global context feature sequence output by the deep network, aggregate it into a global feature vector, and map it to the modulation class space through a fully connected layer to calculate the probability distribution to complete the classification decision.

The application discloses a multi-field fake news detection method based on hierarchical potential field modeling, and belongs to the technical field of fake news detection. The method comprises the following steps: obtaining multi-angle multi-modal feature embedding of to-be-detected news by using a pre-training feature extractor; enhancing original field labels based on multi-modal features by using a field label enhanced memory network; constructing a single-modal and multi-modal processing flow by using a layer-stacking Transformer block, thereby modeling the field dependency relationship in news features in a hierarchical manner while adaptively fusing multi-modal features; introducing a variational mutual information decoupling module to supervise each layer of expert mixed network; obtaining a fake judgment feature by using a field perception Transformer, and outputting a fake judgment result through a classification head. By means of hierarchical field modeling and multi-modal feature fusion, the application effectively solves the distribution difference challenge existing in the multi-field adaptation of social media news.

The application discloses a method for adaptive secure consensus control of a multi-agent spatio-temporal dynamic system with unknown boundary nonlinearity under hybrid attacks, comprising the following steps: constructing a multi-agent spatio-temporal dynamic system model with unknown boundary nonlinearity and a virtual leader model; establishing a hybrid network attack model containing deception attacks and denial of service (Dos) attacks, and using an adaptive radial basis neural network to approximate the unknown boundary nonlinearity function; defining a consensus error signal to obtain a consensus state error system; designing a composite adaptive neural network secure boundary consensus control scheme, constructing a Lyapunov function for the error system, and obtaining a sufficient condition for the error system to realize mean square secure consensus control. The application effectively solves the problem that the traditional method is difficult to simultaneously cope with complex network attacks and unknown boundary nonlinear disturbances, and significantly improves the security and robustness of the system.

The application discloses a resource allocation method of a hybrid space division multiplexing elastic optical network and belongs to the technical field of optical communication. The application constructs a hybrid network model containing a multi-fiber bundle link and a multi-core optical fiber link, and designs a load and crosstalk aware heuristic algorithm. By means of modulation format step degradation, spectrum window plane construction to implicitly satisfy spectrum continuity constraints, load aware routing to balance network load, crosstalk aware core selection to suppress inter-core crosstalk, and optical signal-to-noise ratio verification to ensure transmission quality, joint optimization of routing, spectrum and spatial channels is realized. The application solves the problem that the prior art is only directed to a single type of space division multiplexing network and is difficult to adapt to a hybrid scene, effectively reduces network blocking rate under the premise of ensuring transmission quality, takes into account crosstalk suppression and spectrum efficiency, and has good engineering practicability.

Provided in the present disclosure is a cross-layer fault diagnosis method. The method comprises: acquiring a fault phenomenon, wherein the fault phenomenon has a fault manifestation in a hybrid network, and the hybrid network comprises two or more networks; and on the basis of the fault phenomenon, using a first large model to perform cross-layer fault diagnosis on the hybrid network, so as to obtain a diagnosis result, wherein the first large model is a model trained on the basis of knowledge in a knowledge base, and the knowledge comprises a cross-layer fault diagnosis rule. Further provided in the present disclosure are an electronic device, a readable medium and a program product.

The application provides a long-short-term combined reinforcement learning charging load prediction method and system, and belongs to the technical field of intelligent power distribution network optimal operation. The method comprises the following steps: acquiring multi-dimensional charging feature data of multiple charging stations in a region at continuous time steps to form a three-dimensional data matrix; extracting a spatial feature vector of each charging station; extracting a time feature vector of each charging station; generating a space-time fusion feature vector; constructing a multi-agent reinforcement learning prediction model, each charging station is defined as an agent, a global state vector is generated based on the local observation state of each agent and the global power grid state information, and the local value function of each agent is combined through a hybrid network to generate a global value function under the condition of meeting the monotonicity constraint to guide each agent to output a charging load prediction value at the next moment. The application deeply fuses the space-time features of the charging load, realizes multi-site collaborative prediction, and improves the prediction accuracy and robustness.

This invention discloses an oracle bone script recognition method, system, electronic device, and storage medium based on a hybrid CNN-Transformer network. The method first preprocesses the oracle bone script image and constructs a multi-source feature extraction network including local perception and global cognition branches. Then, a frequency-enhanced Transformer module is introduced into the global cognition branch to adaptively suppress high-frequency background noise through a frequency domain filtering mechanism, and a compact cross-channel interactive attention mechanism is combined to extract fine-grained stroke features of the oracle bone script to compensate for detail loss. Next, a selective branch gating fusion mechanism is introduced to dynamically evaluate the semantic importance of features and adaptively aggregate local perception and global cognition features. Finally, the oracle bone script recognition result is output. This invention can effectively overcome noise interference from complex rubbings and intra-class differences in character shapes, significantly improving the accuracy and robustness of oracle bone script recognition.

This invention discloses a method and system for intelligent identification and fault diagnosis of disconnector switch status in substations based on an attention mechanism. The method includes: acquiring current signals, vibration signals, and infrared thermal images of the disconnector switch to construct a multimodal time-series data stream; extracting temporal and spatial features using a Transformer-CNN hybrid network, and performing feature interaction through cross-modal cross-attention; constructing a graph neural network to encode physical coupling relationships as graph edges for message passing, and outputting a fault feature map; outputting diagnostic results and thermal fault segmentation masks through multi-task output using a classification head and a segmentation head; predicting current features and fault trend residuals using a multi-granularity self-attention network; and finally performing reliable decision fusion through a Venn-Abers module to output a comprehensive fault type with a confidence interval. This invention significantly improves the accuracy and robustness of disconnector switch fault diagnosis through multimodal feature fusion, graph neural network physical relationship modeling, and reliable decision fusion.

This invention provides a method for fine-grained identification of flexural structures based on the Transformer algorithm, comprising: extracting a set of flexural-sensitive attributes from 3D seismic data, performing attribute fusion and sample construction to obtain fused attribute samples and corresponding label datasets; combining a U-shaped network and the Transformer algorithm to obtain a hybrid network, and performing serialization and semantic embedding on the fused attribute samples to obtain serialized embedding features; inputting the serialized embedding features into the hybrid network, performing hybrid encoding and global context modeling, as well as upsampling reconstruction and skip fusion, to obtain pixel-level flexural segmentation results; performing training configuration and hyperparameter setting to obtain hyperparameter configuration, using the hyperparameters for network training and convergence to obtain an identification model, and outputting flexural structure identification results. This invention improves the expression of complex geometry and scale changes in flexural structures, achieving fine-grained identification and improved accuracy of small-scale flexural structures.

This invention relates to a battery pack balancing control method and system. The method includes: Step S1: treating the battery pack as an intelligent agent network and converting the battery pack balancing problem into a Markov game that can be handled by the DRL algorithm; Step S2: generating online expert data using the FCS-MPC algorithm; Step S3: training the intelligent agent network and a hybrid network based on different intelligent agent networks using the FCS-MPC-QMIX algorithm, and dynamically fusing the expert data into the objective functions of the intelligent agent network and the hybrid network during training to accelerate network convergence; Step S4: using the trained intelligent agent network and the hybrid network to achieve battery pack balancing control. This invention can provide balancing control for distributed battery energy storage systems.

The application discloses a steel rail fatigue crack classification and identification method based on a CNN-BiTCN-CA, a hybrid network architecture combining CNN and BiTCN is constructed, local perception and spatial features of signals are extracted through CNN, and meanwhile, an expanding causal convolution mechanism of BiTCN is used to extract dynamic time sequence dependency of signals in front and backward two time dimensions; a cross attention mechanism is introduced to complementarily fuse spatial features and time sequence features extracted in a cross road; and a steel rail crack evolution state is specifically divided into three physical stages of crack initiation, crack propagation and fracture. Experimental data show that the accuracy of the CNN-BiTCN-CA model in the above multi-classification task can reach 99.83%, and the single sample processing time is less than 0.1 millisecond, which meets the high-precision classification requirement of high-speed railway steel rail fatigue cracks and can provide more detailed and phased evaluation of the fatigue state of the steel rail.

This application belongs to the field of microgrid frequency control, specifically disclosing a model-free adaptive microgrid frequency control method to resist network attacks. The method includes the following steps: considering the mixed effects of denial-of-service attacks and spoofing attacks, a system output model incorporating attack coefficients is established, modeling the microgrid frequency control system as a discrete-time nonlinear system; dynamic linearization technology is used to transform the nonlinear system into an equivalent linear data model; for the linear data model, a partial pseudo-derivative estimation algorithm is designed, and a model-free adaptive control law incorporating a decay function is designed; the time-varying partial pseudo-derivative matrix is ​​iteratively updated using the partial pseudo-derivative estimation algorithm, and based on the updated partial pseudo-derivative matrix, the control input at the current moment is calculated using the model-free adaptive control law to control the microgrid frequency. This application can achieve effective frequency control even when the microgrid frequency control system model is unknown and faces mixed network attack threats.

The management and reporting of congestion in hybrid networks subject to congestion at MAC level requires adaptation of the L4S mechanism. Regarding uplink traffic, an AP receives a MAC frame marked with MAC congestion indication, reports to IP layer, the indication and the IP packet of the received frame, and responsive to the reporting, marks the packet with a CE codepoint. To speed up congestion reporting, a communication device detects congestion in a local transmit MAC queue storing MAC frames of a first flow of data traffic, and responsive to the detection, reports, to IP layer, a MAC congestion indication using a MAC service primitive. In particular, a communication device has a MAC service configured to communicate with an upper layer using a primitive that includes an ECN codepoint to report congestion and / or a MA-UNITDATA-STATUS.indication primitive that includes an ECE bit to report congestion.

A smart park system with hybrid networks based on 5G / 5G-A and IoT technologies is disclosed. The smart park system comprise: terminal device; a cloud network system, which include a 5G network and an all optical network; an IoT platform; an application platform; and an application service sub-system. The all optical network includes a core network switch, an optical line terminal OLT and an optical network unit ONU, wherein the OLT adopts a reverse feedback priority dynamic setting, in which the OLT sends a first data package via a first downlink channel, the OLT dynamically sets a priority of the first downlink channel from the OLT to the ONU according to a feedback for the first data package via a corresponding first uplink channel, and the OLT sends subsequent first data packages via the first downlink channel according to the set priority.

The application discloses a kind of intelligent substation secondary circuit fault arc online monitoring method and device, pulse optical signal is sent to the redundant optical fiber in the substation secondary circuit optical cable by optical time domain reflectometer, and Rayleigh backscattering signal is collected;Wavelet packet transform is carried out on scattering signal, the energy characteristics and entropy characteristics of each frequency band are calculated, and multidimensional feature vector is obtained;Non-dominated solution set is filtered by applying Pareto Front Optimization algorithm, and the Pareto optimal solution set of arc characteristics is obtained;The directed graph model of optical fiber sensing network is constructed, the arc candidate position is determined by applying directed minimum cut algorithm, and positioning is carried out by tree filling algorithm;Feature and position information are constructed into time sequence feature matrix, input into the CNN-LSTM hybrid network of Bayes online resource allocation optimization for classification identification;Locking decision strategy is generated based on fault diagnosis result, encapsulated as MMS message and sent to station control layer, and protection linkage is realized.

This disclosure provides a cross-layer fault diagnosis method, comprising: acquiring fault phenomena, wherein the fault phenomena are fault manifestations in a hybrid network, the hybrid network including two or more networks; based on the fault phenomena, performing cross-layer fault diagnosis on the hybrid network using a first major model to obtain diagnostic results, wherein the first major model is trained based on knowledge in a knowledge base, the knowledge including cross-layer fault diagnosis rules. This disclosure also provides an electronic device, a readable medium, and a program product.

The application discloses a kind of computing and communication efficiency cross-modal model collaborative training method and system, training method includes the following steps: using the hybrid network structure of fusion local perception and global semantic association, the model size is compressed by combining structure reparameterization technology to construct light cross-modal basic model architecture;Constructing a minimalist feature adaptation module, adding a minimalist feature adaptation module at the output end of the cross-modal basic model architecture, by gradually reducing its parameter quantity to a minimalist linear transformation form, local feature differentiation tuning is realized, while reducing the amount of communication transmission data;In local training, by comparing the feature distribution difference of public reference data set and local data, the distribution distance loss function is minimized, to promote the feature space of each participant to gradually align;Server side maintains global feature adaptation module parameters, each participant uploads updates after fine-tuning the module based on local data, and the global model is optimized by weighted average.

The application discloses a fuel cell multi-nozzle ejector control strategy based on a MOHSAC algorithm, relates to the technical field of proton exchange membrane fuel cell anode gas supply system control, and establishes an actuator characteristic submodel, obtains a mapping relationship from a discrete nozzle gear to a continuous valve opening mixed action to a total flow of primary flow through offline calibration; obtains actual state data of the ejector, constructs a multi-target state space and a mixed action space; constructs a layered multi-target reward function covering excess ratio tracking, pressure safety, ejector ratio efficiency and actuator life; establishes independent double Critic networks and mixed Actor networks based on the MOHSAC algorithm to perform iterative updating, and maintains a Pareto optimal strategy set; and in the deployment stage, calls a matching strategy according to a running condition to output a cooperative control instruction. The application realizes global cooperative optimization of the mixed action under multi-target conflicts.

The application belongs to the technical field of deep reinforcement learning, and discloses a multi-agent multi-task cooperative reinforcement learning method based on a space-time fusion architecture, which comprises the following steps: step 1, initializing a task sampling probability, forming an entity embedding vector sequence and a task embedding vector; step 2, inputting the entity embedding vector sequence and the task embedding vector into a noise-resistant feature extraction layer to obtain deep time sequence features; step 3, generating a dynamic weight matrix in real time to obtain local Q values output by an agent; step 4, a Transformer hybrid network receiving a global state vector of an environment, a task embedding vector and a local Q sequence to give a global action value; step 5, calculating a total loss; and step 6, based on the total loss, training an optimization unit to update all network parameters, and simultaneously, judging whether a preset evaluation round is reached. The application has stronger anti-interference ability and more stable control performance, and realizes adaptive control of multiple heterogeneous tasks.

The application relates to a multi-unmanned aerial vehicle cooperative confrontation method and system based on role representation and mutual information constraint, a multi-unmanned aerial vehicle cooperative confrontation scene is established; local observation information and joint action trajectory are acquired, a role coding network is trained by sampling batch data, potential semantic role representation of the unmanned aerial vehicle of the own side is extracted by using the role coding network after training; according to a dynamic reward threshold, the collected trajectory data and the corresponding potential semantic role representation are stored in a two-channel storage structure in a hierarchical manner, a mutual information estimator is synchronously trained; based on an internal reward and punishment mechanism and the mutual information estimator after training, a corrected total reward signal is calculated, a strategy network is evaluated and optimized, a strategy is updated through gradient back propagation of a mixed network, and the cooperative confrontation strategy of the multi-unmanned aerial vehicle of the own side is optimized; and an antagonistic system is realized based on the method. The application provides high-level semantic support for decision-making in a complex confrontation environment, improves convergence efficiency and robustness, and improves the accuracy of global value evaluation and credit distribution.

The application discloses a dynamic threshold leakage identification method based on hierarchical time series prediction, and relates to the field of leakage detection of water supply systems. By obtaining water quantity or water meter data of different levels with hour granularity, standard time series data sets are obtained through format conversion, time series completion and data merging processing. The top, middle and bottom sequences are identified. For water types lacking direct monitoring data, a virtual sequence is constructed by using hierarchical aggregation constraint coupled moving average filtering to complete the aggregation constraint matrix. A time series prediction model with hour granularity is constructed based on a hierarchical hybrid network to determine the prediction performance with coefficient of determination and root mean square error. The upper limit of the interval is used as the normal water consumption boundary by using the prediction interval of different confidence levels output by the model. The optimal dynamic threshold proportion is determined by traversing the threshold candidate range. When the actual water quantity exceeds the proportion, it is determined as potential leakage. The accuracy, precision and recall rate are used as evaluation criteria to maximize the recall rate and reduce the missed detection loss.

Techniques for a hybrid network directory service are described. Messages forming a request to launch an instance within a cloud provider network are received, the messages including an identifier of a customer virtual network within the cloud provider network, the customer virtual network having connectivity to another network outside of the cloud provider network, the other network outside of the cloud provider network having a directory service. An instance is launched, the instance having connectivity to the customer virtual network. A server of the directory service on the other network outside of the cloud provider network is identified. The identified server is caused to add the instance as a node of the directory service. Directory service data received from the identified server is stored. Directory service requests originating from the customer virtual network are processed.

This invention discloses a spatiotemporal traffic flow prediction method based on dual-flow decoupling. The method first decomposes the temporal data of road network traffic flow into trend and periodic components using the exponential moving average method. Then, features are extracted and fused using a deep linear network and a local feature hybrid network with temporal block embedding to obtain the global temporal prediction component. After the traffic flow temporal data is encoded in the temporal domain by gated dilated convolution, multi-view spatial aggregation is performed using forward, backward, and normalized graph convolutions with adaptive adjacency matrices to obtain the local spatiotemporal prediction component. Finally, dynamic weights are generated by a gated network, and the two types of components are weighted and fused to obtain the final prediction result. This invention accurately separates long-term and short-term traffic flow features, adaptively balances global patterns and local details, improves the accuracy and robustness of long-term temporal prediction, and reduces computational complexity, making it suitable for intelligent traffic flow management scenarios.

This invention provides a method for ultra-short-term load forecasting of urban distribution networks based on deep embedding clustering and ICEEMDAN-iTransformer. Deep embedding clustering is used to refine the classification of user load curves and extract typical regional electricity consumption patterns to achieve scenario dimensionality reduction. Furthermore, an innovative iTransformer prediction module based on a hierarchical multi-scale attention mechanism and a dynamic sparse expert hybrid network is constructed using the load sequences after ICEEMDAN adaptive decomposition. This effectively captures multivariate temporal dependencies, enabling accurate prediction of load sequences after removing redundant features, significantly improving the accuracy and computational efficiency of ultra-short-term load forecasting. This invention helps power grid companies achieve accurate ultra-short-term load forecasting for multiple regions and types of users in urban distribution networks, providing reliable data support for power generation planning, power flow optimization, and demand-side response scheduling, thereby improving the safety and economy of power grid operation.

The application provides a cell selection method and related device. In the method, a dynamic effective factor (DECF) mechanism is proposed. A first parameter is determined based on the mechanism. The first parameter is used to quantify the link resource cost of a cell. The NTN cell and the TN cell can intelligently switch under actual service demand and link state. The mobility management performance of the NTN / TN hybrid network is improved. The utilization rate of network resources is further improved. And the user experience is optimized.

The application provides a multi-center node cross-domain heterogeneous interconnection proxy network architecture design method and system, comprising the following steps: S1, connecting multiple star networks to build a hybrid network structure; wherein each star network comprises a center node and multiple heterogeneous systems, and the center node is directly interconnected with the multiple heterogeneous systems; S2, deploying a center proxy of a corresponding simulation engine, a service proxy corresponding to each subsystem, and a transfer proxy for managing data forwarding between the multiple center nodes in each center node; S3, setting a heterogeneous proxy corresponding to the multiple center nodes in each heterogeneous system connected with the center nodes; and S4, configuring the IP address of the DDS communication middleware, the subscription and publication of the domain, and the topic strategy for the four proxies in steps S2 and S3.

This disclosure relates to the field of communication technology, and in particular to a method, apparatus, device, medium, and product for packet loss detection in a hybrid optoelectronic network. The method includes: receiving at least one message container, the message container including message container information and at least one data packet; the message container information including a data stream identifier, a detection type, and coloring information; the detection type including a packet loss detection type or a non-packet loss detection type; and the coloring information including the coloring period and coloring identifier of the data packets in the data stream; determining a first data stream that meets the packet loss detection conditions based on the message container information of each message container; and if packet loss occurs in the data packets received by the first data stream, sending a packet loss event to a management node, the packet loss event being used to indicate that packet loss has occurred in the first data stream. The technical solution of this disclosure provides a packet loss detection method for a hybrid optoelectronic network, achieving effective packet loss detection.