39 results about "Balancing network" patented technology

The embodiment of the invention relates to an analog simulation method based on electric power spot transaction. The analog simulation method comprises the steps of obtaining static structure data such as power grid topology and unit parameters and dynamic time sequence data such as load and new energy prediction; performing preprocessing and multi-time scale division on the dynamic data; constructing a multi-agent model with differentiated characteristics based on the processed data; each agent generates and reports energy block information such as a time-sharing block, a continuous block, a curve block or a variable block according to a perceived market environment; constructing a market clearing model considering system power balance, network power flow, unit operation and energy block characteristic constraint based on all declarations, and solving to obtain clearing electricity price and electric quantity; a result is fed back to the intelligent agent, and the strategy is updated through deep reinforcement learning; and performing rolling co-simulation between the day-ahead market and the real-time market, wherein a day-ahead clearing result is used as a boundary condition of the real-time market. The authenticity, accuracy and practicability of market simulation are improved.

The invention provides a hybrid routing method and device based on congestion awareness, and the method comprises the steps: determining a source node and a target node of a data packet, the source node representing a starting router node which initiates data transmission, the target node representing a termination router node which finally transmits data, and the destination node representing a termination router node which finally transmits the data; acquiring an input buffer occupation state of a target adjacent node corresponding to the source node, if the input buffer occupation state of the target adjacent node is an idle state, transmitting the data packet through a target shortest path corresponding to the target adjacent node, and if the input buffer occupation state of the target adjacent node is a congestion state, transmitting the data packet through a target shortest path corresponding to the target adjacent node. If yes, other nodes of the source node in the second target direction are determined, other shortest paths corresponding to the other nodes are obtained, and the data packet is transmitted to the target node through the other shortest paths. After the output port with the lightest load is selected for the data packet to bypass the congestion area near the source node, the algorithm is simplified while the network load is balanced, and overhigh routing overhead is avoided.

The application relates to the technical field of high-frequency inverter output current sharing, and discloses a high-frequency inverter parallel current sharing method and system. According to the application, for multiple parallel high-frequency inverters, the output currents of the high-frequency inverters flow into a parallel confluence node and subsequent circuits after being balanced by a preset impedance balancing network. The impedance balancing network is modular, and comprises multiple modular current sharing branches. One end of each current sharing branch is connected to the output end of each high-frequency inverter in one-to-one correspondence, and the other end of each current sharing branch is connected to be suspended. The parameter design of the impedance balancing network is determined based on the connection between the output end of each high-frequency inverter and the parallel confluence node and the subsequent circuits. The balancing method has the characteristics of few devices, wide frequency range and strong extensibility of multi-inverter current balancing.

The application discloses a kind of anonymous routing methods of double-layer satellite network for delay-sensitive service, it is applied to VLEO / LEO double-layer satellite network, this network is based on software-defined network SDN architecture implementation, including data plane, control plane and application plane;Data plane includes VLEO layer, VLEO / LEO cross layer and LEO layer;Application plane is deployed with the routing method of the application upper;The method comprises: constructing layered decoupling routing calculation framework, to divide routing decision into including LEO layer inner routing, VLEO / LEO cross layer routing and VLEO layer inner routing multilayer path selection;Using the collaborative computing capability of SDN controller and satellite node, multilayer path selection is reconfigured and weight disturbance in real time, generates optimal routing path.The application can effectively reduce end-to-end delay, balance network load, improve the real-time performance, robustness and security anonymity of double-layer satellite network.

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.

The invention provides a dialogue reply method and device, equipment, a medium and a program product, and the method comprises the steps: obtaining dialogue data, inputting the dialogue data into a trained large language model based on a deep learning network, and obtaining a dialogue reply feature, the deep learning network comprises at least one first module, each first module comprises a second module with at least one first basic unit and a second basic unit which are connected in sequence, or the second module with at least one second basic unit and the first basic unit which are connected in sequence, and the first basic unit and the second basic unit comprise a normalization layer and a sub-layer; the output of the normalization layer is used as the input of the sub-layer, the input of the normalization layer in the first basic unit and the output of the sub-layer are subjected to residual connection, the output of the normalization layer in the second basic unit and the output of the sub-layer are subjected to residual connection, and reply is carried out according to dialogue reply features. The stability and model performance of the network are improved by balancing the internal structure of the network.

The invention discloses a collaborative environment-oriented service deployment and request scheduling optimization method and device, and relates to the technical field of edge-cloud collaborative mobile edge computing in the power industry, and the method comprises the steps: constructing a multi-instance queuing network model and a multi-dimensional hierarchical soft actor-commentator algorithm; based on the multi-instance queuing network model, modeling a multi-instance execution relationship in a power station task processing process, and providing a performance evaluation basis for a request scheduling strategy by quantifying task transmission delay, queuing delay and calculation delay; and based on the multi-dimensional hierarchical soft actor-commentator algorithm and in combination with the current state of the power information system, determining the deployment number and distribution position of each micro-service instance, and formulating a request scheduling strategy. According to the method, the dependency relationship and the multiplexing characteristic between the micro-services are fully considered, the request response delay can be effectively reduced, the network load and the system energy consumption can be balanced, and the service quality and the resource utilization efficiency of the electric power station are improved.

The invention discloses an energy storage grading charging and discharging regulation and control method for stabilizing photovoltaic power fluctuation, and relates to the technical field of photovoltaic power generation and energy storage cooperative control, and the method is applied to a photovoltaic power station system comprising a group cascade self-balancing networking unit, a station-level central networking energy storage system and a station-level cooperative networking controller. Power deviation caused by local shadow is monitored and compensated in real time through a group cascade unit, power cooperative scheduling and reactive power dynamic distribution are carried out on a station level layer, cooperative optimization of photovoltaic array internal power balance and grid-connected point power grid support is achieved, and meanwhile, through virtual synchronous machine cluster control and a protective off-grid / re-grid-connected mechanism, the photovoltaic array internal power balance and the grid-connected point power grid support are optimized. According to the method, power fluctuation can be effectively stabilized in a complex shielding environment, the power generation efficiency of the photovoltaic power station and the interaction capability of a power grid are improved, and the stability and reliability of a power system are enhanced.

This invention relates to a method and apparatus for collaborative trajectory planning and energy efficiency optimization of UAV networks for multi-target surveillance. The method achieves deep integration of routing pairing, trajectory planning and power control by constructing a multi-target collaborative optimization framework, designing a hierarchical linkage algorithm, and enhancing scenario adaptability and dynamic response capabilities. This solves the problem of balancing network performance, lifetime and energy efficiency in multi-target surveillance scenarios and promotes the large-scale application of UAV swarms in complex surveillance tasks.

Embodiments of the present application relate to a kind of simulation method based on power spot trading, comprising: obtaining grid topology, unit parameter and other static structure data and dynamic time series data such as load, new energy prediction;Dynamic data is preprocessed and divided into multiple time scales;Based on the data processed, construct a multi-agent model with different characteristics;Each agent generates and reports energy block information such as time block, continuous block, curve block or variable block according to the perceived market environment;Based on all reports, construct a market clearing model considering system power balance, network power flow, unit operation and energy block characteristic constraints and solve, obtain clearing price and power;The results are fed back to the agent, and the strategy is updated by deep reinforcement learning;Rolling collaborative simulation is carried out between day-ahead and real-time market, wherein the day-ahead clearing result is used as the boundary condition of real-time market.The present application improves the authenticity, accuracy and practicality of market simulation.

The application discloses a link bandwidth allocation method and allocation system for multi-mode mobile network multi-link aggregation, comprising: obtaining mobile network base station, device position, real-time environment data; using intelligent power control algorithm, dynamically adjusting transmission power according to environment data and network load; obtaining real-time signal strength data and historical data of network device to base station; predicting minimum transmission point of signal transmission based on real-time and historical data; adjusting antenna direction and position according to base station position, and designing connection using beamforming antenna; setting threshold value according to real-time weather and minimum transmission point data, and deploying signal enhancer; the application discloses a link bandwidth allocation method and allocation system for multi-mode mobile network multi-link aggregation, through intelligent algorithm, real-time monitoring of link quality, dynamic adjustment of network, balancing of network load, and combination of real-time weather data for evaluating signal transmission environment, deploying signal enhancer, and improving signal coverage range.

The present disclosure generally relates to optimizing load-balancing of network endpoints using tree collectives representing a logical network communication topology for the network endpoints. Systems and methods described herein eliminate the previously restrictive conditions imposed on tree-based communication collectives by generating collective trees with any arity and representing any number of physical network endpoints. The resulting collective trees ensure that each represented network endpoint has a number of outgoing flows and a number of incoming flows that are no more than the arity of the collective tree. In this way, the described systems and methods inject significant efficiencies into communication collectives within networked compute nodes by eliminating communication bandwidth latencies and bottlenecks.

The application belongs to the technical field of distributed online optimization and learning, and discloses a quantized distributed online proximal gradient descent optimization method based on Bandit feedback. The method aims to solve the distributed online composite optimization problem under the condition that the communication resource is limited, the network structure is unbalanced, and the gradient information of the loss function is difficult to obtain or even cannot be obtained. The technical scheme comprises the following steps: constructing an unbalanced network graph and determining the node neighborhood relationship, modeling a distributed online composite optimization problem based on Bandit feedback, designing a quantized distributed online proximal gradient descent algorithm combined with an adaptive uniform quantization strategy and a proximal gradient descent technology, and performing static regret analysis on the algorithm.

The application discloses an output driving circuit, which comprises an input unit, a multi-stage output unit, a load unit, a feedback network, a bias network and a balance network; the input unit, the multi-stage output unit and the load unit are connected in sequence, the feedback network is connected with the output of the multi-stage output unit, the bias network is connected with the input unit and the multi-stage output unit, and the balance network is connected with the multi-stage output unit. The output driving circuit has the advantages of large output swing, high output bandwidth and good reliability.

The present disclosure generally relates to optimizing load-balancing of network endpoints using tree collectives representing a logical network communication topology for the network endpoints. Systems and methods described herein eliminate the previously restrictive conditions imposed on tree-based communication collectives by generating collective trees with any arity and representing any number of physical network endpoints. The resulting collective trees ensure that each represented network endpoint has a number of outgoing flows and a number of incoming flows that are no more than the arity of the collective tree. In this way, the described systems and methods inject significant efficiencies into communication collectives within networked compute nodes by eliminating communication bandwidth latencies and bottlenecks.

Embodiments of the present application provide a data processing method, a network training method, an electronic device and a storage medium. The method comprises: obtaining input data of a network layer of a deep neural network, original weights of the network layer being decomposed into N sub-weights for quantization processing; performing convolution processing on the input data and the N sub-weights respectively to obtain N intermediate data corresponding to the N sub-weights; and performing summation processing on the N intermediate data to obtain output data of the network layer. In this way, the original weights of the network layer are decomposed into the N sub-weights for quantization processing, and the efficiency and performance of the network can be flexibly balanced by adjusting the value of N.

Disclosed are systems, apparatuses, methods, and computer-readable media for balancing replicated frames in a replication network. A method includes: receiving, at a redundancy receiver, first frames from a first path and second frames from a second path, determining a network performance imbalance between a redundancy edge device to the redundancy receiver based on receiving the first frames and the second frames, and in response to determining a threshold of memory will be exceeded based on the delay between the first frames and the second frames, providing an indication to slow frames transmitted along the second path. In some aspects, a path mismatch device is configured to track replication frames and determine network performance imbalance. For example, the path mismatch device can identify one network is slower than another, and may provide remediation to balance network performance that maintains the integrity of the replication network.

The invention belongs to the technical field of remote sensing image processing, and particularly relates to a multi-modal remote sensing data classification method based on a semi-supervised contrast regular network. The method comprises the following steps: generating high-quality pseudo labels for all label-free multi-modal remote sensing data by adopting a multi-modal label distribution method, screening samples with balanced categories according to the pseudo labels, constructing a label-free sample set and a label sample set, and performing enhancement processing on the sample sets; constructing a semi-supervised contrast regular network; performing supervised learning of the network by using the labeled sample set; performing consistent regular learning of the network by using the reliable unmarked sample; contrastive learning is executed, an anchor sample set, a positive sample set and a negative sample set are established according to reliability judgment, contrastive loss is calculated, and unreliable samples provide negative sample features for the loss; and jointly using three loss functions to jointly optimize the classification performance of the network. The multi-modal remote sensing data classification method is beneficial to realizing category-balanced network optimization in training, enhances the network robustness, improves the network classification performance, and can be widely applied to multi-modal remote sensing data classification.

Load balancing is provided between network devices to perform DPI operations. Network devices in a virtualized network device array can balance network traffic by using a network connection between the devices to exchange data packets for DPI operations. Furthermore, the network devices can use dedicated CPUs to offload DPI operations. In this way, the techniques presented allow existing network routing or deterministic routing to be maintained, enabling DPI to be performed on the fly without restricting network traffic to a specific route.

The invention discloses a wireless unmanned aerial vehicle edge network calculation unloading method and device, and belongs to the field of unmanned aerial vehicle edge network calculation, and the method comprises the steps: enabling an unmanned aerial vehicle to be virtualized into an access node and a calculation node, and enabling the unmanned aerial vehicle to flexibly participate in communication and calculation tasks. And then a complex multi-dimensional problem is converted into a trilateral matching optimization problem by the constructed trilateral matching game model, so that the decision space is greatly reduced. And the setting of the dynamic preference rule enables the three-party agent to carry out adaptive adjustment according to the real-time state of the network, then the distributed iterative optimization process can gradually improve the matching result, and finally a better resource allocation state is achieved. And finally, the hierarchical task unloading is realized, so that the high-performance computing unmanned aerial vehicle and the relay unmanned aerial vehicle can reasonably share tasks according to own characteristics and capabilities, the distribution and utilization of network resources are optimized, the overall computing load of the network is effectively balanced, the average service time delay of the computing tasks in the network is reduced, and the utilization rate of the computing resources is improved.

Various configurations of modular data center systems are described to enable a rapidly deployable data center that can be scaled to the demand and deployed more efficiently with significantly less field construction labor requirements. The systems include combinations of any number of separate IT modules, energy storage modules and transformer modules. A control and monitoring system monitors power demand of the IT equipment of the IT module, controlling power flow between the IT equipment of the IT module and the energy storage module, and balancing the network data load in response to compute load variations experienced by the IT equipment. In another configuration, a single module may include all IT equipment, energy storage and transformers / power conversion subsystems necessary for it to serve as a standalone data center unit.

A network device in a network is provided. During operation, the network device can receive a static multicast route associated with a multicast group and a source of the multicast group. Here, the network device can include a plurality of ingress interfaces, each corresponding to a path via which the network device is reachable from the source. The network device can determine, based on an indicator in the static multicast route, that the network device is to dynamically select an ingress interface for the static multicast route. The network device can then determine respective weights allocated to the plurality of ingress interfaces and select the ingress interface for the static multicast route from the plurality of ingress interfaces based on their respective weights. Subsequently, the network device can program the static multicast route in forwarding hardware of the network device with the selected ingress interface.

The invention relates to the technical field of unmanned aerial vehicle safety, in particular to a multi-unmanned aerial vehicle distributed formation safety control method based on a dual network, and the method comprises the steps: building a dynamic model, a connectivity index and a DoS attack sequence, and defining a formation tracking error and a control target. Fusing the characteristic parameters to construct continuous comprehensive communication quality evaluation indexes, constructing a hysteresis adaptive mode switching rule with an anti-jitter characteristic, setting an adaptive contraction network controller based on a recursive balance network, and generating a distributed control law adaptive to a current mode; based on a shrinkage mapping theory and a multi-Lyapunov function method, the index stability of a switching system is analyzed, verification operation is carried out through a simulation experiment, a recursive balance network architecture is adopted to construct an adaptive shrinkage network controller, the real-time control capability of an embedded platform is improved, and the real-time performance of the embedded platform is improved. And the conflict problem between the calculation complexity and the real-time requirement is solved.

The invention provides a network flow data packet balanced loading method and device and a medium, and relates to the technical field of network flow data packet balanced loading, and the method comprises the steps: determining a pressure index used for representing the overall balanced state of a system; judging whether the system is in an equilibrium state or not according to the numerical range of the pressure index; if the system is not in the equilibrium state, adjusting the transfer rate through a preset feedback control algorithm; the control quantity of the feedback control algorithm is comprehensively determined according to the pressure index, the network card cache area occupancy rate relative to a target value and the error of the disk falling cache area occupancy rate relative to the target value; transferring the network flow data packet temporarily stored in the network card cache region to a disk falling cache region according to the adjusted transfer rate; distributing corresponding hard disk identifiers for the network flow data packets in the disk-falling cache region, and writing the network flow data packets into corresponding hard disks in batches; according to the invention, the operation stability and reliability in a high-load or performance heterogeneous environment can be enhanced.

In general, the subject matter of this disclosure relates to a system for optimally balanced network multipathing. An example system includes a fabric adapter communicatively coupled to a plurality of network ports, a plurality of controlling hosts, and a plurality of application hosts. The fabric adapter is configured to: receive contents of one or more headers of a network packet; receive control information for the network packet, the control information describing at least one of a state of a network or a state of an application running on at least one application host from the plurality of application hosts; receive a payload of a network packet; select a network path in the network; and forward the network packet to a network port of the plurality of network ports, the network packet including the one or more headers and the payload.

The application discloses a concrete crack lightweight detection model and a detection method, and the detection model comprises three parts of a feature extraction backbone network, a feature fusion network and a regression prediction network. The backbone network and the feature fusion network are alternately connected by four LGSE modules and four C3-LGSE modules respectively, so that the feature extraction of the concrete crack and the information fusion of the multi-dimensional features are realized; and the regression prediction network is responsible for outputting the detection result of the concrete crack. The lightweight convolution module LGS can balance the network parameters and the detection accuracy, enhance the inter-group information exchange of the feature maps, obtain more features by using fewer feature maps, the channel attention module ECA can further improve the detection accuracy without increasing the parameter amount, and the LGSE module with the simultaneous action of the two can realize the lightweight and high-precision detection at the same time. The application is helpful to realize the lightweight and high-precision detection of the concrete crack, and has important significance for realizing the real-time detection embedded in a low-computing-power platform.

The present disclosure relates to a traffic control method, device, equipment and program product. The traffic control method applied to a client gateway module comprises: determining a target service in response to a data request sent by a client service module; determining a target time window corresponding to the target service; wherein the target time window is determined based on a target traffic model corresponding to the target service, and the target traffic model is used to balance network traffic; and sending the data request to the server gateway module within the target time window to request target data of the target service from a server service module corresponding to the server gateway module. In this way, the data requests of multiple clients can be more evenly dispersed, the network traffic can be better flattened, and traffic peaks and server peak capacity can be avoided, thereby reducing the network cost of data resource requests and the resource consumption of the server.

Various configurations of modular data center systems are described to enable a rapidly deployable data center that can be scaled to the demand and deployed more efficiently with significantly less field construction labor requirements. The systems include combinations of any number of separate IT modules, energy storage modules and transformer modules. A control and monitoring system monitors power demand of the IT equipment of the IT module, controlling power flow between the IT equipment of the IT module and the energy storage module, and balancing the network data load in response to compute load variations experienced by the IT equipment. In another configuration, a single module may include all IT equipment, energy storage and transformers / power conversion subsystems necessary for it to serve as a standalone data center unit.

The application provides a method and device for adjusting configuration parameters of network devices, aiming at balancing performance and energy consumption of a network system, so as to reduce energy consumption of the network system while maintaining normal forwarding of traffic. The method for adjusting configuration parameters of network devices comprises: obtaining network topology of the network system and information of predicted traffic of the network system, determining multiple traffic distribution relationships of the predicted traffic according to the network topology, selecting a traffic distribution relationship from the multiple traffic distribution relationships according to network performance and energy consumption corresponding to the predicted traffic of the network system based on the multiple traffic distribution relationships, and instructing at least one network device in the network system to apply configuration parameters corresponding to the selected traffic distribution relationship.

The invention discloses an output driving circuit. The output driving circuit comprises an input unit, a multi-stage output unit, a load unit, a feedback network, a bias network and a balance network, the input unit, the multi-stage output unit and the load unit are sequentially connected, the feedback network is connected with the output of the multi-stage output unit, the bias network is connected with the input unit and the multi-stage output unit, and the balance network is connected with the multi-stage output unit. The output driving circuit has the advantages of being large in output swing, high in output bandwidth and good in reliability.