POE switch and resource allocation method

Abstract

The invention discloses a POE (Power Over Ethernet) switch and a resource allocation method, belongs to the technical field of communication resource deployment, and solves the problem that an existing method does not dynamically integrate dynamic factors such as equipment load characteristics, service types and emergency levels, so that an allocation strategy is single. Generating a dynamic to-be-allocated queue in combination with the equipment coding priority and the demand emergency level, determining a residual power allocation value in combination with the real-time power information by taking the operation fluency as a constraint, constructing a multi-target optimization model based on the residual power allocation value and the dynamic to-be-allocated queue, and solving the target allocation power of the switch port power receiving equipment; according to the method and the device, comprehensive analysis is performed by combining the multi-dimensional characteristics of the port power receiving equipment, so that the resource demand of the equipment is comprehensively evaluated, and the dynamic to-be-allocated queue is generated by combining the equipment coding priority and the demand emergency level, so that the balanced and reasonable allocation of the POE switch is realized, and the resource utilization rate is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of communication resource deployment, and particularly relates to a POE switch and a resource allocation method. BACKGROUND

[0002] With the rapid development of information technology, network communication is becoming increasingly important in various fields. Among numerous network devices, POE (Power over Ethernet) switches have been widely used due to their unique advantages. POE switches can provide power to terminal devices (such as IP cameras, wireless access points, etc.) through Ethernet lines while transmitting data signals, greatly simplifying the wiring and installation process of devices, reducing deployment costs, and improving the flexibility and scalability of the system Existing POE switches have some problems in resource allocation. Traditional POE switches usually use fixed power allocation or simple on-demand allocation strategies, i.e., allocating the same power value to each port. Although the above strategies are simple, they have many limitations in practical applications. On the one hand, different types of terminal devices have different power requirements, and the traditional power allocation strategy is difficult to match dynamic load requirements; on the other hand, different types of POE powered devices (PD, Powered Device) have significant differences in power supply protocols, voltage stability, and surge tolerance. Existing POE switches rely on the LLDP message or resistance voltage division detection of PD devices to identify device types, but misjudgments may occur in complex scenarios.

[0003] Chinese patent CN113949637B discloses a switch resource intelligent allocation deployment method and device, which can obtain resource application information including network access demand configuration information; generate a machine room resource allocation scheme corresponding to the network access demand configuration information, wherein the machine room resource allocation scheme includes cabinet allocation information; if there is no physical machine resource pool model name identifier in the cabinet allocation information, determine whether there is a switch resource in the target cabinet corresponding to the cabinet allocation information, if not, generate a first switch resource allocation result for allocating switch resources to the target cabinet according to the network access demand configuration information and the preset physical machine network access deployment strategy; and perform network resource deployment based on the first switch resource allocation result; however, the existing method generates an allocation scheme based only on the network access demand configuration information and the physical machine resource pool model name identifier when allocating resources, without dynamically integrating device load characteristics, business types, and emergency level dynamic factors, resulting in a single allocation strategy that is difficult to adapt to complex scenarios. To address the above problems, we propose a POE switch and a resource allocation method. SUMMARY

[0004] The POE switch and resource allocation method solve the problem that the existing method generates an allocation scheme only based on network access demand configuration information and a physical machine resource pool model name identifier, without dynamically fusing device load characteristics, service types, and emergency level dynamic factors, resulting in a single allocation strategy.

[0005] The POE switch resource allocation method comprises the following steps: In response to a resource allocation instruction, a unified identification code of a powered device of a switch port is collected in real time, and device load data is indexed based on the unified identification code, wherein the device load data comprises power consumption demand information, starting current, voltage fluctuation value, link discovery protocol, service type, load characteristic initial sampling, voltage division resistance, and environmental parameter. The device load data is identified and analyzed to generate a load analysis result, a dynamic allocation queue is generated in combination with device code priority and demand emergency level, and the dynamic allocation queue is encapsulated and uploaded based on a TCP / IP protocol stack. Real-time running parameters of the switch are obtained, real-time power occupation information in the real-time running parameters of the switch is identified, and the running fluency of the switch is evaluated based on the real-time running parameters, so as to determine a residual power allocation value in combination with real-time power information and running fluency as a constraint. The residual power allocation value and the dynamic allocation queue are loaded in real time, a multi-objective optimization model is constructed based on the residual power allocation value and the dynamic allocation queue, and the target allocation power of the powered device of the switch port is solved.

[0006] A protocol adaptation model based on deep learning combined with a particle optimization algorithm is pre-constructed, the load analysis result is obtained, the link discovery protocol and the voltage division resistance in the load analysis result are identified, the protocol adaptation model verifies the link discovery protocol and the voltage division resistance twice, the device access protocol is determined based on the double verification result, and a device access instruction is generated, and in response to the device access instruction, a switch port powered device resource allocation task is performed.

[0007] Preferably, the method for generating a dynamic allocation queue in combination with device code priority and demand emergency level comprises the following steps: The unified identification code is parsed based on regular matching combined with a field segmentation strategy, a device code field value is obtained, and the device code field value is mapped to a priority weight table of a protocol database. A real-time weight table is generated based on the unified identification code, and it is judged whether the real-time weight table exists for the switch port powered device corresponding to the unified identification code. If the real-time weight table exists for the switch port powered device corresponding to the unified identification code, the device code priority corresponding to the switch port powered device is output. If the power receiving device corresponding to the uniform identification code does not exist in the real-time weight table, a similarity algorithm is used to determine a similar priority of the power receiving device of the switch port, and the similar priority is used as a device code priority corresponding to the power receiving device of the switch port; The load analysis result is obtained, a load multi-dimensional feature index is extracted based on the load analysis result, an emergency level of the multi-dimensional feature is quantified through a preset fuzzy logic rule, the emergency levels of the multi-dimensional features are weighted and fused to obtain a demand emergency level corresponding to the power receiving device of the switch port; A multi-dimensional weighting strategy combining the device code priority and the demand emergency level is defined, a comprehensive ranking value of the power receiving device of the switch port is calculated based on the multi-dimensional weighting strategy, and an initial allocation queue is generated in combination with the comprehensive ranking value of the power receiving device of the switch port; The initial allocation queue is loaded, the power receiving devices of the switch port with the same comprehensive ranking value are dynamically adjusted based on a conflict detection mechanism, and a dynamic allocation queue is generated, and the dynamic allocation queue is encapsulated and uploaded based on a TCP / IP protocol stack.

[0008] Preferably, the method for determining the similar priority of the power receiving device of the switch port based on the similarity algorithm comprises: The service type, load feature initial sampling, and power consumption demand information in the load analysis result are identified, service feature vectors, load feature vectors, and power consumption feature vectors corresponding to the service type, load feature initial sampling, and power consumption demand information are extracted and integrated into a service-load-power consumption feature set; The service-load-power consumption feature set is loaded, the feature vectors in the service-load-power consumption feature set are normalized by using a maximum-minimum normalization method, a combined similarity algorithm is used to calculate the combined similarity of the normalized service-load-power consumption feature set, and K devices with the highest combined similarity in the protocol database are determined as candidate similar devices; The combined similarity is calculated by the following formula: wherein, The combined similarity, the service similarity, the load similarity, and the power consumption similarity are respectively, The service feature weight coefficient, the load feature weight coefficient, and the power consumption feature weight coefficient are respectively determined based on an entropy weight method, The standard device startup current peak value and the switch port power receiving device startup current peak value are respectively, The standard device startup maximum power consumption and the switch port power receiving device startup maximum power consumption are respectively, respectively represent the maximum allowable difference of current, the maximum allowable difference of power consumption, respectively represent the switch port powered device, the standard device; The device coding priority consistency of the candidate similar device is checked, and if the device coding priority of the candidate similar device is consistent, the device coding priority of the candidate similar device is inherited as the similar priority of the switch port powered device. If the device coding priority of the candidate similar device is inconsistent, the device coding priority of the candidate similar device is weightedly averaged, and the average priority after the weighted averaging is taken as the similar priority of the switch port powered device.

[0009] Preferably, the method for determining the residual power allocation value by taking the running fluency as a constraint and combining real-time power information comprises: Loading the real-time running parameters of the switch, reading the utilization rate parameters, delay parameters and packet loss rate parameters in the real-time running parameters of the switch; Based on the parameter sampling period, the utilization rate curve, the delay curve and the packet loss rate curve are obtained, and the utilization rate curve, the delay curve and the packet loss rate curve are integrated into a parameter curve set; Identifying extreme points in the parameter curve set, and using a high-order statistical analysis method to statistically analyze the difference mean between the extreme points and the extreme points before and after the sampling time, and normalizing the difference mean, so that the difference mean is taken as the first fluency coefficient of the switch; The slope mean of the parameter slope between adjacent sampling time is calculated, and the slope mean of the parameter slope is taken as the second fluency coefficient of the switch; The first fluency coefficient and the second fluency coefficient are obtained, and the running fluency of the switch is evaluated based on the reward function of reinforcement learning combined with the first fluency coefficient and the second fluency coefficient; Wherein, the running fluency of the switch is calculated by the following formula: Wherein, represents the running fluency of the switch, respectively represent the first fluency coefficient and the second fluency coefficient, and is the reward function, represents the sampling interval time, is the penalty term of the reward function; The load vacancy rate at the sampling time in the real-time running parameters of the switch is analyzed by the principal component analysis method, and the vacancy rate time sequence queue is obtained, and the average vacancy rate of the vacancy rate time sequence queue is taken as the fluency adjustment weight; The vacancy rate time sequence queue is combined with multiple sampling time points to obtain multiple vacancy clustering clusters, the Euclidean distance between the vacancy clustering cluster and the ideal vacancy rate is analyzed, and the vacancy clustering cluster corresponding to the minimum Euclidean distance is selected as the constraint vacancy rate; The switch running smoothness, smoothness adjustment weight, and constraint vacancy rate are obtained, and the residual power distribution value is determined by taking the switch running smoothness, smoothness adjustment weight, and constraint vacancy rate as constraints. The residual power distribution value calculation formula is represented as: Wherein, The residual power distribution value, the switch theoretical total power, and the real-time power occupancy value are respectively, The smoothness adjustment weight and the constraint vacancy rate are respectively.

[0010] Preferably, the method for constructing a multi-objective optimization model based on the residual power distribution value and the dynamic to-be-allocated queue comprises: The constraint conditions of the multi-objective optimization model are defined as the single-port device power distribution, the constraint vacancy rate, and the port device load peak value, and a multi-objective optimization function of the multi-objective optimization model is constructed by taking the single-port device power distribution, the constraint vacancy rate, and the port device load peak value as constraints. The multi-objective optimization function is represented as: Wherein, The multi-objective optimization function is represented as, The power redundancy weight coefficient and the queue sorting weight coefficient searched based on the genetic algorithm are respectively, The single-port device maximum distribution power, the single-port device distribution power constraint, and the residual power distribution value are respectively, The comprehensive sorting value corresponding to the port device, the total number of devices, and the constraint vacancy rate are respectively. The multi-objective optimization function is solved by taking the minimum power redundancy and the maximum comprehensive sorting value as the optimization objectives of the multi-objective optimization function by combining the genetic algorithm, and the Pareto optimal solution of the multi-objective optimization function in the feasible solution space is searched. The Pareto optimal solution is taken as the target distribution power of the switch port powered device, and the target distribution power of the switch port powered device is output.

[0011] Preferably, the method for determining the device access protocol based on the double verification result comprises: The load analysis result is obtained, and the link discovery protocol and the voltage division resistance in the load analysis result are identified. The double verification module in the protocol adaptation model performs double verification on the link discovery protocol and the voltage division resistance to generate a double verification result. In response to the dual verification result, and based on the adaptive kernel density estimation algorithm, the link discovery protocol and the voltage division resistance are jointly probabilistically modeled to obtain a link protocol density function; The link protocol density function is parallelly decomposed to obtain a protocol response feature vector and an impedance response feature vector, and the protocol response feature vector and the impedance response feature vector are multi-scale fused based on a perception-enhanced Transformer network to obtain multi-scale fusion features; An interaction relationship between the multi-scale fusion features and standard link protocols in a protocol database is established, the top M groups of standard link protocols with high compatibility in the protocol database are captured by a lightweight identification module, and the standard link protocol with the highest compatibility is taken as an initial access protocol; The standard link protocol with the highest compatibility is taken as the initial access protocol, and other standard link protocols are searched by a particle optimization algorithm to optimize the initial access protocol and output a device access protocol; When the protocol adaptation model based on deep learning combined with the particle optimization algorithm is pre-constructed, a support vector machine model is taken as a basic framework, the basic framework is connected with an input layer and an output layer, a dual verification module is introduced between the support vector machine model and the input layer, the dual verification module is used for dual verification of the link discovery protocol and the voltage division resistance, the support vector machine model is used for joint probabilistic modeling of the link discovery protocol and the voltage division resistance based on the adaptive kernel density estimation algorithm to obtain a link protocol density function, a parallel space joint attention module is set between the support vector machine and the output layer, a bidirectional feature pyramid and a perception-enhanced Transformer network are introduced into the parallel decomposition attention module, the parallel decomposition attention module is used for parallel decomposition of the link protocol density function to obtain a protocol response feature vector and an impedance response feature vector, and the protocol response feature vector and the impedance response feature vector are multi-scale fused based on the perception-enhanced Transformer network to obtain multi-scale fusion features, an interaction relationship between the multi-scale fusion features and standard link protocols in a protocol database is established, a lightweight identification module is introduced between the parallel decomposition attention module and the output layer, the lightweight identification module is a spine-leaf architecture based on a multi-head self-attention mechanism, and the lightweight identification module is used for capturing the top M groups of standard link protocols with high compatibility in the protocol database, taking the standard link protocol with the highest compatibility as the initial access protocol, and searching other standard link protocols by the particle optimization algorithm to optimize the initial access protocol and output a device access protocol.

[0012] Preferably, the dual verification module in the protocol adaptation model performs dual verification on the link discovery protocol and the voltage division resistance by the method comprising: Identify the link discovery protocol, voltage division resistance, index the voltage division resistance, traverse the protocol database, classify the voltage division resistance based on the fuzzy clustering algorithm, and obtain the voltage division resistance clustering center and the device category label corresponding to the clustering center; Based on the standard link protocol associated with the device category label corresponding to the clustering center, the consistency of the link discovery protocol is verified; If the standard link protocol associated with the device category label is consistent with the link discovery protocol, the standard link protocol associated with the device category label is used as the device access protocol; If the standard link protocol associated with the device category label is inconsistent with the link discovery protocol, feedback the double verification failure, take the double verification failure as the double verification result, and perform joint probability modeling on the link discovery protocol and the voltage division resistance based on the adaptive kernel density estimation algorithm.

[0013] On the other hand, the application also provides a POE switch, which comprises: The allocation response module acquires the uniform identification code of the powered device of the switch port in real time in response to the resource allocation instruction, and indexes the device load data based on the uniform identification code; The allocation queue generation module is used to identify and analyze the device load data, generate a load analysis result, generate a dynamic allocation queue in combination with the device code priority and the demand urgency level, and upload the dynamic allocation queue based on the TCP / IP protocol stack; The allocation value determination module is used to obtain real-time running parameters of the switch, identify real-time power occupation information in the real-time running parameters of the switch, and determine a residual power allocation value based on the real-time running parameters and the real-time power information, with the running fluency as a constraint; The target allocation power solving module is used to load the residual power allocation value and the dynamic allocation queue in real time, construct a multi-objective optimization model based on the residual power allocation value and the dynamic allocation queue, and solve the target allocation power of the powered device of the switch port; The allocation execution module is used to obtain the load analysis result, identify the link discovery protocol and the voltage division resistance in the load analysis result, perform double verification on the link discovery protocol and the voltage division resistance based on the protocol adaptation model, and determine the device access protocol based on the double verification result.

[0014] Preferably, the allocation execution module comprises: The adaptation model construction unit is used to pre-construct a protocol adaptation model based on deep learning combined with a particle optimization algorithm; The device verification unit is used to obtain the load analysis result, identify the link discovery protocol and the voltage division resistance in the load analysis result, and perform double verification on the link discovery protocol and the voltage division resistance based on the protocol adaptation model; The access protocol determination unit determines a device access protocol based on the dual verification result, and generates a device access instruction; The resource allocation unit executes a switch port powered device resource allocation task in response to the device access instruction.

[0015] Compared with the prior art, the embodiments of the present application have the following beneficial effects: In the embodiments of the present application, by comprehensively analyzing the multi-dimensional characteristics of the port powered device, the resource demand of the device is comprehensively evaluated, and then a dynamic allocation queue is generated in combination with the device coding priority and the demand urgency level, so that the resource allocation demand of high priority services is guaranteed while the resource utilization rate is maximized, thereby realizing balanced and reasonable POE switch allocation and improving the resource utilization rate.

[0016] In the embodiments of the present application, when generating a dynamic allocation queue in combination with the device coding priority and the demand urgency level, the problem that the newly accessed device may not exist in the real-time weight table can be considered, and the similar priority of the switch port powered device is determined based on the similarity algorithm, so that the dynamic change of the network environment can be adapted to, resource allocation lag or unreasonableness caused by static priority can be avoided, the timeliness of resource allocation is improved, the allocation result is more suitable for the actual demand of the device, resource waste is reduced, and the resource utilization rate is improved.

[0017] In the embodiments of the present application, when determining the similar priority of the switch port powered device based on the similarity algorithm, by extracting the service type, the load feature initial sampling, the power consumption demand information corresponding to the service feature vector, the load feature vector and the power consumption feature vector, the problem of priority misjudgment caused by the traditional method of relying on only a single feature for judgment can be avoided, and the integration into the service-load-power feature set can provide a unified data basis for subsequent similarity calculation, ensure the comparability of different device features, and determine a reasonable priority for newly accessed or non-standard devices through feature similarity matching without a device pre-registration protocol library, thereby expanding the device compatibility of the POE switch, and through multi-dimensional feature fusion and similarity calculation, the one-sidedness of a single feature is avoided, priority verification and conflict processing ensure the reliability of the result, and resource allocation errors are reduced.

[0018] In the embodiments of the present application, when determining the residual power allocation value in combination with the running fluency as a constraint and real-time power information, by the first fluency coefficient, the second fluency coefficient and the constraint vacancy rate, the fluency evaluation can be dynamically adjusted according to real-time data to ensure the timeliness of the evaluation result, avoid system collapse or performance degradation caused by resource overload or idling, and through reinforcement learning and clustering analysis, the evaluation model and the constraint condition can be automatically adjusted, so that the switch resource allocation can adapt to different service loads and network environments. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 A POE switch resource allocation method implementation flowchart is shown.

[0020] Figure 2 A method implementation flowchart for generating a dynamic allocation queue in combination with device coding priority and demand urgency level is shown.

[0021] Figure 3 A method implementation flowchart for determining the similarity priority of the powered device of the switch port based on the similarity algorithm is shown.

[0022] Figure 4 A method implementation flowchart for determining the residual power allocation value in combination with real-time power information as a constraint is shown.

[0023] Figure 5 A method implementation flowchart for constructing a multi-objective optimization model based on the residual power allocation value and the dynamic allocation queue is shown.

[0024] Figure 6 A method implementation flowchart for determining the device access protocol based on the dual verification result is shown.

[0025] Figure 7 A method implementation flowchart for the dual verification module in the protocol adaptation model to perform dual verification on the link discovery protocol and the voltage division is shown.

[0026] Figure 8 The figure is a schematic diagram of the POE switch architecture provided by the present application. DETAILED DESCRIPTION

[0027] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used in the description herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the description and claims of this application as well as the above abstract are intended to cover all alternatives, modifications, and equivalents of the application falling within the scope of the application; the terms "comprising", "having", "including", and "containing" used herein are meant to be broad and encompass the terms "consisting of", "consisting essentially of", and "substantially consisting of". The terms "first", "second", and the like, used in the description and in the claims, are used for distinguishing between similar objects and are not necessarily used in a sequential order.

[0028] The existing method generates an allocation scheme only based on network access demand configuration information and a physical machine resource pool model name identifier when allocating resources, without dynamically fusing device load characteristics, service types, and emergency level dynamic factors, resulting in a single allocation strategy that is difficult to adapt to complex scenarios. To address the above problems, a POE switch and resource allocation method is proposed. In short, when the method is implemented, first, in response to a resource allocation instruction, the unified identification code of the power supply equipment of the switch port is collected in real time, the device load data is indexed based on the unified identification code, then the dynamic allocation queue is generated based on the device code priority and the demand emergency level, the dynamic allocation queue is encapsulated and uploaded based on the TCP / IP protocol stack, then the real-time power occupation information in the real-time running parameters of the switch is identified, and the running smoothness of the switch is evaluated based on the real-time running parameters, the remaining power allocation value is determined based on the running smoothness as a constraint and the real-time power information, a multi-objective optimization model is constructed based on the remaining power allocation value and the dynamic allocation queue, the target allocation power of the power supply equipment of the switch port is solved, a protocol adaptation model based on deep learning combined with a particle optimization algorithm is pre-constructed, and finally the protocol adaptation model verifies the link discovery protocol and the voltage division resistance, and determines the device access protocol based on the double verification results. In the embodiment of the application, the multi-dimensional characteristics of the port power supply equipment are combined for comprehensive analysis, so as to comprehensively evaluate the resource demand of the equipment, and then the dynamic allocation queue is generated based on the device code priority and the demand emergency level, so as to maximize the resource utilization rate while guaranteeing the resource allocation demand of high-priority services, so as to realize balanced and reasonable POE switch allocation, and improve the resource utilization rate.

[0029] The embodiment of the application provides a POE switch resource allocation method, Figure 1 The POE switch resource allocation method is shown in a flowchart, and the POE switch resource allocation method specifically comprises: S10, in response to a resource allocation instruction, the unified identification code of the power supply equipment of the switch port is collected in real time, and the device load data is indexed based on the unified identification code, wherein the device load data includes but is not limited to power consumption demand information, starting current, voltage fluctuation value, link discovery protocol, service type, load characteristic initial sampling, voltage division resistance and environmental parameters; It should be noted that the resource allocation instruction is a specific command or request for triggering or guiding the switch to perform the resource allocation task, and is usually generated by an upper management system (such as a network management platform, an automatic operation and maintenance system) or a control logic of the switch itself, and the unified identification code is a unique identifier allocated to each powered device, which is used for uniquely identifying and managing the device in the POE switch system. The unified identification code can contain various information of the device, such as device type, location, function, etc., so that the system can quickly and accurately index and manage the device, and in the embodiment, the structure of the unified identification code can be type code-management domain-device ID-priority field. The powered device of the switch port includes but is not limited to a wireless access point, an IP camera, a network printer, a networking sensor, a display, a lighting device, and an industrial computer.

[0030] S20, identifying and analyzing the device load data to generate a load analysis result, generating a dynamic allocation queue in combination with the device code priority and the demand urgency level, and encapsulating and uploading the dynamic allocation queue based on the TCP / IP protocol stack; S30, obtaining real-time running parameters of the switch, identifying real-time power occupation information in the real-time running parameters of the switch, and evaluating the running fluency of the switch based on the real-time running parameters, to determine a residual power allocation value in combination with the real-time power information and the running fluency as a constraint; S40, loading the residual power allocation value and the dynamic allocation queue in real time, constructing a multi-objective optimization model based on the residual power allocation value and the dynamic allocation queue, and solving the target allocation power of the powered device of the switch port.

[0031] S50, pre-constructing a protocol adaptation model based on deep learning combined with a particle optimization algorithm, obtaining the load analysis result, identifying the link discovery protocol and the voltage division resistance in the load analysis result, double verifying the link discovery protocol and the voltage division resistance by the protocol adaptation model, determining the device access protocol based on the double verification result, and generating a device access instruction, and in response to the device access instruction, performing a resource allocation task of the powered device of the switch port.

[0032] In the embodiment of the application, the multi-dimensional characteristics of the powered device of the port are combined for comprehensive analysis, so as to comprehensively evaluate the resource demand of the device, and then generate a dynamic allocation queue in combination with the device code priority and the demand urgency level, thereby maximizing the resource utilization rate while guaranteeing the resource allocation demand of high-priority services, so as to realize balanced and reasonable POE switch allocation and improve the resource utilization rate.

[0033] The embodiment of the application provides a method for generating a dynamic allocation queue in combination with a device code priority and a demand urgency level, Figure 2This diagram illustrates the implementation flow of the method for generating a dynamic queue to be assigned by combining device coding priority and demand urgency level. The method specifically includes: S101, based on regular expression matching combined with field segmentation strategy, parse the unified identifier code, obtain the device code field value, and map the device code field value to the priority weight table of the protocol database; It should be noted that parsing the unified identifier code based on regular expression matching combined with field segmentation strategies can accurately obtain the device code field value and map it to the priority weight table in the protocol database. This precise parsing method ensures the accurate acquisition of device information and provides a reliable basis for subsequent priority determination.

[0034] S102, Generate a real-time weight table based on the unified identifier code, and determine whether there is a switch port powered device corresponding to the unified identifier code in the real-time weight table; In this embodiment, considering that the newly accessed device may not have a real-time weight table, in order to avoid allocation errors caused by the lack of priority of non-standard devices (without protocol library records), the present invention determines the similar priority of the powered devices on the switch port based on the similarity algorithm. This can adapt to the dynamic changes in the network environment, avoid resource allocation delays or unreasonableness caused by static priorities, and improve the timeliness of resource allocation.

[0035] S103, If there is a switch port powered device with a unified identifier code in the real-time weight table, output the device code priority corresponding to the switch port powered device. S104. If there is no switch port powered device corresponding to the unified identifier code in the real-time weight table, determine the similarity priority of the switch port powered device based on the similarity algorithm, and use the similarity priority as the device code priority corresponding to the switch port powered device. S105, Obtain the load analysis results, extract multi-dimensional feature indicators of the load based on the load analysis results, quantify the urgency level of the multi-dimensional features through preset fuzzy logic rules, and weight and fuse the urgency levels of the multi-dimensional features to obtain the demand urgency level corresponding to the powered device of the switch port. It should be noted that the method for weighted fusion of the urgency levels of multi-dimensional features can be average weighting or exponential weighting, while the urgency levels of multi-dimensional features can be quantified by pre-set fuzzy logic rules through principal component analysis or expert consultation.

[0036] S106 defines a multi-dimensional weighting strategy that combines device coding priority and demand urgency level. Based on the multi-dimensional weighting strategy, the comprehensive ranking value of the powered devices on the switch port is calculated, and the initial queue to be assigned is generated by combining the comprehensive ranking value of the powered devices on the switch port. S107: Load the initial queue to be allocated, dynamically adjust the powered devices on the switch ports with the same comprehensive sorting value based on the conflict detection mechanism, generate a dynamic queue to be allocated, and encapsulate and upload the dynamic queue to be allocated based on the TCP / IP protocol stack.

[0037] In this embodiment, when dynamically adjusting the power-receiving devices of switch ports with the same comprehensive ranking value based on the conflict detection mechanism, when the comprehensive ranking values ​​of the power-receiving devices of switch ports are the same, priority is considered first, followed by urgency level, so as to ensure that high-priority and high-urgency devices are ranked at the front of the queue.

[0038] In this embodiment of the invention, when generating a dynamic queue to be allocated by combining device coding priority and demand urgency level, it can take into account the problem that newly accessed devices may not have a real-time weight table. Furthermore, by determining the similarity priority of the powered devices on the switch port based on the similarity algorithm, it can adapt to the dynamic changes in the network environment, avoid resource allocation delays or unreasonableness caused by static priorities, improve the timeliness of resource allocation, make the allocation results more in line with the actual needs of the devices, reduce resource waste, and improve resource utilization.

[0039] This invention provides a method for determining the similarity priority of powered devices on a switch port based on a similarity algorithm. Figure 3 This diagram illustrates the implementation flow of the method for determining the similarity priority of powered devices on a switch port based on a similarity algorithm. The method specifically includes: S1041, Identify the service type, initial load feature sampling, and power consumption requirement information in the load parsing results, extract the service feature vector, load feature vector, and power consumption feature vector corresponding to the service type, initial load feature sampling, and power consumption requirement information, and integrate them into a service-load-power consumption feature set; In this embodiment of the invention, by extracting the service feature vector, load feature vector and power consumption feature vector corresponding to the service type, initial load feature and power consumption requirement information, the problem of priority misjudgment caused by relying on only a single feature for judgment in traditional methods can be avoided. Moreover, integrating them into a service-load-power consumption feature set can provide a unified data foundation for subsequent similarity calculation and ensure the comparability of features of different devices.

[0040] S1042, Load the service-load-power consumption feature set, use the maximum-minimum normalization method to normalize the feature vectors in the service-load-power consumption feature set, and perform combination similarity calculation on the normalized service-load-power consumption feature set based on the combination similarity algorithm to determine the K devices with the highest combination similarity in the protocol database as candidate similar devices. The combination similarity is calculated using the following formula: in, These are similarity in terms of combination, business logic, load, and power consumption. These are the weighting coefficients for service, load, and power consumption characteristics determined based on the entropy weighting method. These are the peak startup current of standard equipment and the peak startup current of the powered devices at the switch ports. These are the maximum power consumption for standard device startup and the maximum power consumption for powered devices on switch ports, respectively. These are the maximum allowable difference in current and the maximum allowable difference in power consumption, respectively. These respectively represent the powered device and the standard device on the switch port; S1043, Perform device code priority consistency check on candidate similar devices to determine whether the device code priorities of candidate similar devices are consistent; S1044, if the device code priority of the candidate similar devices is consistent, the device code priority of the candidate similar devices is inherited as the similar priority of the switch port powered device; in this embodiment, the priority inherited by the switch port powered device is consistent with other similar devices in the protocol library, avoiding priority "isolation" caused by the target device being a newly accessed device, and ensuring that the resource allocation strategy of similar devices in the network is unified.

[0041] S1045, if the device code priorities of the candidate similar devices are inconsistent, the device code priorities of the candidate similar devices are weighted and averaged, and the average priority after weighting and averaging is used as the similarity priority of the power-receiving device of the switch port.

[0042] In this embodiment, when candidate devices cause priority conflicts due to historical data errors or device upgrades, a weighted average is used to integrate information from all parties to avoid the problem of priority deviation caused by extreme values.

[0043] In this embodiment of the invention, when determining the similarity priority of powered devices on a switch port based on a similarity algorithm, the extraction of service feature vectors, load feature vectors, and power consumption feature vectors corresponding to service type, initial load feature sampling, and power consumption requirement information avoids the problem of misjudgment of priority caused by relying on only a single feature in traditional methods. Furthermore, integrating these into a service-load-power consumption feature set provides a unified data foundation for subsequent similarity calculations, ensuring the comparability of features of different devices. Moreover, it eliminates the need for pre-registering protocol libraries for devices, and reasonable priorities can be determined for newly accessed or non-standard devices through feature similarity matching, expanding the device compatibility of PoE switches. In addition, the multi-dimensional feature fusion and combined similarity calculation avoid the one-sidedness of a single feature, and priority verification and conflict handling ensure the reliability of the results and reduce resource allocation errors.

[0044] This invention provides a method for determining the remaining power allocation value by combining operational smoothness as a constraint with real-time power information. Figure 4 The diagram illustrates the implementation flow of the method for determining the remaining power allocation value by combining operational smoothness as a constraint with real-time power information. Specifically, this method includes: S201 loads the switch's real-time operating parameters and reads the utilization, latency, and packet error rate parameters from these parameters. The utilization parameters include, but are not limited to, CPU utilization and memory utilization. The latency parameters include, but are not limited to, port queue latency and protocol stack processing latency. The packet error rate parameters include, but are not limited to, the error packet rate. These parameters cover three dimensions: computing resources (CPU / memory), forwarding performance (latency), and reliability (packet error rate), comprehensively reflecting the switch's health status and avoiding the bias caused by a single indicator. S202, based on the parameter sampling period, obtain the utilization curve, delay curve and packet error rate curve, and integrate the utilization curve, delay curve and packet error rate curve into a parameter curve set; S203 identifies the extreme points of the parameter curve and uses a high-order statistical analysis method to calculate the mean difference between the extreme points and the extreme points at the previous and next sampling times. The mean difference is then normalized and used as the first smoothness coefficient of the switch. The extreme points reflect the drastic fluctuations of the parameters, and the mean difference quantifies the amplitude and frequency of the fluctuations. After normalization, it can intuitively reflect the operating stability of the switch. S204, calculate the average slope of the parameter slope between adjacent sampling times, and use the average slope of the parameter slope as the second smoothness coefficient of the switch. S205, obtain the first smoothness coefficient and the second smoothness coefficient, and evaluate the smoothness of switch operation based on the reward function of reinforcement learning combined with the first smoothness coefficient and the second smoothness coefficient; The smoothness of switch operation is calculated using the following formula: in, Indicates the smoothness of switch operation. These are the first and second smoothness coefficients, respectively. For the reward function, Indicates the sampling interval time. This is the penalty term for the reward function; S206 uses principal component analysis to perform time-series analysis of the load idle rate at the sampling time in the real-time operating parameters of the switch to obtain the idle rate time-series queue. The average idle rate of the idle rate time-series queue is used as the smoothness adjustment weight. The average idle rate reflects the idle degree of the switch resources. After being used as the idle rate adjustment weight, the power allocation strategy can be dynamically adjusted. S207, the vacancy rate time sequence queue is combined to cluster multiple sampling times to obtain multiple vacancy clusters. The Euclidean distance between the vacancy clusters and the ideal vacancy rate is analyzed, and the vacancy cluster corresponding to the minimum Euclidean distance is selected as the constraint vacancy rate. In this embodiment, the cluster closest to the ideal vacancy rate is selected by Euclidean distance to ensure that the vacancy rate after allocation will not deviate from the ideal value, thereby improving the robustness of resource allocation. S208: Obtain the switch's operating smoothness, smoothness adjustment weight, and constraint vacancy rate, and determine the remaining power allocation value using the switch's operating smoothness, smoothness adjustment weight, and constraint vacancy rate as constraints. The formula for calculating the remaining power allocation value is as follows: in, These represent the remaining power allocation value, the theoretical total power of the switch, and the real-time power occupancy value, respectively. These are the smoothness adjustment weight and the constraint vacancy rate, respectively.

[0045] In this embodiment of the invention, when determining the remaining power allocation value by combining operational smoothness as a constraint with real-time power information, the smoothness evaluation can be dynamically adjusted based on real-time data through the first smoothness coefficient, the second smoothness coefficient, and the constraint vacancy rate, ensuring the timeliness of the evaluation results and avoiding system crashes or performance degradation caused by resource overload or idleness. At the same time, through reinforcement learning and cluster analysis, the evaluation model and constraints can be automatically adjusted, enabling the switch resource allocation to adapt to different service loads and network environments.

[0046] This invention provides a method for constructing a multi-objective optimization model based on remaining power allocation values ​​and a dynamic queue of unallocated power. Figure 5This diagram illustrates the implementation flow of the method for constructing a multi-objective optimization model based on remaining power allocation values ​​and a dynamic queue of unallocated power. The method specifically includes: S301 defines the constraints of the multi-objective optimization model as power allocation to a single-port device, idle rate, and peak load of the port device. A multi-objective optimization function is constructed using these constraints. It's important to note that the multi-objective function clearly defines the optimization direction—to allocate as much power as possible while prioritizing the needs of high-priority devices. Compared to traditional single-objective optimization, the multi-objective function better aligns with the real-world need to balance efficiency and business priority.

[0047] The multi-objective optimization function is expressed as: in, Represents a multi-objective optimization function. These are the power redundancy weight coefficients based on genetic algorithm search and the queue sorting weight coefficients, respectively. These represent the maximum power allocated to a single port device, the power allocation constraint for a single port device, and the remaining power allocation value, respectively. These are the overall ranking value, total number of devices, and constraint vacancy rate for the port devices, respectively. S302 combines a genetic algorithm to solve a multi-objective optimization function with the optimization objectives of minimizing power redundancy and maximizing the overall ranking value. It then searches for the Pareto optimal solution of the multi-objective optimization function in the feasible solution space. It should be noted that traditional optimization methods have difficulty handling multi-objective conflicts, while the genetic algorithm, by simulating the global search mechanism of natural selection, can find multiple Pareto optimal solutions in the feasible solution space, covering the balance point of different objective weights.

[0048] S303 uses the Pareto optimal solution as the target power allocation for the powered devices at the switch port and outputs the target power allocation for the powered devices at the switch port.

[0049] In this embodiment of the invention, constraints such as single-port power, vacancy rate, and peak load are used to avoid problems such as device overload and link congestion caused by resource allocation, ensuring stable system operation. A multi-objective optimization function is constructed using single-port device allocation power, constrained vacancy rate, and peak load of port devices as constraints. It should be noted that the multi-objective function clearly defines the optimization direction—to allocate as much power as possible while prioritizing the needs of high-priority devices. Compared to traditional single-objective optimization, the multi-objective function better aligns with the need for both efficiency and business priority in real-world scenarios. Finally, a genetic algorithm is used to search for a Pareto optimal solution, ensuring a global balance between the allocation results and the multi-objectives, avoiding resource waste or business impairment caused by local optima.

[0050] This invention provides a method for determining device access protocols based on dual authentication results. Figure 6 The diagram illustrates the implementation flow of the method for determining the device access protocol based on the dual authentication result. The method specifically includes: S401, Obtain the load analysis results and identify the link discovery protocol and voltage resistance in the load analysis results; where the link discovery protocol is the basic protocol for device communication (LLDP is used for neighbor discovery), and voltage resistance is the electrical characteristic of the physical link (such as impedance matching). Together, they constitute the identity and physical constraints of device access, providing clear input objects for subsequent dual verification. In this embodiment, when determining the device access protocol, both logical protocol (link discovery) and physical layer characteristics (voltage resistance) can be considered simultaneously, avoiding physical incompatibility problems caused by only verifying the logical protocol.

[0051] In S402, the dual verification module in the protocol adaptation model performs dual verification on the link discovery protocol and voltage divider, generating a dual verification result. In this embodiment of the invention, it is considered that verifying only the link discovery protocol may overlook physical layer incompatibility, and verifying only the voltage divider may ignore logical protocol errors. Dual verification, through a combination of logical and physical checks, significantly improves the legitimacy of access and avoids device damage or communication failure caused by single verification. The dual verification result can directly exclude incompatible devices, thereby reducing invalid calculations in subsequent processing and improving resource allocation efficiency.

[0052] S403 responds to the dual verification results and performs joint probability modeling of link discovery protocol and voltage resistance based on the adaptive kernel density estimation algorithm to obtain the link protocol density function. By adaptively adjusting the kernel function width, it dynamically fits the local density of the data and more accurately describes the joint distribution of "protocol type-voltage resistance". S404 decomposes the link protocol density function in parallel to obtain the protocol response feature vector and impedance response feature vector. Based on the perceptual enhancement Transformer network, the protocol response feature vector and impedance response feature vector are fused at multiple scales to obtain multi-scale fused features. The Transformer network captures the long-distance dependence between features through the self-attention mechanism, thereby generating a more comprehensive feature representation through multi-scale fusion and enhancing the accuracy of subsequent protocol identification. S405 establishes the interaction relationship between multi-scale fusion features and standard link protocols in the protocol database. A lightweight identification module captures the top M groups of standard link protocols with high compatibility in the protocol database, and uses the standard link protocol with the highest compatibility as the initial access protocol. The spine-leaf architecture, through the hierarchical design of "spine" and "leaf", can quickly filter out standard protocols that highly match the current features, avoiding the problem of inefficient calculation by traversing the entire database.

[0053] S406 uses the most compatible standard link protocol as the initial access protocol and then uses a particle optimization algorithm to search for other standard link protocols to optimize the initial access protocol, outputting the device access protocol. The particle optimization algorithm, through swarm intelligence search, finds the globally optimal protocol in the feasible solution space, avoiding insufficient adaptability caused by local optima in the initial selection, thus ensuring that the final protocol is always the optimal solution.

[0054] In this embodiment of the invention, when determining the device access protocol based on the dual verification results, it is considered that verifying only the link discovery protocol may overlook physical layer incompatibility, and verifying only voltage resistance may ignore logical protocol errors. Dual verification, through a combination of logical and physical checks, significantly improves the legitimacy of access and avoids device damage or communication failure caused by single verification. The dual verification results can directly exclude incompatible devices, thereby reducing unnecessary calculations in subsequent processing, improving resource allocation efficiency, and utilizing the spine-leaf architecture with its layered design of "spine" and "leaf" to quickly filter out standard protocols that highly match the current characteristics, avoiding the inefficient calculation problem of traversing the entire database.

[0055] It should be noted that the protocol database is a system / repository for storing, managing and organizing metadata and rules related to network communication protocols. It is the core infrastructure for network devices (such as switches, routers and terminals) to achieve protocol identification, verification, adaptation and interaction.

[0056] In this embodiment, when pre-constructing the protocol adaptation model based on deep learning combined with particle optimization algorithm, a support vector machine (SVM) model is used as the basic architecture. This architecture connects to the input layer and the output layer. A dual verification module is introduced between the SVM model and the input layer to perform dual verification of the link discovery protocol and voltage resistance. The SVM model uses an adaptive kernel density estimation algorithm to perform joint probability modeling of the link discovery protocol and voltage resistance, obtaining the link protocol density function. A parallel spatial joint concern module is set between the SVM and the output layer. This parallel decomposition concern module introduces a bidirectional feature pyramid and a perceptual enhancement Transformer network. The parallel decomposition concern module is used to decompose the link protocol density function in parallel to obtain the protocol response features. The protocol adaptation model uses a support vector machine (SVM) as its basic architecture, integrating a dual verification module, adaptive kernel density estimation, a parallel spatial joint attention module, a perceptual enhancement Transformer network, and a lightweight recognition module. This results in a multi-scale fused feature, which establishes an interaction relationship between the multi-scale fused feature and standard link protocols in the protocol database. A lightweight recognition module is introduced between the parallel decomposition attention module and the output layer. This module employs a spine-leaf architecture based on a multi-head self-attention mechanism. It extracts the top M groups of standard link protocols with high compatibility from the protocol database, using the most compatible standard link protocol as the initial access protocol. Furthermore, it searches for other standard link protocols using a particle optimization algorithm to optimize the initial access protocol and output the device access protocol. In this embodiment, the protocol adaptation model is based on a support vector machine (SVM) architecture, integrating a dual verification module, adaptive kernel density estimation, a parallel spatial joint attention module, a perceptual enhancement Transformer network, and a lightweight recognition module. The model can adapt to the changes in the "protocol-voltage resistance" relationship under different scenarios by dynamically adjusting the kernel function width according to the data distribution. Furthermore, through parallel decomposition and multi-scale fusion of Transformer networks, it can efficiently process high-dimensional features, avoid the curse of dimensionality, and achieve efficient, accurate, and adaptive determination of device access protocols in complex network environments. This provides key technical support for the intelligent access of network devices such as PoE switches.

[0057] This invention provides a method for dual verification of link discovery protocols and voltage impedance matching in a protocol adaptation model using a dual verification module. Figure 7 This diagram illustrates the implementation flow of the dual verification module in the protocol adaptation model, which performs dual verification on the link discovery protocol and voltage blocking. Specifically, the dual verification method in the protocol adaptation model includes: S4021, Identify the link discovery protocol and voltage divider. Using the voltage divider as an index, traverse the protocol database and classify the voltage dividers based on the fuzzy clustering algorithm to obtain the voltage divider cluster center and the device category label corresponding to the cluster center. The clustering result is directly associated with the standard device type in the protocol database, which can provide a clear "physical layer identity basis" for the subsequent verification of the link discovery protocol and avoid label errors caused by voltage divider fluctuations. S4022, perform consistency verification on the link discovery protocol based on the standard link protocol associated with the device category label corresponding to the cluster center; S4023, If the standard link protocol associated with the device category label is consistent with the link discovery protocol, the standard link protocol associated with the device category label shall be used as the device access protocol; S4024 If the standard link protocol associated with the device category label is inconsistent with the link discovery protocol, feedback indicates that the dual verification failed. The dual verification failure is taken as the dual verification result, and joint probability modeling of the link discovery protocol and voltage divider is performed based on the adaptive kernel density estimation algorithm.

[0058] In this embodiment, traditional verification only checks whether the link discovery protocol conforms to the standard, but may overlook the matching between the physical layer device type and the protocol. This invention ensures that the logical protocol and the physical type of the device are consistent through the association verification of "physical layer labels combined with standard protocols", avoiding communication failures caused by protocol mismatch.

[0059] On the other hand, embodiments of the present invention also provide a PoE switch. Figure 8 The diagram shows the architecture of the PoE switch, which specifically includes: The allocation response module 100 responds to resource allocation commands by collecting the unified identification codes of the powered devices on the switch ports in real time and indexing the device load data based on the unified identification codes. The allocation queue generation module 200 is used to identify and parse the equipment load data, generate load parsing results, and generate a dynamic queue to be allocated based on the equipment coding priority and the urgency level of the demand. The dynamic queue to be allocated is encapsulated and uploaded based on the TCP / IP protocol stack. The allocation value determination module 300 is used to obtain the real-time operating parameters of the switch, identify the real-time power occupancy information in the real-time operating parameters of the switch, evaluate the smoothness of the switch operation based on the real-time operating parameters, and determine the remaining power allocation value by combining the smoothness of operation with the real-time power information. The target power allocation solution module 400 is used to load the remaining power allocation value and the dynamic queue to be allocated in real time, and to build a multi-objective optimization model based on the remaining power allocation value and the dynamic queue to be allocated to solve the target power allocation of the powered devices at the switch port. The allocation execution module 500 is used to obtain the load parsing results, identify the link discovery protocol and voltage divider in the load parsing results, and perform dual verification of the link discovery protocol and voltage divider using the protocol adaptation model. Based on the dual verification results, the device access protocol is determined.

[0060] In this embodiment, the allocation execution module 500 includes: The adaptation model building unit 510 is used to pre-build a protocol adaptation model based on deep learning combined with particle optimization algorithm; The device verification unit 520 is used to obtain the load parsing results, identify the link discovery protocol and voltage blocking in the load parsing results, and verify the link discovery protocol and voltage blocking through the protocol adaptation model. The access protocol determination unit 530 determines the device access protocol based on the dual verification results and generates a device access command; Resource allocation unit 540, in response to device access command, performs resource allocation tasks for powered devices on switch ports.

[0061] It should be noted that the steps of the POE switch provided in this embodiment of the invention correspond to the steps of the above-described POE switch resource allocation method, and will not be repeated here.

[0062] In summary, this invention provides a PoE switch and a resource allocation method. In this embodiment, by combining the multi-dimensional characteristics of the port-powered devices for comprehensive analysis, the resource requirements of the devices are fully evaluated. Then, a dynamic queue to be allocated is generated by combining the device coding priority and the urgency level of the requirements. This maximizes resource utilization while ensuring the resource allocation needs of high-priority services, thereby achieving a balanced and reasonable allocation of resources on the PoE switch and improving resource utilization.

[0063] It should be noted that, for the sake of simplicity, the foregoing embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to the present invention. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0064] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on these embodiments, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art can still combine, add, delete, or otherwise adjust the features of the various embodiments of the present invention according to the circumstances without conflict or creative effort, thereby obtaining different technical solutions that do not fundamentally depart from the concept of the present invention. These technical solutions also fall within the scope of protection of the present invention.

Claims

1. A method for allocating resources on a PoE switch, characterized in that, The method includes: In response to resource allocation commands, the system collects the unified identification codes of powered devices on the switch ports in real time and indexes device load data based on the unified identification codes. The system identifies and analyzes equipment load data, generates load analysis results, and generates a dynamic queue to be allocated by combining equipment coding priority and demand urgency level. The dynamic queue to be allocated is then encapsulated and uploaded based on the TCP / IP protocol stack. Obtain the real-time operating parameters of the switch, identify the real-time power occupancy information in the real-time operating parameters of the switch, evaluate the smoothness of the switch operation based on the real-time operating parameters, and determine the remaining power allocation value by combining the smoothness of operation with the real-time power information. The remaining power allocation value and the dynamic queue to be allocated are loaded in real time. Based on the remaining power allocation value and the dynamic queue to be allocated, a multi-objective optimization model is constructed to solve the target allocation power of the powered devices at the switch port.

2. The PoE switch resource allocation method as described in claim 1, characterized in that: The method further includes: A protocol adaptation model based on deep learning combined with particle optimization algorithm is pre-built to obtain the load parsing results. The link discovery protocol and voltage blocking are identified in the load parsing results. The protocol adaptation model performs dual verification of the link discovery protocol and voltage blocking. Based on the dual verification results, the device access protocol is determined and the device access command is generated. In response to the device access command, the switch port powered device resource allocation task is executed.

3. The PoE switch resource allocation method as described in claim 2, characterized in that: The method for generating a dynamic queue to be assigned by combining device coding priority and demand urgency level includes: The unified identifier code is parsed based on regular expression matching combined with field segmentation strategy to obtain the device code field value, and the device code field value is mapped to the priority weight table of the protocol database. A real-time weight table is generated based on the unified identifier code, and it is determined whether there is a switch port powered device corresponding to the unified identifier code in the real-time weight table. If the switch port powered device has a unified identifier code in the real-time weight table, output the priority of the device code corresponding to the switch port powered device; If there is no switch port powered device corresponding to the unified identifier code in the real-time weight table, the similarity priority of the switch port powered device is determined based on the similarity algorithm, and the similarity priority is used as the device code priority corresponding to the switch port powered device. Obtain the load analysis results, extract multi-dimensional feature indicators of the load based on the load analysis results, quantify the urgency level of the multi-dimensional features through preset fuzzy logic rules, and weight and fuse the urgency levels of the multi-dimensional features to obtain the demand urgency level corresponding to the powered device of the switch port. Define a multi-dimensional weighting strategy that combines device coding priority and demand urgency level. Calculate the comprehensive ranking value of the powered devices on the switch port based on the multi-dimensional weighting strategy, and generate an initial queue to be assigned based on the comprehensive ranking value of the powered devices on the switch port. The initial queue to be assigned is loaded. Based on the collision detection mechanism, the powered devices on the switch ports with the same comprehensive sorting value are dynamically adjusted, and a dynamic queue to be assigned is generated. The dynamic queue to be assigned is encapsulated and uploaded based on the TCP / IP protocol stack.

4. The PoE switch resource allocation method as described in claim 3, characterized in that: The method for determining the similarity priority of powered devices on a switch port based on a similarity algorithm includes: Identify the service type, initial load feature sampling, and power consumption requirement information in the load parsing results, extract the service feature vector, load feature vector, and power consumption feature vector corresponding to the service type, initial load feature sampling, and power consumption requirement information, and integrate them into a service-load-power consumption feature set; Load the service-load-power feature set, use the maximum-minimum normalization method to normalize the feature vectors in the service-load-power feature set, and calculate the combined similarity of the normalized service-load-power feature set based on the combined similarity algorithm to determine the K devices with the highest combined similarity in the protocol database as candidate similar devices. Perform device code priority consistency verification on candidate similar devices. If the device code priorities of candidate similar devices are consistent, inherit the device code priority of candidate similar devices as the similarity priority of the switch port powered device. If the device code priorities of candidate similar devices are inconsistent, the device code priorities of the candidate similar devices are weighted and averaged, and the average priority after weighting and averaging is used as the similarity priority of the power-receiving device on the switch port.

5. The PoE switch resource allocation method as described in claim 1, characterized in that: The method for determining the remaining power allocation value by combining operational smoothness as a constraint with real-time power information includes: Load the switch's real-time operating parameters and read the utilization, latency, and packet error rate parameters from the switch's real-time operating parameters. The utilization curve, delay curve, and packet error rate curve are obtained based on the parameter sampling period, and the utilization curve, delay curve, and packet error rate curve are integrated into a parameter curve set. Identify the extreme points of the parameter curve, and use high-order statistical analysis methods to calculate the mean difference between the extreme points and the extreme points at the previous and next sampling times. Then, normalize the mean difference and use the mean difference as the first smoothness coefficient of the switch. Calculate the average slope of the parameter slope between adjacent sampling times, and use the average slope of the parameter slope as the second smoothness coefficient of the switch. Obtain the first smoothness coefficient and the second smoothness coefficient, and evaluate the smoothness of switch operation based on the reward function of reinforcement learning combined with the first smoothness coefficient and the second smoothness coefficient; Principal component analysis was used to perform time series analysis of the load idle rate at the sampling time in the real-time operating parameters of the switch, and the idle rate time series queue was obtained. The average idle rate of the idle rate time series queue was used as the smoothness adjustment weight. By combining the vacancy rate time series queue, multiple sampling times are clustered to obtain multiple vacancy clusters. The Euclidean distance between the vacancy clusters and the ideal vacancy rate is analyzed, and the vacancy cluster corresponding to the minimum Euclidean distance is selected as the constraint vacancy rate. Obtain the switch's operating smoothness, smoothness adjustment weight, and constraint vacancy rate, and use these factors to determine the remaining power allocation value.

6. The PoE switch resource allocation method as described in claim 5, characterized in that: The method for constructing a multi-objective optimization model based on remaining power allocation values ​​and a dynamic queue of unallocated power includes: The constraints of the multi-objective optimization model are defined as the power allocated to a single-port device, the vacancy rate, and the peak load of the port device. The multi-objective optimization function of the multi-objective optimization model is constructed with the power allocated to a single-port device, the vacancy rate, and the peak load of the port device as constraints. The multi-objective optimization function is solved by combining the genetic algorithm with the optimization objectives of minimizing power redundancy and maximizing the comprehensive ranking value. The Pareto optimal solution of the multi-objective optimization function is then searched in the feasible solution space. Using the Pareto optimal solution as the target power allocation for the powered devices at the switch port, output the target power allocation for the powered devices at the switch port.

7. The PoE switch resource allocation method as described in claim 2, characterized in that: The method for determining the device access protocol based on the dual authentication result includes: Obtain the load analysis results and identify the link discovery protocol and voltage blocking in the load analysis results; In the protocol adaptation model, the dual verification module performs dual verification on the link discovery protocol and voltage blocking, and generates dual verification results. In response to the dual verification results, and based on the adaptive kernel density estimation algorithm, joint probability modeling of the link discovery protocol and voltage blocking is performed to obtain the link protocol density function; The link protocol density function is decomposed in parallel to obtain the protocol response feature vector and impedance response feature vector. The protocol response feature vector and impedance response feature vector are then fused at multiple scales based on the perceptual enhancement Transformer network to obtain multi-scale fused features. Establish the interaction relationship between multi-scale fusion features and standard link protocols in the protocol database. Use a lightweight identification module to capture the top M groups of standard link protocols with high compatibility in the protocol database, and use the standard link protocol with the highest compatibility as the initial access protocol. The initial access protocol is the standard link protocol with the highest compatibility. Other standard link protocols are searched using the particle optimization algorithm to optimize the initial access protocol and output the device access protocol.

8. The PoE switch resource allocation method as described in claim 7, characterized in that: The dual verification module in the protocol adaptation model performs dual verification of the link discovery protocol and voltage divider, including: The link discovery protocol and voltage divider are identified. Using the voltage divider as an index, the protocol database is traversed, and the voltage divider is classified according to the fuzzy clustering algorithm to obtain the voltage divider cluster center and the device category label corresponding to the cluster center. The link discovery protocol is validated for consistency based on the standard link protocol associated with the device category label corresponding to the cluster center. If the standard link protocol associated with the device category label is consistent with the link discovery protocol, the standard link protocol associated with the device category label shall be used as the device access protocol. If the standard link protocol associated with the device category label is inconsistent with the link discovery protocol, the double verification fails. The double verification failure is taken as the double verification result, and the link discovery protocol and voltage divider are jointly probabilistically modeled based on the adaptive kernel density estimation algorithm.

9. A PoE switch for implementing the PoE switch resource allocation method as described in any one of claims 1-8, characterized in that: The PoE switch includes: The allocation response module responds to resource allocation commands by collecting the unified identification codes of the powered devices on the switch ports in real time and indexing the device load data based on the unified identification codes. The allocation queue generation module is used to identify and parse equipment load data, generate load parsing results, and generate a dynamic queue to be allocated based on equipment coding priority and demand urgency level. The dynamic queue to be allocated is encapsulated and uploaded based on the TCP / IP protocol stack. The allocation value determination module is used to obtain the real-time operating parameters of the switch, identify the real-time power occupancy information in the real-time operating parameters of the switch, evaluate the smoothness of the switch operation based on the real-time operating parameters, and determine the remaining power allocation value by combining the smoothness of operation with the real-time power information. The target power allocation solution module is used to load the remaining power allocation value and the dynamic queue to be allocated in real time, and to build a multi-objective optimization model based on the remaining power allocation value and the dynamic queue to be allocated to solve the target power allocation of the powered devices at the switch port. The allocation execution module is used to obtain the load parsing results, identify the link discovery protocol and voltage divider in the load parsing results, and perform dual verification of the link discovery protocol and voltage divider in the protocol adaptation model. Based on the dual verification results, the device access protocol is determined.

10. The PoE switch as described in claim 9, characterized in that: The allocation execution module includes: The adaptation model building unit is used to pre-build protocol adaptation models based on deep learning combined with particle optimization algorithms; The device verification unit is used to obtain the load analysis results, identify the link discovery protocol and voltage blocking in the load analysis results, and verify the link discovery protocol and voltage blocking through the protocol adaptation model. The access protocol determination unit determines the device access protocol based on the dual verification results and generates a device access command; The resource allocation unit, in response to the device access command, performs the task of allocating resources to powered devices on the switch port.

Images