A method, device and equipment for detecting power supply of POE power supply

Abstract

The invention relates to the technical field of POE power supply, and discloses a power supply detection method, device and equipment of a POE power supply. The method comprises: collecting voltage data and current data of a POE power supply system to perform characteristic curve fitting operation, and obtain a power supply parameter sensitivity matrix; forming a multi-level detection system with a voltage detection unit, a current detection unit and a power detection unit; establishing a coupled equation group including a voltage equation, a current equation and a power equation and performing numerical solution to obtain a power supply state parameter matrix; inputting a dual-loop power supply control model, performing abnormality detection through an inner-loop response network and an outer-loop control network in the dual-loop power supply control, and obtaining a target abnormality type; performing multi-level judgment operation based on the target abnormality type, generating a power supply control instruction sequence, and executing power supply protection control. The invention realizes accurate prediction of the power supply state, and effectively reduces the probability of power supply accidents.

Description

Technical Field

[0001] The present invention relates to the technical field of POE power supply, and in particular to a power supply detection method, device and equipment of a POE power supply. Background Art

[0002] The POE power supply system is widely used in the field of network equipment power supply due to its advantage of being able to transmit data and power at the same time. However, the POE power supply system faces challenges in power supply stability and reliability during actual operation, especially when the load changes dynamically, voltage fluctuations and current mutations occur frequently, which seriously affects the normal operation and service life of the equipment.

[0003] Traditional POE power supply detection methods mainly rely on simple threshold judgment and linear analysis models, which cannot effectively deal with complex power supply abnormalities. These methods have large errors when dealing with nonlinear load characteristics, and the detection response speed is slow, making it difficult to detect and deal with power supply abnormalities in a timely manner, resulting in overvoltage, overcurrent and other faults in the power supply system. Summary of the invention

[0004] The present invention provides a POE power supply detection method, device and equipment, which realizes accurate prediction of power supply status and effectively reduces the probability of power supply accidents.

[0005] In a first aspect, the present invention provides a POE power supply detection method, the POE power supply detection method comprising: Collect voltage data and current data of the POE power supply system, perform characteristic curve fitting operation on the voltage data and current data to obtain a voltage-current characteristic curve, and calculate the first-order derivative and second-order derivative of the voltage-current characteristic curve to obtain a power supply parameter sensitivity matrix; Constructing a voltage detection unit, a current detection unit and a power detection unit according to the power supply parameter sensitivity matrix, and connecting the voltage detection unit, the current detection unit and the power detection unit through a data bus to form a multi-level detection system; Collecting real-time detection data through the multi-level detection system, establishing a coupled equation group including a voltage equation, a current equation and a power equation, and numerically solving the coupled equation group to obtain a power supply state parameter matrix; Inputting the power supply state parameter matrix into a dual-loop power supply control model, performing anomaly detection through an inner-loop response network and an outer-loop control network in the dual-loop power supply control, and obtaining a target anomaly type; A multi-level judgment operation is performed based on the target abnormality type to obtain a hierarchical alarm control strategy, and a power supply control instruction sequence is generated according to the hierarchical alarm control strategy to execute power supply protection control.

[0006] In a second aspect, the present invention provides a power supply detection device for a POE power supply, the power supply detection device for the POE power supply comprising: The acquisition module is used to collect voltage data and current data of the POE power supply system, perform characteristic curve fitting operation on the voltage data and current data to obtain a voltage-current characteristic curve, and calculate the first-order derivative and second-order derivative of the voltage-current characteristic curve to obtain a power supply parameter sensitivity matrix; A construction module, used to construct a voltage detection unit, a current detection unit and a power detection unit according to the power supply parameter sensitivity matrix, and connect the voltage detection unit, the current detection unit and the power detection unit through a data bus to form a multi-level detection system; A solution module, used for collecting real-time detection data through the multi-level detection system, establishing a coupled equation group including a voltage equation, a current equation and a power equation, and numerically solving the coupled equation group to obtain a power supply state parameter matrix; A detection module, used for inputting the power supply state parameter matrix into a dual-loop power supply control model, performing anomaly detection through an inner-loop response network and an outer-loop control network in the dual-loop power supply control, and obtaining a target anomaly type; The execution module is used to perform multi-level judgment operations based on the target abnormality type to obtain a hierarchical alarm control strategy, and generate a power supply control instruction sequence according to the hierarchical alarm control strategy to execute power supply protection control.

[0007] The third aspect of the present invention provides a computer device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the computer device executes the above-mentioned POE power supply detection method.

[0008] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned POE power supply detection method.

[0009] In the technical solution provided by the present invention, by establishing a voltage-current characteristic curve and calculating a sensitivity matrix, an accurate description of the electrical characteristics of the POE power supply system is achieved, a multi-level detection system structure is adopted, and detection units in three dimensions of voltage, current and power are combined to improve the integrity and accuracy of the detection data; a dual-loop power supply control model is introduced, and the detection sensitivity and response speed of the system to power supply anomalies are enhanced through the synergy of inner loop fast response and outer loop state tracking; a numerical solution method based on a coupled equation group is used to solve the nonlinear coupling problem between power supply parameters, so that the detection result is more accurate; a deep neural network is used for anomaly classification, combined with a multi-level judgment mechanism, to improve the accuracy of anomaly type identification; an autoregressive model and trend analysis are used to achieve accurate prediction of the power supply status, providing a basis for preventive maintenance; a complete hierarchical alarm and protection control strategy is designed to effectively reduce the probability of power supply accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0011] Figure 1 Schematic diagram of the steps of the power supply detection method of the POE power supply in an embodiment of the present invention; Figure 2 Schematic diagram of the structure of a POE power supply detection device in an embodiment of the present invention; Figure 3 It is a schematic block diagram of the structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0012] Embodiments of the present invention provide a method, device and equipment for detecting power supply of a POE power supply. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0013] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , an embodiment of the power supply detection method of the POE power supply in the embodiment of the present invention includes: Step S1, collecting voltage data and current data of the POE power supply system, performing characteristic curve fitting operation on the voltage data and the current data to obtain a voltage-current characteristic curve, and calculating the first-order derivative and the second-order derivative of the voltage-current characteristic curve to obtain a power supply parameter sensitivity matrix; It is understandable that the execution subject of the present invention may be a power supply detection device of a POE power supply, or may be a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0014] Specifically, the voltage data and current data of the POE power supply system are collected. The POE power supply system is sampled under different load conditions to obtain a voltage sampling sequence and a current sampling sequence. The voltage sampling sequence and the current sampling sequence are digitally processed, and the sampling sequence is filtered, denoised and normalized to ensure the accuracy and reliability of the data, and a voltage data set and a current data set are obtained. Feature extraction and data fitting operations are performed on the voltage data set and the current data set to obtain a voltage-current characteristic curve. The relationship between voltage and current is expressed by a suitable mathematical model. The polynomial fitting method is adopted to fit the voltage and current data points to obtain a voltage-current characteristic polynomial equation to describe the nonlinear relationship between voltage and current. The coefficients of the voltage-current characteristic polynomial equation are normalized to eliminate the differences between different data scales, so that the characteristic curve equation can adapt to different working conditions and load conditions, and a standardized characteristic curve equation is obtained. Based on the standardized characteristic curve equation, the first-order derivative curve of the voltage-current relationship is calculated. By differentiating the voltage-current characteristic equation, a voltage-current sensitivity function is obtained, which reflects the degree of influence of voltage change on current and the responsiveness of the power supply under different loads and working conditions. The voltage-current sensitivity function is subjected to second-order differential operation to obtain the second-order sensitivity function of the voltage-current characteristic. The second-order sensitivity function reflects the quadratic interaction between voltage and current changes, and can reveal the dynamic response characteristics of the power supply when facing rapid load changes. The calculation of the second-order derivative helps to identify potential nonlinear problems, especially when the load changes drastically or the power supply is unstable. The second-order sensitivity function effectively captures these abnormal conditions. At the preset operating point, numerical calculations are performed based on the voltage-current sensitivity function and the second-order sensitivity function to generate the power supply parameter sensitivity matrix.

[0015] Step S2, constructing a voltage detection unit, a current detection unit and a power detection unit according to the power supply parameter sensitivity matrix, and connecting the voltage detection unit, the current detection unit and the power detection unit through a data bus to form a multi-level detection system; Specifically, according to the power supply parameter sensitivity matrix, the characteristic decomposition operation is performed to extract the sensitivity relationship between the various parameters in the power supply system, and the voltage sensitivity characteristic vector, the current sensitivity characteristic vector and the power sensitivity characteristic vector are obtained, which represent the change law of voltage, current and power under different loads and working conditions and their influence on the performance of the power supply system. The voltage sampling circuit is parameterized according to the voltage sensitivity characteristic vector. By optimizing the working parameters of the voltage sampling circuit, it is ensured that it can provide stable and accurate voltage data under various working conditions. After the voltage signal is collected, the analog signal is converted into a digital signal through an analog-to-digital converter, and the data is cached and stored to obtain a voltage detection unit. Based on the current sensitivity characteristic vector, the detection parameters of the Hall current sensor are set. The Hall current sensor plays an important role in POE power supply detection. It can sense the current size passing through the current conductor through the Hall effect. Setting appropriate detection parameters can ensure that the current signal is collected accurately and can reflect the changes in the power supply load in real time. The collected current signal is properly conditioned and amplified to form a signal suitable for subsequent analysis to obtain a current detection unit. The output data of the voltage detection unit and the output data of the current detection unit are input into the digital multiplier, and the real-time power calculation is performed through the digital multiplier to obtain the power detection unit. The sampling timing of the voltage detection unit, the current detection unit and the power detection unit is synchronously controlled according to the power sensitivity characteristic vector to generate a unified sampling synchronization signal to ensure that the output data of all detection units remain consistent in time and avoid data deviation caused by inconsistent sampling. Through the sampling synchronization signal, the voltage detection unit, the current detection unit and the power detection unit are connected through the data bus to form a multi-level detection system. The multi-level detection system can work together to monitor the voltage, current and power status of the POE power supply in real time, and ensure the accuracy and consistency of various data through synchronous control.

[0016] Step S3, collecting real-time detection data through a multi-level detection system, establishing a coupled equation group including a voltage equation, a current equation and a power equation, and numerically solving the coupled equation group to obtain a power supply state parameter matrix; Specifically, the multi-level detection system is controlled in time sequence based on the sampling synchronization signal to ensure that all detection units (voltage detection unit, current detection unit and power detection unit) can collect data at a unified time point to obtain a synchronous collection time series. According to the synchronous collection time series, the voltage detection unit, current detection unit and power detection unit are collected to obtain real-time detection data. The real-time detection data is digitally filtered and normalized to obtain pre-processed detection data. The purpose of digital filtering is to remove high-frequency noise in the signal, so that the acquisition results of voltage, current and power are smoother and more accurate. Normalization can eliminate the differences between different measurement units and data scales, so that the data are consistent and comparable. Based on the pre-processed data, the nonlinear state expressions of the voltage equation and the current equation are established to describe the variation of voltage and current under different loads and different working points. The power equation is constructed based on the voltage equation and the current equation, and the variation relationship of voltage and current is linked to the variation of power. Through these coupled equations, the behavior and state of the POE power system are described. The coupled equations are subjected to variable separation and linearization. Variable separation is to transform the coupled nonlinear equations into multiple relatively independent equations to simplify the calculation process. The linearization process converts the nonlinear equation into an approximate linear form by performing Taylor expansion or other linearization methods on the nonlinear equation, which is convenient for subsequent numerical calculations. After linearization, a linearization coefficient matrix is ​​obtained, which represents the linear relationship between voltage, current and power. The linearization coefficient matrix is ​​input into the iterative solver for numerical calculation. The solution method includes the least squares method, which minimizes the error by optimizing the solution process. The least squares method can effectively deal with the noise and errors in the data. Through the iterative optimization process, the solution vector of the coupled equations is obtained, which represents the numerical solution of parameters such as voltage, current and power. The solution vector of the equation system is analyzed by eigenvalue, and the key characteristic parameters in the solution vector are extracted to reveal the essence of the system behavior. The characteristic parameters include the stability of the system, the load response capability and the fault indication. Based on the characteristic parameter set, the state space transformation and matrix reconstruction are performed to obtain the power supply state parameter matrix. State space transformation refers to the conversion of the state of the system from the original parameter space to a new space that is more suitable for analysis. Matrix reconstruction is to reconstruct the power supply state matrix according to the transformed characteristic parameters to provide a higher level of power supply state information.

[0017] The coupled equations are separated by variables. The variables in the voltage equation, current equation, and power equation are distinguished according to linear terms and nonlinear terms. The relationship between voltage, current, and power is usually complex and contains nonlinear terms. These nonlinear terms are processed by appropriate mathematical means to make the equations easier to solve. In the process of variable separation, the linear terms and nonlinear terms of voltage, current, and power are extracted to form a separation variable matrix, which represents the relative importance of each parameter in each equation and their relationship. Based on the separation variable matrix, Taylor expansion is performed on the nonlinear terms. Taylor expansion is a method of approximating a nonlinear function as a polynomial. By expanding the function at a certain operating point, it can be approximated as a linear function. Taylor expansion expands the nonlinear terms in the voltage equation, current equation, and power equation into first-order, second-order, or higher-order polynomial terms. Since the influence of high-order terms on the system gradually decreases, the expansion is truncated to remove the high-order terms and retain only the first-order or second-order terms to obtain a linearized approximate equation group. The linearized approximate equation group is converted into a matrix form, and the linearized coefficient matrix is ​​obtained by the coefficient comparison method. The matrix represents the linear relationship between parameters such as voltage, current, and power. Condition number analysis is performed on the linearization coefficient matrix. The numerical stability of the matrix is ​​revealed, that is, whether the solution process of the matrix is ​​easily affected by errors. Matrices with smaller condition numbers are more stable in numerical calculations, while matrices with larger condition numbers lead to larger errors in the calculation results. Based on the condition number analysis, the Jacobian iteration matrix is ​​constructed. The Jacobian matrix is ​​an approximate matrix used for iterative solution. It can gradually approach the true solution by continuously updating the solution vector to obtain an iterative initial solution. The initial solution is input into the least squares iterative solver, and the least squares iterative operation is performed according to the set convergence threshold and maximum number of iterations. The least squares method is an optimization algorithm whose goal is to minimize the error in the system of equations and obtain the best approximate solution to the system of equations. In each iteration, the accuracy of the current solution is evaluated by calculating the residual vector. The residual vector represents the gap between the current solution and the actual solution. The smaller the residual, the more accurate the current solution. The residual vector is calculated after each iteration. As the number of iterations increases, the residual gradually decreases, and the solution vector gradually approaches the true solution. In order to ensure the effectiveness and accuracy of the solution process, the iterative sequence is judged for convergence. When the residual is less than the preset threshold, the system stops iterating and considers that the solution has converged, obtaining the final converged solution. The converged solution is normalized, the numerical value of the solution is mapped to a standard range, the differences between different dimensions and units are eliminated, and the solution vector of the system of equations is generated.

[0018] Step S4, inputting the power supply state parameter matrix into the dual-loop power supply control model, performing anomaly detection through the inner-loop response network and the outer-loop control network in the dual-loop power supply control, and obtaining the target anomaly type; Specifically, the power supply state parameter matrix is ​​analyzed according to the voltage parameter, current parameter and power parameter to obtain the inner loop detection data sequence and the outer loop control data sequence, which reflect the operation state and control effect of the power supply system. The inner loop detection data sequence is mainly used to monitor the changes of parameters such as voltage and current in real time, while the outer loop control data sequence focuses on the long-term operation trend and control strategy of the system. The inner loop detection data sequence is input into the inner loop response network of the dual-loop power supply control model for rapid response processing. The purpose of the inner loop response network is to respond quickly when an abnormality occurs in the power supply system, so as to adjust the power supply parameters in time. The response network consists of multiple layers, including a rapid response layer, a feature extraction layer and a state judgment layer. In the rapid response layer, data is efficiently collected at a preset sampling frequency to ensure that the voltage and current change information can be obtained in real time. The feature extraction layer uses wavelet transform for feature decomposition. Wavelet transform can effectively decompose the signal into components of different frequencies, thereby extracting important features in the signal and distinguishing abnormal fluctuations in voltage and current signals. In this way, potential abnormal patterns are identified to provide necessary information for subsequent judgment. The state judgment layer uses the threshold judgment method to detect anomalies, judge whether the voltage and current fluctuate beyond the preset threshold, so as to determine whether there is anomaly and generate a voltage-current anomaly feature sequence. At the same time, the outer loop control data sequence is input into the outer loop control network of the dual-loop power supply control model to track and calculate the power supply state. The outer loop control network is used to monitor the overall operation trend of the power supply system and help the system judge the power supply state changes over a long period of time. The outer loop control network includes a data cache layer, a trend analysis layer and a state prediction layer. In the data cache layer, historical data is stored in a circular queue to ensure that the historical data used by the system during processing is always the latest and most relevant data. The trend analysis layer uses a sliding window method to calculate trends. The sliding window technology can calculate the average trend of data in different time windows, so that the system can smoothly capture the changing trend of the power system state. In the state prediction layer, an autoregressive model is used to predict the future power supply state. The autoregressive model can predict the future state trend based on the changing law of historical data. The voltage-current anomaly feature sequence and the power supply state change trend are input into the feature fusion of the dual-loop power supply control model. The feature fusion includes a feature mapping unit and a decision unit. In the feature mapping unit, the voltage-current abnormal feature sequence and the power supply status change trend are converted into a unified feature space through the feature mapping process. This space can effectively integrate data from different sources into a comprehensive abnormal feature vector. The abnormal feature vector contains the abnormal feature information of the power system at the current moment and in the future. The generated abnormal feature vector is classified by deep neural network to identify the target abnormal type. Through multi-layer nonlinear mapping, deep neural network can efficiently extract meaningful patterns from complex data.The neural network contains three fully connected layers, each of which uses the ReLU activation function and Dropout regularization. The ReLU activation function can effectively improve the network training speed and solve the gradient vanishing problem, while the Dropout regularization helps prevent overfitting and enhance the generalization ability of the model. The output layer of the network uses the Softmax function for multi-classification processing, which enables the system to classify abnormal situations into different categories and help locate specific abnormal types, such as excessive voltage fluctuations, abnormal current, or insufficient power.

[0019] The outer loop control data sequence is input into the circular queue of the data cache layer. The length of the circular queue is set to 1024 data points to ensure that the system obtains enough historical data during the analysis process and can accommodate the control data of the most recent period. The circular queue adopts a first-in-first-out storage strategy to ensure that the data will not be missed during the continuous update process. Each storage operation is updated according to the setting of the read and write pointers. The old data will be automatically overwritten when the new data is written, realizing the real-time update and storage of historical data. In order to ensure the stability of the queue and prevent data overflow, an overflow processing mechanism is set to ensure that the data that is about to be lost can be correctly processed when the queue storage is full, and the integrity and continuity of the data are maintained. The historical data sequence is divided into time windows to obtain an overlapping time window sequence. The long data sequence is divided into multiple shorter time segments, each of which contains certain control data. The overlapping time window technology can better capture the local change trend in the data and avoid the loss of information caused by the data cutting process through overlapping windows. Based on the time window sequence, multi-dimensional statistical feature extraction is performed to extract various features of the data, such as mean, standard deviation, peak, skewness and kurtosis, to obtain statistical feature vectors, which help the model identify the change law of the power system and its potential abnormal patterns. The statistical feature vectors are grouped according to voltage features, current features and power features, and each group of features is input into the trend analysis layer in the outer loop control network. The trend analysis layer consists of three trend fitting units, corresponding to the characteristics of voltage, current and power respectively. Each trend fitting unit calculates the linear trend coefficient and quadratic trend coefficient of the corresponding group of features. These two coefficients reflect the change law of voltage, current and power in the power system in different time periods. The linear trend coefficient represents the linear change trend of the system state, while the quadratic trend coefficient can reveal more complex nonlinear change trends. By calculating these trend coefficients, the change trend of each group of features is obtained to form a complete trend feature sequence. The trend feature sequence is input into the state prediction layer in the outer loop control network. The state prediction layer uses the autoregressive model (AR model) for state prediction. The optimal order of the autoregressive model is determined by AIC (Akaike Information Criterion), and the AR model with an order of 8 (i.e., AR(8) model) is selected. The AIC criterion helps to select the model order that best suits the current data among many models. By comparing the goodness of fit and complexity of models of different orders, the optimal model is finally selected. The Levinson-Durbin recursive algorithm is used to calculate the autoregressive coefficients of the AR(8) model. These coefficients reflect the degree of influence of the system's historical state on the future state and can accurately predict the future voltage, current and power states. By using these autoregressive coefficients, the state prediction equation is constructed, and the predicted state sequence of the voltage state, current state and power state is calculated through recursive iteration, which represents the expected operating state of the system at the future moment.In order to evaluate the accuracy of the prediction results, the root mean square error (RMSE) of the prediction state sequence of 1 step, 4 steps, 8 steps and 16 steps is calculated respectively. The root mean square error is a standard for measuring the accuracy of prediction. An error value is obtained by calculating the difference between the predicted value and the actual value. The smaller the RMSE value, the more accurate the prediction. By calculating the root mean square error of the prediction results at different time scales, the prediction credibility of each time period is evaluated, and a corresponding prediction confidence score is given. Based on the prediction confidence score, the prediction results of different time scales are adaptively weighted and fused. The weighting coefficient is adjusted according to the prediction confidence score of each time scale, so that the prediction results with higher confidence have a larger proportion in the final fusion. The power supply status change trend is obtained through the adaptive weighted fusion method.

[0020] Step S5: Perform multi-level judgment operations based on the target abnormality type to obtain a hierarchical alarm control strategy, generate a power supply control instruction sequence according to the hierarchical alarm control strategy, and execute power supply protection control.

[0021] Specifically, the abnormalities are graded and judged according to the identified target abnormality types. Different abnormalities are graded and handled according to their severity and urgency, so as to minimize the impact of the fault on the power supply system without affecting the overall operation of the system. The abnormality type is compared with a set of preset judgment rules, and different abnormalities are divided into different severity levels according to the rules. The generated graded alarm control strategy is a specific action plan for the system to deal with different types of abnormalities. The strategy includes response measures for each alarm level and stipulates the timing and handling method of taking measures. The graded alarm control strategy needs to be dynamically adjusted in combination with the actual operating status of the power supply system, load demand and the specific nature of the abnormality to ensure the stability and security of the system. Based on the graded alarm control strategy, a power supply control instruction sequence is generated. The power supply control instruction sequence is the actual operation command for the system to execute protection control, and is adjusted accordingly according to different abnormality levels and strategies. The instruction sequence generates specific control operations according to the target abnormality type, the graded alarm strategy and the power supply status detected in real time, including voltage adjustment instructions, current limit instructions, power adjustment instructions or switching operation instructions. The control instructions are executed in sequence to ensure that each instruction can be responded to and executed at the right time. During this process, the power supply protection control mechanism monitors various parameters in real time to ensure that the system can continue to operate stably during the execution of instructions and adjust the operation strategy in a timely manner. During the execution process, new voltage, current and power data are obtained, and the control instructions are adjusted according to these real-time data to respond to changes in the system status. If a new anomaly occurs, the current power supply status is re-evaluated, and a new instruction sequence is generated according to the new anomaly type and severity to ensure that the system is always in the best working state.

[0022] In the embodiment of the present invention, by establishing a voltage-current characteristic curve and calculating a sensitivity matrix, an accurate description of the electrical characteristics of the POE power supply system is achieved, and a multi-level detection system structure is adopted, combining detection units of three dimensions of voltage, current and power, thereby improving the integrity and accuracy of the detection data; a dual-loop power supply control model is introduced, and the detection sensitivity and response speed of the system to power supply anomalies are enhanced through the synergy of inner loop fast response and outer loop state tracking; a numerical solution method based on a coupled equation group is used to solve the nonlinear coupling problem between power supply parameters, making the detection result more accurate; a deep neural network is used for anomaly classification, combined with a multi-level judgment mechanism, to improve the accuracy of anomaly type identification; an autoregressive model and trend analysis are used to achieve accurate prediction of the power supply status, providing a basis for preventive maintenance; a complete hierarchical alarm and protection control strategy is designed to effectively reduce the probability of power supply accidents.

[0023] In a specific embodiment, the process of executing step S1 may specifically include the following steps: The POE power supply system is sampled under different load conditions to obtain a voltage sampling sequence and a current sampling sequence, and the voltage sampling sequence and the current sampling sequence are digitally processed to obtain a voltage data set and a current data set; Perform feature extraction and data fitting operations on the voltage data set and the current data set to obtain a voltage-current characteristic polynomial equation, and perform coefficient normalization processing on the voltage-current characteristic polynomial equation to obtain a standardized characteristic curve equation; Based on the standardized characteristic curve equation, the first-order derivative curve of the voltage-current relationship is calculated to obtain the voltage-current sensitivity function, and the voltage-current sensitivity function is subjected to a second-order differential operation to obtain the second-order sensitivity function of the voltage-current characteristic; The voltage-current sensitivity function and the second-order sensitivity function are numerically calculated at a preset working point to generate a power supply parameter sensitivity matrix.

[0024] Specifically, the POE (Power Over Ethernet) power supply system provides power to devices through Ethernet cables while transmitting data. Sampling voltage and current helps analyze the stability and efficiency of the system. By sampling under different load conditions, we can have a more comprehensive understanding of the response characteristics of the power supply system, especially under changing load conditions. During the sampling process, the values ​​of voltage and current will fluctuate as the load changes. When collecting data for the POE power supply system, it is necessary to ensure that the sampling frequency is high enough to capture the details of these fluctuations. Assume that the voltage sequence is collected under different load conditions. and current sequence ,in is the number of sampling points. In order to ensure the validity and stability of these data, digital signal processing is performed to remove noise and improve the accuracy of the signal. Common digital signal processing techniques include filtering, smoothing and other methods. By filtering the original voltage and current data, the voltage data set is obtained. and current data sets , reflecting the actual working state of the system under different loads. Perform feature extraction and data fitting operations on the voltage data set and the current data set. Feature extraction is to identify the relationship between voltage and current and the law of their change with the load. In order to better describe the relationship between voltage and current, a voltage-current characteristic polynomial equation is constructed using polynomial fitting. Assume that the relationship between voltage and current is described by a quadratic polynomial, and its expression is: ; in, , and are the coefficients of the polynomial, calculated by the fitting algorithm. These coefficients are obtained by fitting using the least squares method. The goal of the least squares method is to minimize the fitting error, that is, to minimize the sum of squares of the following errors: ; By solving the equation, we get the fitted coefficients , and , and obtain the voltage-current characteristic polynomial equation. Normalize the coefficients of the voltage-current characteristic polynomial equation to obtain the standardized characteristic curve equation. For example, divide all coefficients by the highest voltage value. , thus standardizing the voltage-current relationship and obtaining the standardized equation: ; Based on the normalized characteristic curve equation, the first-order derivative of the voltage-current relationship is calculated. The voltage-current derivative represents the rate at which the current changes with the voltage, i.e., the voltage-current sensitivity function. By taking the derivative of the normalized characteristic curve equation, the voltage-current sensitivity function is obtained: ; It indicates the sensitivity of current to voltage change, indicating how the current changes accordingly when the voltage changes. Performing a second-order differential operation on the voltage-current sensitivity function, we get the second-order sensitivity function of the voltage-current characteristic. The second-order sensitivity function describes the acceleration of current change, that is, the acceleration of the effect of the voltage change rate on the current, and its expression is: ; The second-order sensitivity function can reveal the nonlinear characteristics of the system current response. When the second-order derivative is large, it means that the current is more sensitive to the change of voltage, and vice versa. The nonlinear response characteristics of the system under different voltage changes are analyzed by calculating the second-order sensitivity function. At the preset working point, the voltage-current sensitivity function and the second-order sensitivity function are numerically calculated to generate the power supply parameter sensitivity matrix. Assume that the voltage at the preset working point is , the current is , at this operating point, the sensitivity matrix is ​​calculated by numerical methods, where the sensitivity matrix contains the values ​​of the first-order and second-order derivatives. Assuming that the power supply parameter sensitivity matrix is ​​represented by S, it is obtained by the following relationship: ; The sensitivity matrix can provide key parameters for the regulation and optimization of the power supply system and help identify the response characteristics of the system under different load conditions.

[0025] In a specific embodiment, the process of executing step S2 may specifically include the following steps: Performing eigendecomposition operation on the power supply parameter sensitivity matrix to obtain a voltage sensitivity eigenvector, a current sensitivity eigenvector, and a power sensitivity eigenvector; The voltage sampling circuit is parameterized according to the voltage sensitivity characteristic vector, a voltage signal is collected through the voltage sampling circuit, and analog-to-digital conversion and data caching are performed on the voltage signal to obtain a voltage detection unit; The detection parameters of the Hall current sensor are set based on the current sensitivity characteristic vector, the current signal is collected through the Hall current sensor, and the current signal is conditioned and amplified to obtain a current detection unit; Inputting the output data of the voltage detection unit and the output data of the current detection unit into a digital multiplier, performing real-time power calculation through the digital multiplier, and obtaining a power detection unit; Performing sampling timing synchronization control on the voltage detection unit, the current detection unit and the power detection unit according to the power sensitivity characteristic vector to obtain a sampling synchronization signal; Based on the sampling synchronization signal, the voltage detection unit, the current detection unit and the power detection unit are connected through the data bus to form a multi-level detection system.

[0026] Specifically, when performing eigendecomposition, the power supply parameter sensitivity matrix is ​​composed of first-order derivatives and second-order derivatives, representing the mutual influence between voltage, current and power. Assume that the power supply parameter sensitivity matrix It is expressed as follows: ; in, It represents the sensitivity of current to voltage. Indicates the sensitivity of current to power, represents the sensitivity of power to voltage, and It represents the sensitivity of power to current. By performing eigendecomposition on the sensitivity matrix, the voltage sensitivity eigenvector is obtained: , current sensitivity characteristic vector and the power sensitivity eigenvector , thereby determining the parameter configuration and functional design of different detection units. The voltage sensitivity characteristic vector represents the impact of voltage changes on the overall performance of the power supply system. In the voltage sampling circuit, the parameters are configured according to the voltage sensitivity characteristic vector. By analyzing the voltage sensitivity characteristic vector, the gain, sampling frequency, and resolution of the analog-to-digital converter of the sampling circuit are configured to ensure accurate capture of the voltage signal. The output signal of the sampling circuit is digitized by the analog-to-digital converter to obtain a voltage data sequence. Assuming that the voltage is sampled and converted into a digital signal, the voltage sequence obtained is The voltage data is stored and cached through the data cache unit to obtain the output of the voltage detection unit. The detection parameters of the Hall current sensor are set based on the current sensitivity characteristic vector. The current sensitivity characteristic vector can reveal the impact of current changes on the overall performance of the system, thereby guiding the parameter configuration of the Hall current sensor. When designing the current detection unit, the sensitivity, gain, sampling frequency and other parameters of the Hall sensor are set based on the current sensitivity characteristic vector, and ensure that the current signal can be accurately collected. Collecting current signals through Hall sensors , and then these current signals are conditioned (such as amplification, filtering, etc.) to obtain the output data of the current detection unit , these data reflect the changes in the current in the system. The output data of the voltage detection unit and the output data of the current sensing unit Input into the digital multiplier, and the multiplier performs real-time power calculation. The power calculation formula is: ; in, and Represent the values ​​of voltage and current respectively. Assume that the output of the voltage detection unit is , the output of the current detection unit is , then the real-time calculation of power is expressed as: ; The power data obtained Indicates the power consumption of the system in the current working state. Through the calculation of the power detection unit, the real-time power data is passed to the subsequent processing unit. The sampling timing of the voltage detection unit, current detection unit and power detection unit is synchronously controlled. The power sensitivity characteristic vector represents the impact of power changes on system performance, which can help the system adjust the sampling timing to ensure that the sampling data of voltage, current and power are synchronized at the same time, thereby improving the accuracy of the measurement data. The sampling timing of each detection unit is adjusted according to the characteristic vector so that they can sample and output data at the same time. Assume that the synchronization signal is , then through synchronous control, all detection units collect data at the same time to obtain sampling synchronization signals. The voltage detection unit, current detection unit and power detection unit are connected together through the data bus to form a multi-level detection system. The data bus provides an information transmission channel between the detection units to ensure that the outputs of different detection units can be transmitted to the main controller in time for processing. In this way, a complete multi-level detection system is constructed, which can monitor the changes in voltage, current and power in real time and perform effective power supply status detection.

[0027] In a specific embodiment, the process of executing step S3 may specifically include the following steps: Based on the sampling synchronization signal, the multi-level detection system is controlled in time sequence to obtain a synchronous acquisition time series; Data is collected from the voltage detection unit, the current detection unit and the power detection unit according to the synchronous acquisition time sequence to obtain real-time detection data; Perform digital filtering and normalization processing on the real-time detection data to obtain pre-processed detection data; Based on the preprocessed detection data, nonlinear state expressions of the voltage equation and the current equation are established, and the power equation is constructed according to the voltage equation and the current equation to obtain a coupled equation group; The coupled equations are separated and linearized to obtain a linearized coefficient matrix, which is then input into an iterative solver and numerically calculated using the least squares method to obtain a solution vector for the equations. The solution vector of the equation group is subjected to eigenvalue analysis to obtain a characteristic parameter set, and state space transformation and matrix reconstruction are performed based on the characteristic parameter set to obtain a power supply state parameter matrix.

[0028] Specifically, the multi-level detection system is controlled in time sequence based on the sampling synchronization signal to obtain a synchronous acquisition time sequence. The timing control is performed by setting the sampling frequency To ensure the synchronization of the system, the sampling interval is set to: ; The sampling synchronization signal is generated by the clock signal to ensure that all units collect data at a specific time. ,in Indicates The voltage detection unit, current detection unit and power detection unit are sampled at each moment according to the synchronous acquisition time sequence. Data collection is performed to obtain real-time detection data. The voltage detection unit measures the voltage in real time through the voltage sensor, the current detection unit obtains the current value in real time through the Hall sensor, and the power detection unit obtains the power by calculating the product of voltage and current in real time. Assume that the voltage collection data is , the current acquisition data is , the power data is , these data form a multi-dimensional real-time data set. The real-time detection data is digitally filtered and normalized. Digital filtering is used to remove high-frequency noise. The filtering methods include low-pass filtering and median filtering. Assume that the voltage data after filtering is , the current data is , the power data is , effectively eliminating the measurement error and environmental noise through filtering. Normalization is to map data of different dimensions and amplitudes to a unified standard range, usually normalized to the range of [0, 1]. Assume that the normalized voltage data is , the current data is , the power data is , its normalized formula is: ; in and are the minimum and maximum values ​​of the voltage, and the same is true for other data. Based on the pre-processed detection data, the nonlinear state expressions of the voltage equation and the current equation are established. The relationship between voltage and current is described by a set of nonlinear equations that reflect the complex interaction between voltage, current and load. Assume that the voltage equation is: ; in is a nonlinear function describing the relationship between voltage and current. is the external environment or system parameter, and the current equation is: ; in It is a nonlinear function that describes the relationship between current and voltage. These two equations are combined to form a coupled equation group that describes the mutual influence between voltage and current. On this basis, the power equation is constructed. The relationship between power, voltage and current is expressed by the power formula Express: ; Through these equations, the coupled equations are obtained: ; The coupled equations are separated and linearized. Variable separation is to separate the variables in the voltage, current and power equations into linear terms and nonlinear terms, while linearization is to simplify the equations into linear form by Taylor expansion of the nonlinear terms. Assume that the above coupled equations are linearized to obtain the linearization coefficient matrix : ; After linearization, the system equations become a linear system, which is easy to solve. The linearization coefficient matrix is ​​input into the iterative solver, and the least squares method is used for numerical calculation to obtain the solution vector of the system of equations. The least squares method solves the optimal solution by minimizing the sum of squared errors. Assume that the solution vector of the system of equations is , optimize this solution vector by the least squares method to minimize the error. The objective function of the least squares method is: ; in is the experimental value, is the model prediction value, and the optimal solution is obtained by minimizing this function. After obtaining the solution vector of the system of equations, the solution vector is subjected to eigenvalue analysis to obtain the characteristic parameter set and evaluate the performance and stability of the system. Through eigenvalue decomposition, the eigenvalue and eigenvector are obtained to analyze the behavior of the system under different working conditions. Assume that the obtained eigenvalue is , the eigenvector is , these eigenvalues ​​and eigenvectors reveal the inherent characteristics of the system. Based on the characteristic parameter set, the state space transformation and matrix reconstruction are performed to obtain the power supply state parameter matrix The power supply state parameter matrix reflects the state of each part of the system and helps to adjust and optimize the performance of the power supply system. Through state space transformation, the nonlinear system is converted into a linear system, and further analysis and control are carried out on this basis.

[0029] In a specific embodiment, the execution step performs variable separation and linearization processing on the coupled equations to obtain a linearization coefficient matrix, and the linearization coefficient matrix is ​​input into an iterative solver, and numerical calculation is performed by the least squares method to obtain a solution vector of the equations. The process can specifically include the following steps: Perform variable separation operation on the coupled equations, separate the variables in the voltage equation, current equation and power equation according to linear terms and nonlinear terms, and obtain a separation variable matrix; Based on the separation variable matrix, Taylor expansion is performed on the nonlinear terms, and the expansion is truncated to obtain a linearized approximate equation group, which is converted into a matrix form, and a linearized coefficient matrix is ​​obtained by coefficient comparison; Perform condition number analysis on the linearization coefficient matrix, construct the Jacobi iteration matrix, obtain the initial iterative solution, and input the initial iterative solution into the least squares iterative solver, set the convergence threshold and the maximum number of iterations, and obtain the iterative calculation parameters; Perform the least squares iteration operation according to the iteration calculation parameters, and calculate the residual vector after each iteration to obtain an iteration sequence; The convergence of the iterative sequence is judged. When the residual is less than the preset threshold, the iteration is stopped to obtain the converged solution. The converged solution is normalized to generate the solution vector of the equation system.

[0030] Specifically, when performing variable separation operations on the coupled equations, the variables in the voltage equation, current equation, and power equation are reasonably decomposed to distinguish linear terms from nonlinear terms. They are expressed in a form that includes linear and nonlinear parts. Assume that the voltage equation is: ; in Indicates voltage, current and system parameters The nonlinear relationship. Similarly, the current equation is: ; Power equation Then it is the product of voltage and current: ; These three equations describe the coupling relationship between voltage, current and power. In order to separate the variables in these equations, the variables in these equations are distinguished into linear terms and nonlinear terms. For example, in the voltage equation, if the current is a nonlinear function, which can be expressed as: ; in, is the linear relationship between voltage and current, and This is the nonlinear relationship between voltage and current. Similarly, the current equation and power equation are processed similarly, and the variable separation matrix obtained is as follows: ; Taylor expansion is performed on nonlinear terms based on the separation variable matrix. Taylor expansion is a common method used to approximate complex nonlinear functions. and Perform Taylor expansion on nonlinear terms and ignore higher-order terms to obtain a linear approximation. , its Taylor expansion is expressed as: ; in and is the operating point of the system, is the partial derivative of the nonlinear term. Similar Taylor expansion is performed as for the power equation to obtain the linearized approximation of the current and power equations. After Taylor expansion of the nonlinear terms and truncation of the high-order terms, the linearized approximate equations are obtained and converted into matrix form. Assume that the following linearized equations are obtained: ; in, are linearization coefficients, which describe the linear relationship between voltage and current and the linear relationship between current and voltage respectively. The linearization coefficient matrix is ​​subjected to condition number analysis. The condition number is an indicator of the numerical stability of a matrix. A large condition number indicates that the numerical solution of the matrix is ​​unstable. , its condition number Calculated by the following formula: ; in is a matrix The norm of , using the 2 norm or the Frobenius norm, is the inverse of the matrix. By calculating the condition number, the stability of the matrix is ​​evaluated and it is determined whether additional measures are needed to ensure the accuracy of the numerical calculation. After the condition number analysis, the Jacobi iteration matrix is ​​constructed. The Jacobi method is one of the methods for solving linear equations and is applicable to sparse matrices. The core of the Jacobi iteration method is to approximate the solution of the equation system by continuously updating the iteration value. Set the initial solution to ,Right now: ; Then the new solution is calculated according to the Jacobi iteration formula: ; in is the inverse of the linearization coefficient matrix, is the error vector of the current solution. Through continuous iterative calculation, the solution vector is approximated. During the iteration process, the convergence threshold and the maximum number of iterations are set. If the residual vector of the iteration When it is less than the preset convergence threshold, the iteration is considered to have converged and the iteration is stopped. The residual vector is calculated as: ; If the residual is less than the set threshold ,Right now: ; The iteration is considered to have converged and the final solution vector is obtained After the iteration, the final solution is normalized to eliminate the influence of different dimensions. The normalization formula is: ; in and are the minimum and maximum values ​​of the solution vector respectively. The normalized solution vector The final solution vector to the system of equations provides accurate estimates of voltage, current, and power.

[0031] In a specific embodiment, the process of executing step S4 may specifically include the following steps: According to the voltage parameters, current parameters and power parameters, the power supply state parameter matrix is ​​analyzed to obtain the inner loop detection data sequence and the outer loop control data sequence; The inner loop detection data sequence is input into the inner loop response network of the dual-loop power supply control model for rapid response processing to obtain a voltage-current abnormal feature sequence. The inner loop response network includes a rapid response layer, a feature extraction layer and a state judgment layer. The rapid response layer uses a preset sampling frequency for data acquisition, the feature extraction layer uses wavelet transform for feature decomposition, and the state judgment layer uses threshold judgment for abnormality detection; The outer loop control data sequence is input into the outer loop control network of the dual-loop power supply control model for state tracking calculation to obtain the power supply state change trend. The outer loop control network includes a data cache layer, a trend analysis layer and a state prediction layer. The data cache layer uses a circular queue to store historical data, the trend analysis layer uses a sliding window to perform trend calculation, and the state prediction layer uses an autoregressive model to perform state prediction. The voltage-current abnormal feature sequence and the power supply state change trend are input into the feature fusion of the dual-loop power supply control model. The feature fusion includes a feature mapping unit and a decision unit. The feature mapping unit performs abnormal feature mapping to obtain an abnormal feature vector. The abnormal feature vector is classified by deep neural network. The deep neural network contains three fully connected layers. Each layer uses ReLU activation function and Dropout regularization. The output layer uses Softmax function for multi-classification to obtain the target abnormal type.

[0032] Specifically, according to the voltage parameters, current parameters and power parameters, the power supply state parameter matrix is ​​analyzed to obtain the inner loop detection data sequence and the outer loop control data sequence. According to the time series data of parameters such as voltage, current and power, the relevant matrix is ​​obtained through sampling and processing. For example, the voltage data sequence , current data series And power data series It is expressed as: ; ; ; in, Indicates the length of the data sequence, is the sampling time. Through these data sequences, the power supply state parameter matrix is ​​formed : ; This matrix contains information on all power supply status parameters, and the inner-loop detection data sequence and outer-loop control data sequence are parsed based on these data. The inner-loop detection data sequence mainly focuses on abnormal behaviors of voltage and current, while the outer-loop control data sequence focuses on the changing trend of power. In the dual-loop power supply control model, the inner-loop response network is used to quickly respond to abnormalities, and the outer-loop control network is used for state prediction and trend analysis. These two networks play different roles. The former performs real-time data collection and feature extraction through the fast response layer, while the latter performs long-term tracking and prediction of system status through the data cache layer and trend analysis layer. The inner-loop detection data sequence is input into the inner-loop response network of the dual-loop power supply control model. The inner-loop response network includes a fast response layer, a feature extraction layer, and a state judgment layer. In the fast response layer, a preset sampling frequency is used. To collect data: ; in, is the sampling interval. Through the fast response layer, the inner loop network can receive the voltage in real time. and current The feature extraction layer uses wavelet transform to decompose the collected voltage and current signals. The formula of wavelet transform is: ; in, is the mother wavelet function, and are the scale and translation parameters, is the input signal. Through wavelet transform, the instantaneous change characteristics in the voltage and current signals are extracted to capture potential abnormal patterns. After feature extraction, the state judgment layer uses the threshold judgment method for abnormality detection. When the abnormal characteristic values ​​of voltage and current exceed the set threshold, When the event is detected, it is considered an abnormal event: ; The inner loop response network identifies the abnormal conditions of voltage and current in real time and generates a voltage-current abnormal feature sequence. At the same time, the outer loop control data sequence is input into the outer loop control network of the dual-loop power supply control model. The outer loop control network includes a data cache layer, a trend analysis layer, and a state prediction layer. The data cache layer stores historical data in a circular queue. The length of the circular queue is Set to 1024 data points, and use the first-in-first-out method to update data, ensuring that the system has the latest historical data available for analysis at any time: ; In the trend analysis layer, a sliding window method is used to calculate the trend of historical data. The sliding window size is Each time the calculation is performed, the linear fitting trend of the data in the window is analyzed to obtain the trend coefficient and the intercept , through the trend formula: ; In the state prediction layer, the autoregressive (AR) model is used to predict the trend. Assume that the state prediction model is: ; in, is the autoregressive coefficient of the model, is the error term. In the autoregressive model, the future power change trend is predicted by historical data to obtain the power supply status change trend. The voltage-current abnormal feature sequence and the power supply status change trend are input into the feature fusion of the dual-loop power supply control model. The feature fusion consists of a feature mapping unit and a decision unit. In the feature mapping unit, the voltage-current abnormal feature and the power supply status change trend are fused to generate an abnormal feature vector. : ; The generated abnormal feature vector Input to the deep neural network for classification. The deep neural network consists of three fully connected layers. Each hidden layer uses the ReLU activation function and Dropout regularization to avoid overfitting. The output layer of the neural network uses the Softmax function for multi-classification to obtain the target anomaly type. : ; in, It is The scores of the categories, is the total number of categories. Through the results of Softmax output, various abnormal types are obtained, such as overvoltage, overcurrent, power abnormality, etc., and then the corresponding power supply control instructions are generated.

[0033] In a specific embodiment, the process of executing the step of inputting the outer loop control data sequence into the outer loop control network of the dual-loop power supply control model for state tracking calculation to obtain the power supply state change trend may specifically include the following steps: The outer loop control data sequence is input into the circular queue of the data cache layer. The length of the circular queue is 1024 data points. The read-write pointer and overflow processing mechanism are set to perform circular storage. Each storage operation performs a first-in-first-out data update to obtain a historical data sequence. Divide the historical data sequence into time windows to obtain overlapping time window sequences, and extract multidimensional statistical features based on the time window sequences to obtain statistical feature vectors; The statistical feature vectors are grouped according to voltage features, current features and power features, and are input into three trend fitting units of the trend analysis layer in the outer loop control network respectively. The linear trend coefficient and quadratic trend coefficient of each group of features are calculated, and the trend coefficients are combined to form a trend feature sequence. The trend feature sequence is input into the autoregressive model of the state prediction layer in the outer loop control network, the optimal order p=8 is determined by the AIC information criterion, and the autoregressive coefficient is calculated by the Levinson-Durbin recursive algorithm to obtain the AR(8) model parameters; The state prediction equation is constructed according to the AR(8) model parameters, and the prediction state sequence including voltage state, current state and power state is obtained through recursive iterative calculation; The root mean square error of the 1-step, 4-step, 8-step and 16-step predictions of the prediction state sequence is calculated to obtain the prediction confidence score. Based on the prediction confidence score, the prediction results of different time scales of the prediction state sequence are adaptively weighted fused to obtain the power supply state change trend.

[0034] Specifically, the outer loop control data sequence is input into the circular queue of the data cache layer. The length of the circular queue is set to 1024 data points to store enough historical data for further processing. The read and write pointers and overflow handling mechanism are set to ensure the circular storage of data. When the queue is full, the new data point will overwrite the oldest data point to ensure that the latest historical data is always stored in the queue. For example, if the outer loop control data sequence is The storage method of the circular queue is first-in-first-out. Each time data is written, the new data point Will overwrite the earliest data point , forming a new queue . Circular storage ensures that the data is always up to date and can reflect the real-time changes in the system status. Divide the historical data sequence into time windows to obtain overlapping time window sequences. Divide the continuous data sequence into multiple subsequences, each of which contains a fixed time window of data. Assume that the historical data is divided into time windows, overlapping time window sequences It is expressed as: ; Among them, each subsequence Include By analyzing each time window, we can extract multi-dimensional statistical features, such as mean, variance, maximum, minimum, etc., to reflect the changing trend and fluctuation of the power supply system in different time periods. For example, for voltage data, we calculate the mean voltage in each window. , Standard Deviation , maximum value and minimum value , and use these values ​​as part of the voltage eigenvector: ; Similarly, for current and power data, the corresponding statistical features are extracted to obtain the current feature vector and power feature vector. All features will constitute the statistical feature vector: ; These statistical feature vectors are used as input data and enter the trend analysis layer in the outer loop control network for analysis. The trend analysis layer of the outer loop control network includes three trend fitting units, which are used to analyze the trend changes of voltage, current and power characteristics respectively. Each trend fitting unit obtains the corresponding trend coefficient by performing linear fitting and quadratic fitting on its own feature sequence. For example, the linear trend coefficient of the voltage feature sequence is and the quadratic trend coefficient By using the least squares method, we can obtain: ; ; in, and are the coefficients of the linear and quadratic fits, and is the intercept term. By similar methods, the trend coefficients of current and power are obtained. Combining these trend coefficients forms a trend characteristic sequence : ; The trend feature sequence Input the state prediction layer in the outer loop control network. The state prediction layer uses the autoregressive (AR) model for time series prediction. The optimal autoregressive order is selected by the AIC (Akaike Information Criterion) information criterion. , and then use the Levinson-Durbin recursive algorithm to calculate the autoregressive coefficients , thus establishing the AR(8) model: ; in, is a white noise error term. Through recursive iteration, the predicted power supply state sequence is obtained, including the voltage state , current state and power status When performing state prediction, the state prediction equation is constructed based on the obtained AR(8) model parameters, and the recursive method is used for iterative calculation to obtain the predicted future states of voltage, current and power: ; ; ; After obtaining the predicted state sequence, the root mean square error (RMSE) of 1 step, 4 steps, 8 steps and 16 steps is calculated respectively: ; in, is the predicted value, is the actual value, is the prediction step size. By calculating the RMSE of different time steps, the accuracy of the prediction results is evaluated and the prediction confidence score is obtained. Based on the RMSE scores of different time scales, the prediction state sequence is adaptively weighted and fused. The prediction results of 1 step, 4 steps, 8 steps and 16 steps are weighted according to the RMSE score by using the weighted average method: ; in, is the weight reflected by the RMSE score, It is The fusion result obtained is That is, the changing trend of the power supply status, reflecting the state evolution of the system in the future.

[0035] The above describes the power supply detection method of the POE power supply in the embodiment of the present invention. The following describes the power supply detection device of the POE power supply in the embodiment of the present invention. Figure 2 , an embodiment of a POE power supply detection device in an embodiment of the present invention includes: The acquisition module is used to collect voltage data and current data of the POE power supply system, perform characteristic curve fitting operation on the voltage data and current data to obtain a voltage-current characteristic curve, and calculate the first-order derivative and second-order derivative of the voltage-current characteristic curve to obtain a power supply parameter sensitivity matrix; A construction module, used to construct a voltage detection unit, a current detection unit and a power detection unit according to a power supply parameter sensitivity matrix, and connect the voltage detection unit, the current detection unit and the power detection unit through a data bus to form a multi-level detection system; A solution module is used to collect real-time detection data through a multi-level detection system, establish a coupled equation group including a voltage equation, a current equation and a power equation, and numerically solve the coupled equation group to obtain a power supply state parameter matrix; A detection module is used to input the power supply state parameter matrix into the dual-loop power supply control model, perform anomaly detection through the inner loop response network and the outer loop control network in the dual-loop power supply control, and obtain the target anomaly type; The execution module is used to perform multi-level judgment operations based on the target abnormality type, obtain a hierarchical alarm control strategy, generate a power supply control instruction sequence according to the hierarchical alarm control strategy, and execute power supply protection control.

[0036] Through the coordinated cooperation of the above-mentioned components, by establishing the voltage-current characteristic curve and calculating the sensitivity matrix, the electrical characteristics of the POE power supply system are accurately described. The multi-level detection system structure is adopted, combining the detection units of the three dimensions of voltage, current and power, which improves the integrity and accuracy of the detection data; the dual-loop power supply control model is introduced, and the detection sensitivity and response speed of the system to power supply anomalies are enhanced through the synergy of the inner loop fast response and the outer loop state tracking; the numerical solution method based on the coupled equation group solves the nonlinear coupling problem between the power supply parameters, making the detection results more accurate; the deep neural network is used for anomaly classification, combined with the multi-level judgment mechanism, to improve the accuracy of anomaly type identification; through the autoregressive model and trend analysis, the accurate prediction of the power supply status is achieved, which provides a basis for preventive maintenance; a complete hierarchical alarm and protection control strategy is designed to effectively reduce the probability of power supply accidents.

[0037] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0038] Those skilled in the art will understand that Figure 3The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0039] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0040] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0041] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0042] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0043] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A POE power supply detection method, characterized in that: The method comprises: Collect voltage data and current data of the POE power supply system, perform characteristic curve fitting operation on the voltage data and current data to obtain a voltage-current characteristic curve, and calculate the first-order derivative and second-order derivative of the voltage-current characteristic curve to obtain a power supply parameter sensitivity matrix; Constructing a voltage detection unit, a current detection unit and a power detection unit according to the power supply parameter sensitivity matrix, and connecting the voltage detection unit, the current detection unit and the power detection unit through a data bus to form a multi-level detection system; Collecting real-time detection data through the multi-level detection system, establishing a coupled equation group including a voltage equation, a current equation and a power equation, and numerically solving the coupled equation group to obtain a power supply state parameter matrix; Inputting the power supply state parameter matrix into a dual-loop power supply control model, performing anomaly detection through an inner-loop response network and an outer-loop control network in the dual-loop power supply control, and obtaining a target anomaly type; A multi-level judgment operation is performed based on the target abnormality type to obtain a hierarchical alarm control strategy, and a power supply control instruction sequence is generated according to the hierarchical alarm control strategy to execute power supply protection control.

2. The POE power supply detection method according to claim 1, characterized in that: The collecting of voltage data and current data of the POE power supply system, performing characteristic curve fitting operation on the voltage data and current data to obtain a voltage-current characteristic curve, and calculating the first-order derivative and the second-order derivative of the voltage-current characteristic curve to obtain a power supply parameter sensitivity matrix, including: Sampling the POE power supply system under different load conditions to obtain a voltage sampling sequence and a current sampling sequence, and performing digital signal processing on the voltage sampling sequence and the current sampling sequence to obtain a voltage data set and a current data set; Performing feature extraction and data fitting operations on the voltage data set and the current data set to obtain a voltage-current characteristic polynomial equation, and performing coefficient normalization processing on the voltage-current characteristic polynomial equation to obtain a standardized characteristic curve equation; Based on the standardized characteristic curve equation, a first-order derivative curve of the voltage-current relationship is calculated to obtain a voltage-current sensitivity function, and a second-order differential operation is performed on the voltage-current sensitivity function to obtain a second-order sensitivity function of the voltage-current characteristic; The voltage-current sensitivity function and the second-order sensitivity function are numerically calculated at a preset working point to generate a power supply parameter sensitivity matrix.

3. The POE power supply detection method according to claim 2, characterized in that: The voltage detection unit, the current detection unit and the power detection unit are constructed according to the power supply parameter sensitivity matrix, and the voltage detection unit, the current detection unit and the power detection unit are connected through a data bus to form a multi-level detection system, including: Performing eigendecomposition operation on the power supply parameter sensitivity matrix to obtain a voltage sensitivity eigenvector, a current sensitivity eigenvector, and a power sensitivity eigenvector; According to the voltage sensitivity characteristic vector, a voltage sampling circuit is configured with parameters, a voltage signal is collected through the voltage sampling circuit, and analog-to-digital conversion and data caching are performed on the voltage signal to obtain a voltage detection unit; Setting detection parameters of a Hall current sensor based on the current sensitivity characteristic vector, collecting a current signal through the Hall current sensor, and conditioning and amplifying the current signal to obtain a current detection unit; Inputting the output data of the voltage detection unit and the output data of the current detection unit into a digital multiplier, performing real-time power calculation through the digital multiplier, and obtaining a power detection unit; Performing sampling timing synchronization control on the voltage detection unit, the current detection unit and the power detection unit according to the power sensitivity characteristic vector to obtain a sampling synchronization signal; Based on the sampling synchronization signal, the voltage detection unit, the current detection unit and the power detection unit are connected through a data bus to form a multi-level detection system.

4. The POE power supply detection method according to claim 3, characterized in that: The multi-level detection system is used to collect real-time detection data, establish a coupled equation group including a voltage equation, a current equation and a power equation, and numerically solve the coupled equation group to obtain a power supply state parameter matrix, including: Performing timing control on the multi-stage detection system based on the sampling synchronization signal to obtain a synchronous acquisition time series; Performing data acquisition on the voltage detection unit, the current detection unit and the power detection unit according to the synchronous acquisition time sequence to obtain real-time detection data; Performing digital filtering and normalization processing on the real-time detection data to obtain pre-processed detection data; Based on the pre-processed detection data, a nonlinear state expression of a voltage equation and a current equation is established, and a power equation is constructed according to the voltage equation and the current equation to obtain a coupled equation group; Performing variable separation and linearization processing on the coupled equations to obtain a linearization coefficient matrix, and inputting the linearization coefficient matrix into an iterative solver, performing numerical calculations through a least squares method, and obtaining a solution vector of the equations; Performing eigenvalue analysis on the solution vector of the set of equations to obtain a characteristic parameter set, and performing state space transformation and matrix reconstruction based on the characteristic parameter set to obtain a power supply state parameter matrix.

5. The POE power supply detection method according to claim 4, characterized in that: The step of performing variable separation and linearization processing on the coupled equations to obtain a linearization coefficient matrix, and inputting the linearization coefficient matrix into an iterative solver, performing numerical calculation by the least squares method, and obtaining a solution vector of the equations includes: Performing a variable separation operation on the coupled equation group, separating the variables in the voltage equation, the current equation and the power equation according to linear terms and nonlinear terms, and obtaining a separation variable matrix; Based on the separation variable matrix, Taylor expansion is performed on the nonlinear terms, and the expansion is truncated to obtain a linearized approximate equation group, the linearized approximate equation group is converted into a matrix form, and a linearized coefficient matrix is ​​obtained by coefficient comparison; Performing condition number analysis on the linearization coefficient matrix, constructing a Jacobi iteration matrix, obtaining an iterative initial solution, and inputting the iterative initial solution into a least squares iterative solver, setting a convergence threshold and a maximum number of iterations, and obtaining iterative calculation parameters; Performing a least squares iterative operation according to the iterative calculation parameters, and calculating a residual vector after each iteration to obtain an iterative sequence; The convergence of the iterative sequence is judged, and when the residual is less than a preset threshold, the iteration is stopped to obtain a converged solution, and the converged solution is normalized to generate a solution vector of the equation system.

6. The POE power supply detection method according to claim 5, characterized in that: The power supply state parameter matrix is ​​input into the dual-loop power supply control model, and anomaly detection is performed through the inner-loop response network and the outer-loop control network in the dual-loop power supply control to obtain the target anomaly type, including: According to the voltage parameters, current parameters and power parameters, the power supply state parameter matrix is ​​analyzed to obtain an inner loop detection data sequence and an outer loop control data sequence; The inner loop detection data sequence is input into the inner loop response network of the dual-loop power supply control model for rapid response processing to obtain a voltage-current abnormality feature sequence, wherein the inner loop response network includes a rapid response layer, a feature extraction layer and a state judgment layer, wherein the rapid response layer uses a preset sampling frequency for data acquisition, the feature extraction layer uses wavelet transform for feature decomposition, and the state judgment layer uses threshold judgment for abnormality detection; The outer loop control data sequence is input into the outer loop control network of the dual-loop power supply control model for state tracking calculation to obtain the power supply state change trend, the outer loop control network includes a data cache layer, a trend analysis layer and a state prediction layer, the data cache layer uses a circular queue to store historical data, the trend analysis layer uses a sliding window to perform trend calculation, and the state prediction layer uses an autoregressive model to perform state prediction; Inputting the voltage-current abnormal feature sequence and the power supply state change trend into the feature fusion of the dual-loop power supply control model, the feature fusion includes a feature mapping unit and a decision unit, and performing abnormal feature mapping through the feature mapping unit to obtain an abnormal feature vector; The abnormal feature vector is classified by a deep neural network, wherein the deep neural network includes three fully connected layers, each layer uses a ReLU activation function and Dropout regularization, and the output layer uses a Softmax function for multi-classification to obtain the target abnormal type.

7. The POE power supply detection method according to claim 6, characterized in that: The step of inputting the outer loop control data sequence into the outer loop control network of the dual-loop power supply control model to perform state tracking calculation to obtain a power supply state change trend includes: The outer loop control data sequence is input into the circular queue of the data cache layer, the length of the circular queue is 1024 data points, and circular storage is performed by setting the read-write pointer and overflow processing mechanism, and each storage operation performs a first-in-first-out data update to obtain a historical data sequence; Dividing the historical data sequence into time windows to obtain an overlapping time window sequence, and performing multidimensional statistical feature extraction based on the time window sequence to obtain a statistical feature vector; The statistical feature vectors are grouped according to voltage features, current features and power features, and are respectively input into three trend fitting units of the trend analysis layer in the outer loop control network, a linear trend coefficient and a quadratic trend coefficient of each group of features are calculated, and the trend coefficients are combined to form a trend feature sequence; Input the trend feature sequence into the autoregressive model of the state prediction layer in the outer loop control network, determine the optimal order p=8 by the AIC information criterion, calculate the autoregressive coefficient by the Levinson-Durbin recursive algorithm, and obtain the AR(8) model parameters; Constructing a state prediction equation according to the AR (8) model parameters, and obtaining a prediction state sequence including voltage state, current state and power state through recursive iterative calculation; The root mean square errors of the 1-step, 4-step, 8-step and 16-step predictions of the prediction state sequence are calculated respectively to obtain the prediction confidence score, and based on the prediction confidence score, the prediction results of different time scales of the prediction state sequence are adaptively weighted fused to obtain the power supply state change trend.

8. A POE power supply detection device, characterized in that: Used to perform the power supply detection method of a POE power supply according to any one of claims 1 to 7, the power supply detection device of the POE power supply comprises: The acquisition module is used to collect voltage data and current data of the POE power supply system, perform characteristic curve fitting operation on the voltage data and current data to obtain a voltage-current characteristic curve, and calculate the first-order derivative and second-order derivative of the voltage-current characteristic curve to obtain a power supply parameter sensitivity matrix; A construction module, used to construct a voltage detection unit, a current detection unit and a power detection unit according to the power supply parameter sensitivity matrix, and connect the voltage detection unit, the current detection unit and the power detection unit through a data bus to form a multi-level detection system; A solution module, used for collecting real-time detection data through the multi-level detection system, establishing a coupled equation group including a voltage equation, a current equation and a power equation, and numerically solving the coupled equation group to obtain a power supply state parameter matrix; A detection module, used for inputting the power supply state parameter matrix into a dual-loop power supply control model, performing anomaly detection through an inner-loop response network and an outer-loop control network in the dual-loop power supply control, and obtaining a target anomaly type; The execution module is used to perform multi-level judgment operations based on the target abnormality type to obtain a hierarchical alarm control strategy, and generate a power supply control instruction sequence according to the hierarchical alarm control strategy to execute power supply protection control.

9. A computer device, characterized in that: It comprises a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and is characterized in that when the processor executes the computer program, the power supply detection method of the POE power supply described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor is enabled to execute the power supply detection method of a POE power supply according to any one of claims 1 to 7.

Images