Search

CN-121984708-A - Lightweight security detection method, system, terminal and storage medium for energy efficiency management of industrial equipment

CN121984708ACN 121984708 ACN121984708 ACN 121984708ACN-121984708-A

Abstract

The invention discloses a lightweight security detection method, a system, a terminal and a storage medium for energy efficiency management of industrial equipment, wherein the method comprises the steps of converting multiprotocol industrial data into a unified feature matrix through zero copy analysis on an edge gateway, generating an anti-interference robust feature matrix through compound normalization processing, inputting the matrix in parallel into a lightweight automatic encoder and an optimized isolation forest for bimodal analysis, outputting a comprehensive anomaly score through a dynamic fusion layer, calculating energy efficiency deviation degree by combining an equipment energy efficiency baseline, forming a security and energy efficiency combined decision vector, matching a progressive response strategy and executing a control instruction in real time according to the security and energy efficiency combined decision vector, quantifying carbon footprint and energy saving strategy emission reduction in the detection process, and optimizing the model through feedback. The invention realizes high-efficiency deployment at the edge side of the resource limitation, has strong anti-interference and high real-time performance, opens up a linkage closed loop with optimized safety protection and energy efficiency, and provides verifiable carbon emission reduction benefits.

Inventors

  • LU WEICHAO

Assignees

  • 深圳开鸿数字产业发展有限公司

Dates

Publication Date
20260505
Application Date
20251225

Claims (20)

  1. 1. The lightweight security detection method for the energy efficiency management of the industrial equipment is characterized by comprising the following steps of: On an industrial Internet of things gateway carrying an edge operating system, carrying out zero copy analysis on heterogeneous industrial equipment data streams through a multi-protocol analysis engine, and converting the heterogeneous industrial equipment data streams into a structured feature matrix with uniform dimension; performing Z-score standardization, logarithmic transformation and quarter bit distance noise suppression processing on the structural feature matrix, and introducing an environment compensation factor to generate a robust feature matrix; The robust feature matrix is input to a lightweight automatic encoder and an isolation forest in parallel to perform bimodal anomaly analysis, and the comprehensive anomaly score is output through a dynamic fusion layer according to the self-adaptive weighted fusion of the system load and the historical anomaly score; judging a safety state level according to the comprehensive abnormal score and the multi-level safety threshold, and calculating the energy efficiency deviation degree by combining the equipment energy efficiency base line to form a combined decision vector of safety and energy efficiency fusion; generating and issuing an energy efficiency optimization or safety protection instruction according to the joint decision vector matching progressive response strategy, and executing in a determined time delay through an operating system real-time scheduler; And converting the carbon footprint of the detection process based on the system power consumption and the processing time, accumulating the carbon emission reduction achieved by the energy saving strategy, and updating the equipment energy efficiency base line and the detection model through state feedback.
  2. 2. The method for detecting the lightweight security of the energy efficiency management of the industrial equipment according to claim 1, wherein the method for carrying out zero copy analysis on the heterogeneous industrial equipment data stream by a multi-protocol analysis engine on the industrial internet of things gateway carrying the edge operation system is converted into a structured feature matrix with uniform dimension, and the method specifically comprises the following steps: On an industrial Internet of things gateway carrying an edge operating system, carrying out zero-copy analysis on original data streams of heterogeneous industrial equipment with different protocols through a multi-protocol analysis engine built in an open source hong Meng operating system; Converting the equipment operation data packets of different protocols into a structured feature matrix with uniform dimensions; Wherein the structured feature matrix comprises power, current, voltage and temperature physical quantity features.
  3. 3. The industrial equipment energy efficiency management oriented lightweight security detection method of claim 2, wherein the raw data stream is: ; Wherein d n represents a protocol data packet of the n-th industrial equipment; the equipment operation data are as follows: ; Wherein d i represents a protocol data packet of the i-th industrial equipment, Representing a protocol parsing function, p representing a protocol set; The structured feature matrix is: ; Where k represents the number of features and m represents the number of rows of the structured feature matrix.
  4. 4. The method for detecting the lightweight security oriented to the energy efficiency management of the industrial equipment according to claim 1, wherein the processing of Z-score normalization, logarithmic transformation and quarter-bit-distance noise suppression is performed on the structured feature matrix, and an environmental compensation factor is introduced to generate a robust feature matrix, and the method specifically comprises: Performing third-order composite normalization processing on the structured feature matrix to obtain a target structured feature matrix, wherein the third-order composite normalization processing comprises Z-score standardization based on feature mean and standard deviation, logarithmic transformation for compressing outliers and impulse noise suppression based on quartile range; and introducing an industrial environment compensation factor into the target structural feature matrix to generate a normalized robust feature matrix.
  5. 5. The industrial equipment energy efficiency management oriented lightweight security detection method of claim 4, wherein the robust feature matrix is: ; Wherein, the And Is characterized by Is defined as the mean and standard deviation of (c), A j-th sample value representing an i-th feature column in the target structured feature matrix, Representing characteristics Medium represents the median, IQR represents the quartile range, , 。
  6. 6. The method for detecting the lightweight security oriented to the energy efficiency management of the industrial equipment according to claim 1, wherein the step of inputting the robust feature matrix in parallel to a lightweight automatic encoder to perform bimodal anomaly analysis with an isolated forest, and outputting a comprehensive anomaly score through a dynamic fusion layer according to self-adaptive weighted fusion of a system load and a historical anomaly score comprises the following steps: Inputting the robust feature matrix to a lightweight automatic encoder path and an optimized isolation forest path in parallel for bimodal analysis; The lightweight automatic encoder access compresses and reconstructs a normal working condition mode through a 32-dimensional hidden space, and outputs a reconstruction error score based on a reconstruction error; The optimized isolation forest path performs isolation analysis on equipment energy consumption abnormal behaviors by constructing a shallow decision tree with limited depth, and outputs an abnormal score based on path length; And through a dynamic fusion layer, the output weights of the lightweight automatic encoder path and the optimized isolation forest path are adaptively adjusted according to the real-time CPU load of the edge gateway and the anomaly score at the previous moment, and the weighted fusion is carried out to generate a comprehensive anomaly fusion score.
  7. 7. The industrial equipment energy efficiency management oriented lightweight security detection method of claim 6, wherein the lightweight automatic encoder path processing procedure is: ; ; ; ; Wherein, the Representing the output result of the encoder, Representing the output result of the decoder, Representing the loss of reconstruction and, The original input value representing the i-th feature, Representing a matrix of weight parameters of the encoder, Representing the bias parameter vector of the encoder, Representing a matrix of weight parameters of the decoder, Represents the bias parameter vector of the decoder, k represents the feature quantity, An input value representing the ith feature from the robust feature matrix, A model reconstruction value representing the i-th feature, The regularization coefficient is represented as a function of the regularization coefficient, Representing a reconstruction error score; The optimized isolated forest passage is processed by the following steps: ; Wherein, the Representing an anomaly score, E representing a desire, The path length is indicated as such, The sub-sample size is indicated and, Represents the normalization factor, H represents the sum of the tones, Representing the sum of k items before the harmonic progression; The processing process of the dynamic fusion layer is as follows: ; ; Wherein, the The weight is represented by a weight that, Indicating the load of the CPU, Representing the composite anomaly fusion score.
  8. 8. The method for detecting the lightweight security oriented to the energy efficiency management of the industrial equipment according to claim 7, wherein the step of determining the security state level according to the comprehensive anomaly score and the multi-level security threshold value, and calculating the energy efficiency deviation degree by combining the equipment energy efficiency base line to form a combined decision vector for fusing the security and the energy efficiency specifically comprises the following steps: comparing the comprehensive abnormal fusion score with a preset multilevel safety threshold value, and judging the current safety state level of the equipment; Inquiring a preset industrial equipment knowledge base, acquiring a reference energy efficiency curve of the current equipment, and obtaining the energy efficiency deviation degree by calculating the deviation degree of real-time operation characteristics of the equipment and the reference energy efficiency curve and quantifying the energy efficiency state; and combining the safety state level and the energy efficiency deviation degree to form a combined decision vector comprising safety risk and energy efficiency level.
  9. 9. The industrial equipment energy efficiency management oriented lightweight security detection method of claim 8, wherein the current security state level is: ; Wherein, the And All represent threshold values, and < , =0.4, =0.7; The threshold is estimated from historical non-outlier data: ; wherein P represents a probability function; The energy efficiency deviation degree is as follows: ; Wherein, the The degree of deviation of the energy efficiency is indicated, Representing a working condition vector, including power, current, temperature and load rate, Representing a reference vector; The joint decision vector is: ; Wherein, the Representing the joint decision vector(s), Representing a function of the reduced power of the device, Indicating an emergency-off function, Indicating no operating function.
  10. 10. The method for detecting the lightweight security oriented to the energy efficiency management of the industrial equipment according to claim 2, wherein the step of generating and issuing the energy efficiency optimization or the security protection instruction according to the matching of the joint decision vector and the progressive response strategy is performed within a determined time delay through an operating system real-time scheduler, and specifically comprises the steps of: according to the joint decision vector, invoking a security policy engine to match a predefined progressive response policy; when the decision indicates that potential safety hazards exist but the emergency degree is not reached and the energy efficiency deviation is remarkable, generating and issuing an optimization instruction for reducing the running power of the equipment; when the decision indicates that high-risk faults exist, generating and issuing a protection instruction for emergency shutdown or power failure; All control instructions are issued to the controlled industrial equipment within deterministic time delay constraints via the real-time task scheduler of the open-source hong-mo operating system.
  11. 11. The method for detecting the lightweight safety of the energy efficiency management of the industrial equipment according to claim 1, wherein the method for detecting the carbon footprint of the process based on the conversion of the system power consumption and the processing time length, accumulating the carbon emission reduction achieved by the energy saving strategy, and updating the equipment energy efficiency baseline and the detection model through state feedback specifically comprises the following steps: based on the real-time power consumption and algorithm processing time of the edge gateway, converting carbon dioxide equivalent generated in a single detection and decision process through a carbon footprint model; recording and accumulating the energy consumption saving amount of the equipment realized after the coordinated control strategy is executed, and quantifying the actual carbon emission reduction benefit by combining the regional power grid carbon emission factor; And returning the state feedback data after the equipment executes the control instruction to the cloud or local knowledge base for updating the equipment energy efficiency baseline and optimizing the detection model parameters.
  12. 12. The industrial equipment energy efficiency management oriented lightweight security detection method of claim 11, wherein the carbon dioxide equivalent is: ; Wherein, the Represents the equivalent amount of carbon dioxide, The usage amount of the CPU is indicated, The time delay is indicated as such, Representing the peak power consumption and, Representing a grid factor; Wherein, the execution delay constraint is: t exec =t detect +t decision +15ms ; Wherein t exec represents the end-to-end total execution delay, t detect represents the anomaly detection delay, and t decision represents the intelligent decision delay; The carbon emission reduction amount is as follows: ΔC=ΔP×t duration × ; Where ΔC represents the carbon reduction amount, ΔP represents the average power reduction value, and t duration represents the strategy duration.
  13. 13. The utility model provides a lightweight security detection system towards industrial equipment energy efficiency management which characterized in that, lightweight security detection system towards industrial equipment energy efficiency management includes: The protocol adaptation and feature extraction module is used for carrying out zero-copy analysis on the heterogeneous industrial equipment data stream through the multi-protocol analysis engine on the industrial Internet of things gateway carrying the edge operation system and converting the heterogeneous industrial equipment data stream into a structured feature matrix with uniform dimension; the robust preprocessing and noise suppression module is used for performing Z-score standardization, logarithmic transformation and quarter bit distance noise suppression processing on the structural feature matrix, introducing an environment compensation factor and generating a robust feature matrix; The bimodal dynamic fusion analysis module is used for inputting the robust feature matrix to the lightweight automatic encoder in parallel to perform bimodal anomaly analysis with the isolation forest, and outputting comprehensive anomaly score through the dynamic fusion layer according to the adaptive weighted fusion of the system load and the historical anomaly score; the safety and energy efficiency combined decision module is used for judging the safety state level according to the comprehensive abnormal score and the multi-level safety threshold value, and calculating the energy efficiency deviation degree by combining the equipment energy efficiency base line to form a combined decision vector of safety and energy efficiency fusion; The progressive strategy executing and controlling module is used for generating and issuing an energy efficiency optimizing or safety protecting instruction according to the matching progressive response strategy of the joint decision vector, and executing the instruction in a determined time delay through an operating system real-time scheduler; And the carbon benefit quantification and closed-loop optimization module is used for resolving the carbon footprint of the detection process based on the system power consumption and the processing time, accumulating the carbon emission reduction achieved by the energy saving strategy, and updating the equipment energy efficiency baseline and the detection model through state feedback.
  14. 14. The industrial equipment energy efficiency management oriented lightweight security detection system of claim 13, wherein the protocol adaptation and feature extraction module comprises: the data analysis unit is used for carrying out zero-copy analysis on the original data streams of heterogeneous industrial equipment with different protocols on an industrial Internet of things gateway carrying an edge operating system through a multi-protocol analysis engine built in an open source hong Meng operating system; the device comprises a feature conversion unit, a feature conversion unit and a processing unit, wherein the feature conversion unit is used for converting device operation data packets of different protocols into a structured feature matrix with uniform dimensions, and the structured feature matrix comprises power, current, voltage and temperature physical quantity features.
  15. 15. The industrial equipment energy efficiency management oriented lightweight security detection system of claim 13, wherein the robust pre-processing and noise suppression module comprises: The normalization processing unit is used for carrying out third-order composite normalization processing on the structural feature matrix to obtain a target structural feature matrix, wherein the third-order composite normalization processing comprises Z-score normalization based on a feature mean value and a standard deviation, logarithmic transformation for compressing outliers and impulse noise suppression based on a quartile range; And the characteristic compensation unit is used for introducing an industrial environment compensation factor into the target structural characteristic matrix to generate a normalized robust characteristic matrix.
  16. 16. The industrial equipment energy efficiency management oriented lightweight security detection system of claim 13, wherein the bimodal dynamic fusion analysis module comprises: The matrix analysis unit is used for inputting the robust feature matrix to the lightweight automatic encoder path and the optimized isolation forest path in parallel to perform bimodal analysis; the feature learning path processing unit is used for compressing and reconstructing a normal working condition mode through a 32-dimensional hidden space by the lightweight automatic encoder path and outputting a reconstruction error score based on a reconstruction error; The behavior analysis path processing unit is used for performing isolation analysis on the abnormal behavior of the equipment energy consumption by constructing a shallow decision tree with limited depth by the optimized isolation forest path and outputting an abnormal score based on the path length; The weighting fusion unit is used for adaptively adjusting the output weights of the lightweight automatic encoder path and the optimized isolation forest path according to the real-time CPU load of the edge gateway and the abnormal score at the previous moment through the dynamic fusion layer, and carrying out weighting fusion to generate the comprehensive abnormal fusion score.
  17. 17. The industrial equipment energy efficiency management oriented lightweight security detection system of claim 13, wherein the security and energy efficiency joint decision module comprises: The safety level judging unit is used for comparing the comprehensive abnormal fusion score with a preset multilevel safety threshold value and judging the current safety state level of the equipment; The deviation quantization unit is used for inquiring a preset industrial equipment knowledge base, acquiring a reference energy efficiency curve of the current equipment, calculating the deviation degree of real-time operation characteristics of the equipment and the reference energy efficiency curve, and quantifying the energy efficiency state to obtain the energy efficiency deviation degree; And the decision vector generation unit is used for combining the safety state level and the energy efficiency deviation degree to form a combined decision vector containing safety risk and energy efficiency level.
  18. 18. The industrial equipment energy efficiency management oriented lightweight security detection system of claim 13, wherein the progressive policy enforcement and control module comprises: the strategy matching unit is used for calling a security strategy engine to match a predefined progressive response strategy according to the joint decision vector; the optimizing instruction generating unit is used for generating and transmitting an optimizing instruction for reducing the running power of the equipment when the decision indicates that the potential safety hazard exists but the emergency degree is not reached and the energy efficiency deviation is obvious; the protection instruction generation unit is used for generating and issuing a protection instruction of emergency shutdown or power failure when the decision indicates that the high-risk fault exists; All control instructions are issued to the controlled industrial equipment within deterministic latency constraints via a real-time task scheduler of an open-source hong-mo operating system.
  19. 19. The industrial equipment energy efficiency management oriented lightweight security detection system of claim 13, wherein the carbon benefit quantification and closed loop optimization module comprises: the carbon footprint calculation unit is used for converting carbon dioxide equivalent generated in a single detection and decision process through a carbon footprint model based on real-time power consumption and algorithm processing time of the edge gateway; The carbon emission reduction unit is used for recording and accumulating the energy consumption saving amount of the equipment realized after the linkage control strategy is executed, and quantizing the actual carbon emission reduction benefit by combining the regional power grid carbon emission factor; The closed loop optimization unit is used for transmitting state feedback data after the equipment executes the control instruction back to the cloud or local knowledge base and updating the equipment energy efficiency base line and optimizing the detection model parameters.
  20. 20. A terminal comprising a memory, a processor and an industrial equipment energy efficiency management oriented lightweight security detection program stored on the memory and operable on the processor, the industrial equipment energy efficiency management oriented lightweight security detection program when executed by the processor implementing the steps of the industrial equipment energy efficiency management oriented lightweight security detection method of any of claims 1-12.

Description

Lightweight security detection method, system, terminal and storage medium for energy efficiency management of industrial equipment Technical Field The invention relates to the technical field of intelligent equipment networks in industrial scenes, in particular to a lightweight security detection method, a system, a terminal and a computer readable storage medium for energy efficiency management of industrial equipment. Background Along with the deepening application of the industrial Internet of things technology in power, mines, manufacturing and other industries with high energy consumption endangered, real-time, accurate energy efficiency and safety collaborative management on industrial equipment become key demands for digital transformation and sustainable development of the industry. The field devices represented by the photovoltaic inverter, the mining sensor and the motor controller are accessed into a network through an industrial Internet of things edge gateway, and data communication is performed by widely adopting industrial protocols such as Modbus/OPC UA and the like. These devices are typically deployed under severe conditions such as high temperature, high voltage, and strong electromagnetic interference, and their operating conditions directly affect production safety, energy consumption, and carbon emission levels. Therefore, an intelligent management scheme which is deployed at the edge side and can deeply integrate safety detection and energy efficiency optimization is needed to realize early warning, quick response and energy efficiency optimization closed-loop control on abnormal states of equipment. However, the current technical solutions for energy efficiency and safety management of industrial equipment face a series of key technical bottlenecks on the landing edge side, which makes it difficult to effectively apply in actual complex industrial scenarios: First, the contradiction between edge side resource constraints and detection model computational overhead stands out. The existing mainstream security detection method, such as an anomaly detection model based on a depth automatic encoder, generally has a complex network structure and a huge parameter scale, the memory occupation is often more than 200MB, and the edge gateway with limited computing capacity, storage space and power consumption is difficult to bear. If the system is forcefully deployed, the system load is too high, the response is delayed, and even the normal control function is affected. Therefore, developing a lightweight detection model with extremely small memory occupation and high calculation efficiency is a precondition for realizing real-time safety monitoring of the edge side. Secondly, detection accuracy and reliability are difficult to guarantee in a high-noise industrial environment. Industrial field data acquisition is often accompanied by intense impulse noise, electromagnetic interference and operating mode fluctuations. Traditional single data preprocessing methods (such as Z-score standardization) have limited effects under such non-stationary and strong interference signals, and are extremely easy to cause a large number of false alarms generated by a subsequent detection algorithm, the false alarm rate can exceed 15%, and the false alarms seriously interfere with operation and maintenance judgment and possibly cause unnecessary production interruption. How to design a robust data preprocessing mechanism and a detection algorithm adapting to the industrial data characteristics so as to still keep high detection precision and low false alarm rate under a strong noise background is a difficult problem which needs to be solved by actual deployment. Thirdly, the two large systems of safety protection and energy efficiency management are fractured for a long time, and the synergy is lacking. In conventional industrial systems, the safety system and the energy efficiency management system often operate independently. The safety system usually adopts simple rough power-off or emergency stop measures when serious threat is detected, but the safety is guaranteed, the unplanned production stop and energy waste are caused, and the energy efficiency optimization system is focused on energy-saving scheduling and can ignore the potential safety risk or sub-health state of equipment. The system can not realize optimal energy efficiency on the premise of ensuring safety, and can not deepen the perception of safety state through energy efficiency data feedback, so that a collaborative management framework capable of deeply linking a safety event with equipment energy consumption and operation efficiency is needed. Fourth, there is a lack of quantitative support and compliance demonstration for sustainable development. With global importance of environmental protection and carbon emissions reduction, many energy-intensive enterprises face stringent ESG (environmental, social, and regulatory) and carbon regulator