376 results about "Baseline model" patented technology

The invention relates to the technical field of network security and data protection, and discloses a data leakage prevention method and system based on user behavior perception, and the method comprises the steps: obtaining historical operation data, and constructing a personalized behavior reference library; monitoring a current access behavior in real time by sliding a time window, calculating a deviation degree and triggering anomaly detection; performing multi-level feature analysis on the abnormal behavior and calculating a comprehensive abnormal score; dynamically adjusting the access authority according to the score, recording an abnormal behavior and carrying out relevance matching; optimizing the reference model through feedback learning; and evaluating the credibility and the risk level based on an accurate modeling result, and adaptively adjusting permission configuration. According to the method, the user behavior rule can be accurately captured, the data leakage risk can be efficiently identified, the access permission can be dynamically adjusted, the false alarm rate can be reduced, adaptive protection can be realized, and data security and business smoothness can be guaranteed.

The invention provides a patient vital sign anomaly detection method based on an artificial intelligence technology, and relates to the technical field of data processing, and the method comprises the steps: collecting original sign data of a patient; calculating multi-dimensional characteristic parameters; establishing an individual baseline model, and determining a comprehensive reference interval in the model; the vital sign features monitored in real time are constructed into multi-dimensional feature vectors, the multi-dimensional feature vectors are input into the individual baseline model, and the deviation degree of the real-time multi-dimensional feature vectors in the comprehensive reference interval is calculated; performing a clustering analysis to identify an anomalous aggregation region; performing trend analysis, calculating change direction consistency and continuous change amplitude of the multi-dimensional feature vector, and generating a trend analysis result; calculating an accumulated change index in a continuous time window according to a trend analysis result to obtain a dynamic confidence score; when the dynamic confidence score continuously exceeds an adaptive threshold value, determining that a vital sign abnormal event exists; the autonomy and accuracy of the method for detecting the vital sign abnormity of the patient are improved.

The invention relates to the technical field of multi-modal behavior anomaly detection scheme design, in particular to a multi-modal behavior anomaly detection method and system under the condition that encrypted traffic is not decrypted. The method comprises the following steps: capturing a network encrypted traffic data packet in real time; on the premise that decryption is not carried out, multi-mode non-decryption features such as flow layer statistics, time sequence interaction, encryption handshake and context association are extracted in parallel; establishing a dynamic normal behavior baseline model of each feature based on historical data; the real-time features are compared with the baseline model, comprehensive judgment is carried out through a multi-mode correlation analysis algorithm, and an anomaly detection result is output; and when the abnormal condition is judged, the network control equipment is automatically linked for blocking. The method thoroughly gets rid of dependence on traffic decryption, realizes high-precision and self-adaptive detection and rapid automatic response to abnormal behaviors in encrypted traffic through multi-dimensional feature fusion and dynamic baseline technologies, and effectively solves the problem of failure of a traditional detection technology in an encrypted environment.

The invention discloses an oil and gas storage and transportation oil leakage detection system based on multi-source data analysis, and relates to the technical field of oil leakage detection. An original sensing data set is constructed, the original sensing data set is preprocessed, an alignment data set is generated, a pipeline digital twin baseline model in a leakage-free state is established, and a baseline feature set is generated; according to the method, micro-disturbance active detection and dynamic compliance inversion are implemented, an active feature set is generated, the diffusion trend of potential leakage liquid and an earth surface exposure time window are predicted, earth surface evidence features are generated, multi-modal features are fused, and a leakage event and a response scheme are output. A diffusion closed loop from the underground to the earth surface is realized through multiphase seepage and remote sensing verification, the detection precision is improved by utilizing adaptive fusion and dynamic threshold control, self-learning updating is realized through model recharge, and the method has the advantages of high real-time performance, strong anti-interference and adaptive optimization.

The invention relates to the technical field of big data security analysis and network information security, in particular to a communication data intelligent security supervision system based on big data, which comprises a data acquisition module used for extracting standardized entity behavior feature vectors from multi-source heterogeneous communication data; the baseline modeling module is used for dynamically generating a multi-dimensional behavior baseline portrait based on the historical sequence of the entity behavior feature vectors; the anomaly detection module is used for comparing the real-time entity behavior feature vector with the multi-dimensional behavior baseline portrait so as to calculate a micro anomaly score; the atlas construction module is used for screening the micro-anomaly events according to whether the micro-anomaly score exceeds a preset threshold value or not, and quantifying association confidence among the screened events so as to construct an attack chain atlas; the risk quantification module is used for aggregating the characteristics of the attack chain atlas to determine a systematic risk score; according to the method, the discovery capability and response efficiency of complex attacks such as advanced persistent threats and the like are greatly improved.

The invention relates to the technical field of information flow intelligent recommendation, and discloses a real-time interactive feedback information flow content adaptive optimization method and system.The real-time interactive feedback information flow content adaptive optimization method comprises the steps that user interaction data are collected and subjected to robust preprocessing to obtain standardized interaction feature vectors; constructing a user personalized interaction baseline model to obtain a multi-level behavior reference basis, identifying a real-time interest signal of a user and updating an interest state, dynamically adjusting a recall strategy and executing multi-path recall to obtain a candidate content set, adaptively adjusting a sorting feature weight and executing sorting, and dynamically controlling diversity to execute rearrangement to obtain a final recommendation list, and meanwhile, obtaining an interest prediction score in a sorting stage by adopting a hierarchical reasoning architecture and a model distillation technology. The interest change of the user can be captured in real time, the matching degree of the recommended content and the instant interest of the user is improved, and the user experience and the recommendation effect are improved.

The invention relates to the field of electric power monitoring, in particular to an electric power monitoring system intrusion detection method and system based on flow analysis. The method comprises the following steps: collecting network traffic, analyzing and recombining to obtain structured session data; time sequence behavior features and function code distribution features are extracted to construct a multi-dimensional feature set; inputting the feature set into a compliance rule base and a behavior baseline model in parallel, and respectively outputting a rule matching result and an abnormal deviation degree score; generating a comprehensive threat index by adopting a weighted decision fusion strategy; and when the index exceeds a dynamic threshold value, intrusion is determined and an alarm is given. According to the invention, the problem of insufficient precision and adaptability caused by single feature dimension and isolated detection mechanism is solved.

The invention discloses a data center dynamic air-conditioning load rapid calculation method, which comprises the following steps of: 1, summarizing data center cold load influence factors as an input condition of a calculation model; 2, constructing a standardized data center reference model based on building energy consumption simulation software; 3, based on the building geometrical characteristics and thermal performance of the target data center, correcting the enclosing structure load of the reference model by adopting an area correction method, and constructing an enclosing structure load rapid calculation model; 4, automatically resetting an internal heat source load calculation result, and constructing an internal heat source load rapid calculation model; 5, correcting the fresh air load of the reference model by adopting a volume correction method based on the geometric characteristics of the target data center building, and constructing a fresh air load rapid calculation model; and step 6, forming a target data center dynamic air-conditioning load rapid calculation method, and obtaining a hourly cooling load prediction result required for covering the whole year or a design stage.

The invention relates to the technical field of power system equipment state monitoring and fault diagnosis, in particular to a GIS basin-type insulator operation state evaluation method and system, and the system comprises an asset information and baseline modeling module which is used for building a high-fidelity multi-physics field finite element reference model and generating a defect state-external representation mapping data set; the physical information driven agent model generation module is used for constructing a neural network agent model fusing physical law constraints; the real-time data acquisition and feature extraction module is used for acquiring and processing online monitoring data such as UHF, gas, temperature and vibration; the state inversion and digital twinning calibration module is used for inverting internal defect parameters by adopting a Bayesian inference and MCMC method so as to realize real-time calibration of the model; and the evaluation diagnosis and life prediction module carries out fault mode identification and residual life prediction based on the calibration model.

The invention discloses a pumped storage power station model construction method based on digital twinning, relates to the technical field of power system modeling and intelligent control, and is used for solving the problem of low modeling precision of a unit operation state. According to the method, uniform space-time criterion and asset semantic graph driven twinborn observation frames are constructed, consistent mapping of multi-source heterogeneous data and a pumped storage unit structure is established, fine modeling of a working condition state and a safety boundary is realized, on the basis, an online assimilation mechanism and a sub-scene baseline model set are fused, and the working condition state and safety boundary of the pumped storage unit are improved. Continuous self-correction of hydraulic, electromechanical and control parameters under disturbance of different working conditions is achieved, then a time delay budget-driven hierarchical assembly mechanism is introduced, a high-fidelity model and a data-driven model are dynamically called according to scenes, the prediction precision and response efficiency are improved, the out-of-limit risk is avoided through linkage of uncertainty labeling and a conservative mode, and the reliability of the system is improved. And closed-loop iterative updating is formed based on simulation verification and gray rollback, so that synchronous improvement of model credibility, strategy robustness and power station operation economy is realized.

The invention discloses an industrial network fault intelligent diagnosis method and system based on multi-source data, and the method comprises the steps: synchronously collecting data from a plurality of data sources of an industrial network, and extracting a time sequence statistical feature, a flow entropy feature and a protocol conformity feature to form a multi-dimensional feature vector; establishing a dynamic baseline model by adopting a sliding window online learning method, and calculating a comprehensive anomaly score for anomaly detection; the fault suspicion degree is calculated based on the equipment incidence matrix and the fault propagation model to realize fault source positioning; carrying out fault type identification and root cause analysis by adopting Bayesian reasoning and a knowledge rule base; and outputting a structured diagnosis report containing the fault source, the type, the root cause and the disposal suggestion. According to the invention, early warning, accurate positioning and intelligent diagnosis of industrial network faults are realized, and the operation and maintenance efficiency, safety and reliability of the industrial control network are significantly improved.

The invention relates to the technical field of network security threat detection and attack traceability, and discloses an encrypted traffic behavior and protocol multi-dimensional analysis traceability method, and establishes an encrypted traffic multi-dimensional analysis traceability system. The system is internally provided with a traceability knowledge base, a multi-dimensional data source acquisition module, a multi-dimensional feature extraction module, a multi-dimensional analysis traceability module, a dynamic baseline modeling module, an adversarial training module, an intelligent association study and judgment module and a feedback optimization module, the traceability knowledge base provides knowledge support for the follow-up process, and the multi-dimensional data source acquisition module outputs standardized time sequence data; the multi-dimensional feature extraction module generates a feature vector, a traceability result of the multi-dimensional analysis traceability module, an abnormal identifier of the dynamic baseline modeling module and an adversarial sample feature of the adversarial training module jointly serve as input of the intelligent association study and judgment module, and the intelligent association study and judgment module summarizes a study and judgment report. Models and parameters of all the modules are fed back through the feedback optimization module, and a closed loop of collection, analysis, study and judgment and optimization is formed.

The invention discloses a social engineering self-adaptive dynamic protection method based on post risk assessment, relates to the technical field of network security, and solves the problem that in the prior art, the aspects of employee security awareness modeling, multi-modal data fusion and personalized security policy recommendation still have obvious deficiencies. The method comprises the following steps: firstly, constructing an electric power post knowledge graph, and carrying out structured modeling on four entities of posts, permissions, assets and vulnerabilities and association relationships thereof to form a global semantic network; secondly, a post risk baseline model is established, a normal behavior mode is defined by quantifying inherent risks of posts and analyzing historical operation logs, and a static and dynamic combined risk reference is formed; and finally, designing a post exclusive feature extractor, fusing personal behavior data of the employees with semantic constraints and risk baselines of the knowledge graph, and generating highly personalized feature vectors, thereby laying a foundation for subsequent real-time monitoring of abnormal behaviors of the employees, quantification of safety consciousness and providing of differentiated protection.

The invention is suitable for the field of sensor data analysis, and provides a sensor data analysis method and system based on artificial intelligence, and the method comprises the steps: collecting multi-dimensional historical sensor data of normal operation of equipment, and obtaining a standardized historical feature matrix through preprocessing and feature extraction; constructing an equipment health reference model based on the matrix, and establishing multi-dimensional feature dynamic association; collecting equipment data in real time, and extracting features according to the same process to obtain real-time feature vectors; combining the reference model with a known fault case library, constructing a fault evolution prediction model through adversarial training, and inputting real-time vectors to generate a future multivariable sensor data prediction sequence; and calculating a fault deterioration acceleration and a predicted time for reaching a safety threshold based on the prediction sequence, and comparing to generate a risk level signal and a maintenance strategy suggestion. The whole-process intelligent monitoring analysis of the equipment is realized, and the operation and maintenance accuracy and foresight are improved.

The invention belongs to the technical field of resource aggregation, and discloses a distributed resource aggregation operation boundary construction method, system, device and medium, and the method comprises the steps: constructing a distributed resource aggregation operation feasible region model based on an aggregation operation range, and integrating the model with a balance unit day-ahead scheduling power baseline model to obtain a distributed resource deterministic aggregation model; analyzing the influence of the new energy unit output and the source load uncertainty of the load power utilization power on the distributed resource determinacy aggregation model, and constructing a balance unit baseline operation model; and solving the distributed resource aggregation operation feasible region model, and determining a distributed resource aggregation operation boundary considering uncertainty. According to the invention, the IGDT theory is introduced to characterize the uncertainty of the output at the two sides of the source load, DER day-ahead scheduling baseline models with different risk preferences are constructed, and on the basis, the solution of the DER operation feasible region boundary considering the uncertainty of the source load is realized.

The invention relates to the field of tool management, and provides a coal-fired power plant tool intelligent management method and system. The method comprises the following steps: carrying out digital identity construction on a purchasing tool to obtain a tool digital identity file comprising a unique identification code of the tool; constructing a tool circulation record based on the tool digital identity file, and obtaining a block chain traceability record; monitoring an inventory tool state through the block chain traceability record to obtain a tool behavior baseline model; based on the tool behavior baseline model, attenuation prediction and maintenance scheduling are carried out on an inventory tool, and a tool health attenuation trend prediction result and an optimal maintenance scheduling scheme are obtained; and performing decision analysis based on the optimal maintenance scheduling scheme to obtain a tool management decision instruction set. According to the method and the system, non-tampering recording of tool full-life-cycle data, accurate health state prediction and scientific decision support are realized, and the management efficiency and the data credibility are improved.

The invention relates to the technical field of Internet of Things security, in particular to a network attack early warning method based on an abnormal behavior map. Comprising; constructing a subject-behavior-resource three-layer heterogeneous graph, establishing a low-activity baseline model based on historical data, calculating weak association strength between nodes by adopting wavelet decomposition and a signal enhancement algorithm, and identifying a latent attack chain; the attack chain features are converted into multi-dimensional vectors, and an attack collaboration time window and intensity are predicted through a vector convergence angle and a distance change rate; the vulnerability of the graph topology is quantified, the edge weight is dynamically adjusted by using a gradient descent algorithm, and high-risk nodes are isolated; the convergence parameters are fed back to the baseline model and correlation strength calculation to form closed-loop optimization; and calculating a comprehensive threat index based on three-dimensional weighting of the latent threat intensity, the convergence urgency degree and the topology anti-attack capability, and realizing layered early warning. According to the method, the APT attack detection rate is remarkably improved, and the early warning response time is shortened.

The invention relates to the field of artificial intelligence control, in particular to a world model and reinforcement learning collaborative optimization method based on constraint perception gradient projection, which comprises the following steps: constructing a world model based on Transform architecture, performing unified modeling on an environment state, system dynamics and a reward function, and training the model based on reference model data to predict an environment future state; training a reinforcement learning strategy based on a virtual track and a reward signal generated by the world model, and calculating strategy gradient information; an explicit gradient projection operator of constraint perception is designed, a strategy gradient is corrected in real time according to future constraint conditions predicted by the world model, and it is ensured that the gradient updating direction meets the safety constraint and the optimization target at the same time; and synchronously updating world model parameters and reinforcement learning strategy parameters by using the gradient after projection correction to realize collaborative optimization of the world model parameters and the reinforcement learning strategy parameters. According to the method, the key problems of local optimal trap, low convergence speed, insufficient stability and the like in the collaborative optimization process of a traditional method are solved.

The invention relates to the technical field of computer network security, and discloses a server security protection method and system based on a dynamic gateway policy. The method aims at solving the problems that a traditional security gateway is low in protection efficiency, high in false alarm rate and difficult to cope with complex dynamic network attacks due to static rule matching, single flow analysis dimension, strategy updating lagging and poor response collaboration. The method comprises the following steps: receiving an inbound request through a security gateway agent, extracting a multi-dimensional feature, and constructing a structured traffic feature vector; and establishing a dynamic behavior baseline model based on historical normal traffic. According to the scheme, self-adaptive adjustment and minute-level response of the security policy are realized, the false alarm rate is reduced, the detection accuracy of hidden attacks is improved, the continuity of core services under high load is guaranteed, a self-evolution closed loop of'detection-response-learning 'is formed, and the intelligence and reliability of overall protection are enhanced.

The invention relates to the technical field of liquid chromatographic analysis, and discloses a data acquisition method and system of a liquid chromatograph. The method comprises the following steps: extracting chromatograms without solvent residues and with known solvent residues from a historical database, and training by using a self-encoding neural network and a convolutional neural network to generate a baseline reference model and a ghost peak early warning model; obtaining solvent parameters of the previous analysis method, and calculating the residual solvent concentration through residual solvent information; inputting the residual solvent concentration into the baseline reference model and the ghost peak early warning model, and calculating to obtain a baseline drift prediction correction value and a ghost peak prediction risk area; pre-correcting the chromatographic signal acquired in real time to obtain a purified chromatographic signal; performing qualitative and quantitative analysis on the purification signal according to dynamic allocation resources of the ghost peak risk area to generate final chromatographic data; the data interference of the mobile phase residue is effectively eliminated, and the accuracy of data acquisition is improved.

The invention discloses a blockage prediction and alarm system for an oil absorption filter, and relates to the technical field of prediction and alarm, the system comprises a data acquisition module, a data processing module, a blockage quantification module and a prediction and alarm module, the data acquisition module periodically acquires the running state data of the filter; the data processing module adopts a Pauta criterion to eliminate abnormal data, calculates a unit flow pressure difference and a pressure difference change rate, and eliminates oil temperature viscosity interference through a viscosity-temperature characteristic compensation model to obtain a standard pressure difference; the blockage quantification module obtains an ideal pressure difference through a clean filter element baseline model, calculates an original blockage coefficient, and corrects through a time domain residual fusion mechanism to obtain a residual calibration blockage coefficient; the prediction and alarm module constructs a blockage time sequence prediction model based on the long-short-term memory network, generates a blockage coefficient prediction value in the next acquisition period to decide whether to trigger sound-light alarm, and realizes the preposed sensing and alarm of the blockage risk of the oil absorption filter.

The invention provides an intelligent time sequence security analysis method and device based on a large language model, and relates to the technical field of information security, and the method comprises the steps: collecting an original log from a multi-source heterogeneous system to obtain a standardized log event; inputting the standardized log event into a preset large language model to generate a historical semantic log event; constructing a user digital portrait based on the historical semantic log event, and establishing a dynamic behavior baseline model associated with the user digital portrait by using a long-short term memory neural network model; obtaining a semantic log event stream output by the large language model in real time, and querying a dynamic behavior baseline model of a user corresponding to the semantic log event stream; and evaluating whether the current semantic log event is an abnormal event: if the current semantic log event is an abnormal event, calling a large language model to generate a semantic alarm, and injecting the semantic alarm into a rule engine. The technical problems that a traditional static rule system is poor in adaptability, high in false alarm rate and high in maintenance cost are effectively solved.

The invention belongs to the technical field of building energy conservation, and provides a roof system energy efficiency optimization management method and device based on digital twinning. The method comprises the following steps: acquiring real-time monitoring data, determining a digital twinborn body, and determining an energy consumption evaluation index; calculating a comprehensive energy consumption value to determine a target control parameter generation optimization scheme; a control instruction is generated and sent to the execution terminal; calculating a difference coefficient to calibrate the digital twin; and performing energy-saving transformation and equipment updating by taking the calibrated digital twin as a reference model. According to the invention, accurate monitoring and optimal control of the energy consumption of the roof system can be realized, the energy utilization efficiency is improved, and the building operation cost is reduced.

The invention provides a mine post personnel behavior identification method and system based on artificial intelligence, and the method comprises the steps: constructing a mine post dynamic behavior logic reference model, combing post personnel standard operation behavior and process node association logic, and dynamically adjusting the weight. Acquiring personnel behavior acquisition data, and performing adaptive processing on the data and the model to obtain to-be-identified behavior data; and calling a pre-trained AI behavior difference adaptation model to carry out bidirectional feature adaptation, and generating dynamically optimized behavior feature adaptation parameters. And based on the parameters, performing enhancement processing on the to-be-identified behavior data, mining hidden logic association, and generating an enhanced behavior difference feature set. And performing compliance verification according to a preset rule base to obtain a compliance verification result. And finally, generating a post behavior control instruction according to the result and sending the post behavior control instruction to a mine operation control terminal. According to the invention, behaviors of mine post personnel can be accurately identified, and operation safety and high efficiency are guaranteed.

The invention discloses an energy consumption optimization method for integrating building structure parameters. The method comprises the following steps: acquiring geometric morphology, enclosure structure, equipment configuration, personnel activity and external environment data of a building in real time by using an intelligent sensor and a building information model; based on the collected data, calculating thermal load and energy consumption distribution of the building, extracting energy efficiency performance indexes, and establishing an energy efficiency baseline model; the thermal performance of the building envelope structure is dynamically adjusted by utilizing a self-adaptive envelope structure technology so as to adapt to external climate change and internal thermal load requirements; based on the optimized enclosure structure characteristics, an energy recovery system is started, and waste heat and air flow energy in the building are converted into available energy; in combination with energy efficiency data fed back in real time, the intelligent scheduling system dynamically adjusts the operation strategy of the building equipment, optimizes the energy efficiency, reduces unnecessary energy consumption and improves the comfort level; the building energy efficiency is evaluated regularly, and it is ensured that the building continuously maintains the optimal energy efficiency under different seasons, environment changes and use requirements.

The invention relates to the technical field of heating and ventilation fault recognition, in particular to a heating and ventilation system fault recognition method based on operation state offset modeling and a control system. The method comprises the following steps: acquiring steady-state operation data of the heating and ventilation system and extracting actual parameters of equipment; inputting the environment and load parameters into the lightweight integrated tree base line model, and outputting an equipment operation parameter reference value; performing baseline drift correction on the reference value in combination with the accumulated operation duration of the equipment, calculating a dynamic difference value between an actual value and a calibration reference value, and constructing an operation state offset characteristic matrix; and inputting the matrix into a preset fault diagnosis rule base, matching a hidden fault in combination with the dynamic tolerance interval and the multi-dimensional Boolean logic, and giving an alarm. Performing reference compensation by using the accumulated operation time length in combination with a linear or exponential progressive attenuation model; on the premise of not consuming high computing power to re-train an underlying algorithm model, the interference of natural physical aging of the equipment on residual calculation is eliminated.

A defect detection method and a model transfer method. The model transfer method comprises: acquiring a source-domain image and a corresponding annotation, and a baseline model obtained from a source domain; acquiring target-domain images, which are fully annotated, partially annotated, or unannotated; inputting the annotation of the source-domain image into a denoiser to obtain a denoised annotation; inputting the unannotated target-domain images into an initial baseline model to obtain second target-domain prediction results, performing data augmentation on the unannotated target-domain images to obtain second target-domain augmented images, and using the second target-domain prediction results as pseudo-annotations of the second target-domain augmented images; and using the source-domain image and the corresponding denoised annotation, the annotated target-domain images and the corresponding annotations, and / or the second target-domain augmented images and the corresponding pseudo-annotations to train the baseline model, so as to obtain a transfer model. The method can complete training by using a small number of annotated target-domain images, thereby solving the problem of model transfer performance being poor in the case of insufficient target-domain images.

The invention relates to the technical field of home safety, and discloses an intelligent door lock identification and hierarchical response control method, which comprises the following steps: acquiring multi-modal data in a learning mode to train and construct an authorized behavior baseline model; in the working mode, the current feature vector is extracted in real time, the abnormal score of the current feature vector relative to the baseline model is calculated, when the abnormal score is larger than a threshold value, potential threats are judged, the potential threats are responded, the micro actuator is driven to apply physical disturbance signals, and feedback signals are synchronously collected; analyzing the time sequence characteristics of the feedback signal to discriminate whether the event category of the potential threat is benign or malicious, and executing a hierarchical response strategy; and dynamically correcting the model based on the event category, if the model is benign, updating the authorized behavior baseline model, and if the model is malicious, extracting malicious feature signatures and storing the malicious feature signatures in a high-priority threat signature library. Through active discrimination and closed-loop adaptive learning, the recognition accuracy is significantly improved, and the false alarm rate is reduced.

A pruning distillation method based on a channel dimension algorithm comprises the following steps: S1, model improvement: providing an algorithm improvement scheme for the problem that feature fuzzy targets and small target objects are difficult to detect in road infrared target detection by a reference model YOLOv11n, and obtaining an improved algorithm ITD-YOLO; s2, performing structural pruning of channel dimensions on the ITD-YOLO algorithm by using an amplitude-based layer adaptive LAMP pruning method to obtain a lightweight algorithm model ITD-YOLO-P so as to reduce the parameter quantity and the calculation quantity of the improved algorithm; s3, a CWD method based on channel dimensions is proposed for intensive prediction tasks for feature distillation, so that pruning precision loss achieved based on the LAMP method is recovered; s4, the ITD-YOLO-D obtained after pruning and distillation, the ITD-YOLO obtained before pruning and distillation and a current mainstream lightweight algorithm are used for conducting a contrast experiment, and the pruning and distillation effect is verified. The invention provides a pruning distillation method based on a channel dimension algorithm. The pruning distillation method is used for solving the problem that parameter quantity and calculated quantity rise due to algorithm improvement.

The invention relates to a multi-security protection method for an electric power application shopping mall, and the method comprises the following steps: S1, collecting and recognizing the fingerprint information of equipment, then carrying out the credibility evaluation of the equipment, and outputting an authenticated and credible user session voucher; s2, dynamically judging a function and a data range which are allowed to be accessed in combination with a shopping mall permission model, and verifying whether the permission conforms to the current application or not in real time; s3, when a user accesses the application, deploying a runtime application self-protection component according to each running application, monitoring the running state of the application in real time, and obtaining monitoring content; s4, establishing an application normal behavior baseline based on a machine learning algorithm, and identifying an abnormal behavior by comparing a real-time behavior with a baseline model according to the monitoring content; and S5, when the security risk is detected, the system automatically executes corresponding response measures according to the risk level. The reliability of the electric power application shopping mall is effectively improved.