Multi-stage safety early warning and linkage disposal system of electricity-hydrogen complementary energy station

Abstract

The invention relates to the technical field of safety monitoring, and discloses a multistage safety early warning and linkage processing system for an electricity-hydrogen complementary energy station, and the system comprises a monitoring collection module which is used for deploying a multi-source sensor network in the electricity-hydrogen complementary energy station, collecting key parameter data in real time, and generating a real-time data set comprising a timestamp, a sensor ID and a parameter value; the anomaly detection module is used for performing multi-dimensional analysis by adopting an anomaly detection algorithm according to the real-time data set, identifying an anomaly mode and outputting a graded early warning signal; the evaluation grading module is used for carrying out dynamic risk evaluation in combination with the early warning signal and the system state, and outputting a quantitative risk grade and a disposal suggestion; the linkage triggering module is used for automatically matching and triggering a corresponding plan according to the risk level and outputting a specific linkage control instruction; an execution feedback module; according to the invention, the emergency response speed is improved, the false alarm rate caused by normal working condition fluctuation is reduced, and fundamental conversion from passive response to active early warning is realized.

Description

Technical Field

[0001] This invention relates to the field of safety monitoring technology, and more specifically to a multi-level safety early warning and coordinated response system for electric-hydrogen complementary energy stations. Background Technology

[0002] Traditional safety solutions for hydrogen-powered energy stations employ a decentralized monitoring architecture. The monitoring layer relies on single-point alarms with fixed thresholds, the control layer only has rigid, isolated, and simple device linkage functions, and the management layer depends entirely on human experience for status assessment and emergency response. The entire system lacks a unified risk assessment engine and information correlation analysis capabilities, resulting in a passive, reactive protection system.

[0003] Traditional solutions suffer from the following key drawbacks: They have weak early warning capabilities and a high false alarm rate; detection methods based on fixed thresholds cannot adapt to dynamic system operation conditions, easily generating numerous invalid alarms due to normal fluctuations, while lacking the ability to identify slowly evolving potential risks and complex coupled faults; the system lacks a global risk assessment mechanism, with each alarm point existing in isolation, making it difficult for operators to quickly determine the overall risk level and response priority from a large number of scattered alarms, easily leading to delayed decisions due to information overload in emergency situations; the response process heavily relies on human experience, resulting in slow response speed and poor consistency, with different operators potentially making drastically different decisions, and human error easily occurring under pressure; the system lacks self-optimization capabilities, its thresholds and logic are statically preset, unable to learn and optimize from historical events, causing system performance to gradually lag behind actual needs over time. These shortcomings can pose serious safety hazards and operational risks in complex and high-risk scenarios such as electro-hydrogen complementary systems.

[0004] Therefore, a multi-level safety early warning and coordinated response system for electric-hydrogen complementary energy stations is needed to solve the above problems. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides a multi-level safety early warning and linkage response system for electric-hydrogen complementary energy stations to solve the problems existing in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a multi-level safety early warning and linkage response system for an electric-hydrogen complementary energy station, comprising: a monitoring and acquisition module, used to deploy a multi-source sensor network within the electric-hydrogen complementary energy station, collect key parameter data in real time, and generate a real-time dataset including timestamps, sensor IDs, and parameter values; The anomaly detection module is used to perform multi-dimensional analysis based on real-time datasets using anomaly detection algorithms, identify anomaly patterns, and output graded early warning signals. The assessment and grading module is used to conduct dynamic risk assessment by combining early warning signals and system status, and output quantitative risk levels and handling recommendations. The linkage triggering module is used to automatically match and trigger the corresponding contingency plan according to the risk level and output specific linkage control instructions; The execution feedback module is used to evaluate the actuator after it has responded to the control command and completed the disposal, and to monitor the results in real time and feed them back to the system to start a new monitoring cycle.

[0007] The key parameters include: hydrogen concentration, hydrogen storage tank pressure, temperature, electrolyzer operating status, fuel cell output current and voltage, electrical load, and ambient humidity.

[0008] The anomaly detection algorithm specifically involves: constructing a hybrid model integrating rule-based and machine learning approaches. Rule-based anomaly detection requires creating a configurable rule base and using a rule engine to match anomaly types to the input data stream in real time. Machine learning-based anomaly detection requires training an isolated forest model using historical normal operation data, extracting feature vectors from the real-time dataset, inputting them into the trained model, and outputting an anomaly score. Simultaneously, a dynamic threshold is set. When the anomaly score exceeds the threshold, the system determines that an anomaly type identified by the machine learning model has occurred. The dynamic threshold is achieved by continuously maintaining a set of anomaly scores within a sliding time window and calculating the specified high quantile of this set in real time as the current threshold.

[0009] The multi-dimensional analysis method is as follows: Based on the rule-first principle, the detection result based on the rule will immediately generate a preliminary warning signal; a machine learning model is used for confirmation to check whether the machine learning model also gives an abnormal score in the same or related areas; when the rule engine triggers the warning, if the score given by the machine learning model exceeds its dynamic threshold, the warning is confirmed to be valid; when the rule engine triggers the warning, if the score given by the machine learning model does not exceed its dynamic threshold, a final warning will not be issued immediately, and a short observation period will be initiated. If the rule conditions continue to exist during the observation period, a warning will be issued even if the score given by the machine learning model does not exceed the dynamic threshold; if the rule engine does not trigger the warning, but the score given by the machine learning model exceeds its dynamic threshold, an "unknown pattern anomaly" warning will be generated.

[0010] The dynamic risk assessment method is as follows: a risk assessment model is constructed, a risk base value matrix is ​​predefined, and the matrix is ​​indexed by the warning level and the anomaly type; state factors are introduced, including operation mode factors, equipment redundancy factors, cascading impact factors, and environmental and personnel factors; a state coefficient is calculated by weighted summation of multiple state factors; the state coefficient is multiplied by the risk base value and the state to obtain the final risk value; and the calculated risk value is mapped to a preset risk level.

[0011] The method for triggering the corresponding contingency plan is as follows: a contingency plan library is preset, and each contingency plan includes triggering conditions, a sequence of handling actions, execution priority, and a mutex lock; the risk level and the type of exception are matched in the contingency plan library, and the matching process supports fuzzy matching and precise matching; once the optimal contingency plan is matched, the module will generate all the control instructions defined in the contingency plan in sequence.

[0012] The method for re-evaluating after the treatment is completed is as follows: a short-term effect monitoring window is preset to monitor key parameters directly related to the treatment action in real time, analyze the changing trends of these parameters, and determine whether the treatment action has produced the expected effect; the execution result status and effect verification data are used as input to rerun the risk assessment model to calculate a new risk value. If the risk value drops significantly and is below the threshold, it is determined as "treatment successful"; if the risk value remains high or even rises, it is determined as "treatment failed" or "the effect is poor".

[0013] The real-time monitoring results are fed back to the system in the following way: if the handling is successful, the system records all data of this event, generates a complete report, and restores the system status to normal monitoring mode; if the handling fails or the effect is not good, the module will immediately generate a new warning and feed it back directly to the anomaly detection module to trigger system upgrade handling.

[0014] The technical effects and advantages of this invention are as follows: 1. This invention effectively overcomes the inherent defects of traditional solutions by constructing a multi-level intelligent early warning and coordinated response system. The system adopts a hybrid detection model based on rules and machine learning, combined with a dynamic threshold algorithm, which significantly improves the accuracy of early warning. This not only greatly reduces the false alarm rate caused by fluctuations in normal operating conditions, but also accurately identifies slowly evolving faults and complex risks that are difficult to detect by traditional methods, achieving a fundamental shift from passive response to proactive early warning.

[0015] 2. This invention introduces a quantitative risk assessment mechanism, which transforms discrete alarm signals into a unified, quantitative risk level through comprehensive calculation of the risk base value matrix and dynamic state coefficients. This effectively solves the information silo problem in traditional systems, providing operators with clear risk situation awareness and decision support for prioritizing responses, and avoiding decision-making delays caused by information overload in emergency situations.

[0016] 3. This invention achieves automated coordination from risk assessment to response execution by pre-setting a structured contingency plan library. Once a plan is triggered, the system automatically generates and issues precise control commands, completely changing the traditional response model that relies on human experience. This not only greatly improves the speed of emergency response but also ensures the consistency and reliability of response measures, effectively avoiding human error.

[0017] 4. The unique closed-loop feedback and self-optimization mechanism of this invention constitutes the core advantage of the system. The execution feedback module not only verifies the treatment effect but also re-injects the result data into the risk assessment and detection module, forming a continuous optimization closed loop. This enables the system to learn autonomously from historical events, and the threshold parameters and detection model can continuously improve themselves, always maintaining optimal performance. Attached Figure Description

[0018] Figure 1 This is a system structure block diagram of the present invention.

[0019] Figure 2 This is a flowchart of the collaborative decision-making process for the hybrid model of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Reference Figure 1 This invention provides a multi-level safety early warning and linkage response system for an electric-hydrogen complementary energy station, including a monitoring and acquisition module, an anomaly detection module, an assessment and grading module, a linkage triggering module, and an execution feedback module.

[0022] In a more specific application of the present invention, the monitoring and acquisition module is used to deploy a multi-source sensor network within the electric-hydrogen complementary energy station to collect key parameter data in real time and generate a real-time dataset including timestamps, sensor IDs, and parameter values.

[0023] Key parameter data include: For hydrogen concentration, electrochemical or catalytic combustion sensors are used, installed at leakage risk points such as hydrogen pipeline interfaces and hydrogen storage tank valves to collect concentration values ​​and simultaneously monitor high-concentration alarm signals; For hydrogen storage tank pressure, absolute pressure gauge sensors are used, directly installed at the hydrogen storage tank interface or inlet / outlet gas pipelines to collect real-time pressure data inside the tank and match the design pressure range of the hydrogen storage tank; For temperature, platinum resistance, thermocouple, or integrated temperature and humidity sensors are used, attached to the electrolyzer, fuel cell equipment components, or installed in an open area within the station to collect equipment and ambient temperature; For electrolyzer operating status, current and voltage sensors and liquid level sensors are used, installed in the electrolyzer power supply circuit and electrolyte storage tank to collect operating parameters and determine "normal / fault" status; For fuel cell output current and voltage, high-precision DC sensors are used, connected in series / parallel in the fuel cell output circuit to collect real-time current and voltage data and calculate output power; For electrical load, intelligent power sensors are used, installed in the main power supply circuit or load input terminal to collect voltage and current, calculate and output load data including active power.

[0024] The anomaly detection module is used to perform multi-dimensional analysis based on real-time datasets using anomaly detection algorithms, identify anomaly patterns, and output graded early warning signals.

[0025] The anomaly detection algorithm specifically involves constructing a hybrid model integrating rule-based and machine learning approaches. Rule-based anomaly detection requires creating a configurable rule base, categorized by parameter type (e.g., hydrogen safety, equipment operation, and risk level, e.g., emergency / important / warning), and supporting periodic review and dynamic updates, such as adjusting pressure thresholds based on equipment aging. A rule engine is used to match anomaly types to the input data stream in real time, associating the matching with the device location information corresponding to the sensor ID. Machine learning-based anomaly detection requires preprocessed historical normal operation data, including data cleaning, outlier removal, and feature engineering, such as extracting parameter fluctuation variance and time-series trend features, and training an isolated forest model. The real-time dataset is then used to extract feature vectors using the same feature engineering logic, which are input into the trained model to output a quantified anomaly score. A higher score indicates a more severe anomaly, ranging from 0 to 100. Simultaneously, a dynamic threshold is set. This dynamic threshold is achieved by continuously maintaining a set of anomaly scores within a sliding time window. The window size is set according to the parameter fluctuation frequency, such as 5 minutes for hydrogen concentration and 10 minutes for temperature. A specified high quantile of this set is calculated in real-time, such as the 99th quantile for high-risk parameters and the 95th quantile for ordinary parameters, serving as the current threshold. When the anomaly score exceeds this threshold, the system determines that an anomaly type identified by the machine learning model has occurred.

[0026] Based on the rule-first principle, the rule-based detection results immediately generate a preliminary warning signal with a risk level, such as "emergency" for exceeding the hydrogen concentration limit. A machine learning model is used for confirmation, checking whether it also gives an abnormal score in the same or related areas, such as the same hydrogen storage tank or the same power supply circuit. When the rule engine triggers a warning, if the score given by the machine learning model exceeds its dynamic threshold, the model result is overlaid to confirm the warning's validity and upgrade its level. If the rule engine triggers a warning but the score given by the machine learning model does not exceed its dynamic threshold, a final warning will not be issued immediately. Instead, an observation period is initiated based on the warning level (e.g., 1 minute for "emergency" warning, 5 minutes for "alert" warning). If the rule conditions persist during the observation period (e.g., three consecutive data collections all meet the rule threshold), a warning is issued even if the machine learning model score does not exceed the dynamic threshold. If the rule engine does not trigger a warning, but the machine learning model score exceeds its dynamic threshold, an "unknown pattern anomaly" warning is generated. This warning must include a snapshot of the time-series change curve of the abnormal parameters for manual investigation.

[0027] The assessment and grading module is used to conduct dynamic risk assessments by combining early warning signals and system status, and outputs quantitative risk levels and handling recommendations.

[0028] The dynamic risk assessment method is as follows: A risk assessment model is constructed, pre-defined as a risk baseline matrix. The matrix is ​​indexed by warning level and anomaly type, clearly defining the baseline range for different combinations, such as 8-10 for emergency + hydrogen leak and 1-3 for alert + humidity anomaly. State factors are introduced, including operating mode factors (e.g., 1.2 for full load operation, 1.0 for standby, and 0.8 for maintenance mode), reflecting the current system load pressure; equipment redundancy factors (0.8 for redundant equipment and 1.2 for no redundancy), reflecting fault tolerance capability; cascading impact factors (1.5 for ≥3 related equipment and 0.9 for ≤1), quantifying the risk of anomaly spread; and environmental and personnel factors (1.3 for densely populated areas + severe weather and 0.7 for unoccupied areas + normal weather), combining external influences.

[0029] A state coefficient is calculated by weighted summation of multiple state factors, with weights allocated according to factor importance. For example, the cascading impact factor is 0.3, the environmental and personnel factor is 0.25, the operating mode factor is 0.2, and the equipment redundancy factor is 0.25, with a total sum of 1. The state coefficient is then multiplied by the risk baseline to obtain the final risk value. Based on the calculated risk value, it is mapped to a preset risk level, such as levels 1-3 for low risk, levels 4-6 for medium risk, and levels 7-10 for high risk. Differentiated handling suggestions are then output for different levels and anomaly types, such as regular inspections for low risk, partial shutdown for investigation for medium risk, and emergency shutdown plus activation of the emergency plan for high risk.

[0030] Meanwhile, the model supports a dynamic update mechanism: the risk value is recalculated every 5 minutes based on the latest warning signals and system status data. If the risk level jumps by 2 levels or more, such as from medium risk to high risk, an upgrade reminder is automatically triggered. The weights and base value matrices of the status factors can be periodically iterated and optimized based on the historical handling effects, such as adjusting the weights of the chain reaction factors by analyzing past high-risk events.

[0031] The linkage triggering module is used to automatically match and trigger the corresponding contingency plan according to the risk level and output specific linkage control instructions.

[0032] The corresponding contingency plan is triggered as follows: a structured contingency plan library is preset, and each contingency plan includes triggering conditions, a clear risk level range, anomaly type label, and associated equipment ID, such as "high risk + abnormal pressure in hydrogen storage tank + tank A"; a sequence of actions is divided into steps, each step including the execution object, operation parameters, and timeout threshold; execution priority and mutual exclusion lock marking conflicting operations are specified, such as "only one action instruction is allowed at the same time for the valve of hydrogen storage tank A".

[0033] The system matches risk levels, anomaly types, and associated equipment information against the contingency plan database. The matching process supports both precise and fuzzy matching for elements that fully meet the triggering conditions, such as similar anomaly types or overlapping risk level ranges. The optimal plan is selected based on a matching score (0-100 points). Once the optimal plan is matched, the module generates all control commands defined in the plan in the order of steps and receives real-time feedback from the equipment. If a step times out, a retry mechanism is automatically triggered or the system jumps to the emergency backup step in the plan. If multiple plans with the same priority are matched, they are executed in parallel after conflict resolution using a mutex lock. When a higher-priority plan is triggered, the lower-priority plan is automatically paused.

[0034] The contingency plan library supports dynamic maintenance, allowing manual addition of temporary contingency plans, such as special handling during equipment maintenance, or periodic optimization of action sequences and parameters based on historical execution logs to ensure instruction execution efficiency.

[0035] The execution feedback module is used to evaluate the actuator after it has responded to the control command and completed the disposal, and to monitor the results in real time and feed them back to the system to start a new monitoring cycle.

[0036] The post-response evaluation method is as follows: a short-term effect monitoring window is preset, adapted to the anomaly type and response action. For example, a 30-second window is set for hydrogen leak response, and a 5-minute window is set for equipment parameter adjustment. This avoids misjudgment due to excessively short windows or delays due to excessively long windows. Key parameters directly related to the response action are monitored in real time: the monitoring frequency is increased to 2-3 times that of conventional data collection, such as once every minute normally, and once every 20 seconds during the monitoring period. At the same time, parameter verification thresholds are clearly defined. For example, after hydrogen leak response, the concentration needs to be monitored to drop below the safe value of 0-100ppm. Multi-parameter cross-verification is also supported. For example, after handling abnormal pressure in hydrogen storage tanks, pressure and temperature changes are verified simultaneously to eliminate deviations from a single parameter.

[0037] Analyze the changing trends of these parameters to determine whether the treatment action has produced the expected effect; take the execution result status, such as "valve closed" or "fan started", and the effect verification data, including parameter time series curves and threshold comparison results, as input, and rerun the risk assessment model to calculate the new risk value; if the risk value decreases by ≥50% compared to before the treatment and falls into the low-risk range, it is judged as "treatment successful"; if the risk value does not change or decreases by <20%, it is judged as "ineffective"; if the risk value increases, it is directly judged as "treatment failed".

[0038] The real-time monitoring results are fed back to the system in the following ways: If the handling is successful, the system automatically records a full data report of the event, including the event trigger time, warning level, handling action sequence, parameter change curve, and risk value fluctuation trend. The report supports querying and tracing by device ID and time range, and restores the system status from "emergency monitoring mode" to normal monitoring mode. If the handling fails or is ineffective, the module immediately generates a new warning, which includes "failure-related information," such as parameters that do not meet the standards and the execution status of handling actions. This information is directly fed back to the anomaly detection module, triggering system upgrade handling, such as calling a higher priority plan or increasing the intensity of handling actions. At the same time, a "manual intervention prompt" is pushed to the operation and maintenance terminal, and a snapshot of the current abnormal parameters is displayed to help quickly locate the problem.

[0039] The module supports the marking of the handling effect review: the human can mark the reasons for "poor effect" events, such as the deviation of the plan parameters or the failure of the implementing agency. These marked data will serve as the basis for optimizing the risk assessment model and the handling plan library, forming a closed loop iteration.

[0040] Through the above description of the embodiments, those skilled in the art can clearly understand that the various embodiments of this application can be implemented by means of software or software combined with necessary general-purpose hardware platforms, and of course, they can also be implemented by hardware functions; based on this understanding, the technical solution of this application, in essence or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions to cause a computer device, such as including but not limited to a personal computer, server, or network device, to execute all or part of the steps of the method described in any embodiment of this application.

[0041] The foregoing describes exemplary embodiments of this application. It should be understood that the above exemplary embodiments are not restrictive but illustrative, and the scope of protection of this application is not limited thereto. It should be understood that those skilled in the art can make modifications and variations to the embodiments of this application without departing from the spirit and scope of this application, and such modifications and variations should be within the scope of protection of this application.

Claims

1. A multi-stage security early warning and linkage disposal system for an electric-hydrogen complementary energy station, characterized in that, Specifically comprising: A monitoring collection module for deploying a multi-source sensor network inside the electric-hydrogen complementary energy station, collecting real-time key parameter data, and generating a real-time data set including time stamp, sensor ID, and parameter value; An anomaly detection module for multi-dimensional analysis using an anomaly detection algorithm based on the real-time data set, identifying abnormal patterns, and outputting graded early warning signals; An evaluation grading module for dynamic risk assessment combining early warning signals and system status, outputting a quantitative risk level and disposal suggestion; A linkage triggering module for automatically matching and triggering corresponding plans according to the risk level, and outputting specific linkage control instructions; An execution feedback module for re-evaluation after the disposal is completed by the execution mechanism in response to the control instructions, and real-time monitoring of the results feedback to the system to start a new round of monitoring cycle.

2. The multi-stage security early warning and linkage disposal system of the electric-hydrogen complementary energy station according to claim 1, characterized in that: The key parameter data includes hydrogen concentration, hydrogen storage tank pressure, temperature, electrolytic cell operating status, fuel cell output current and voltage, power load, and environmental humidity.

3. The multi-stage security early warning and linkage disposal system of the electric-hydrogen complementary energy station according to claim 1, characterized in that: The anomaly detection algorithm specifically includes: constructing a hybrid model based on rules and machine learning, creating a configurable rule base for rule-based anomaly detection, and using a rule engine to match abnormal types in real time for the input data stream; For machine learning-based anomaly detection, a historical normal operation data is used to train an isolation forest model, a feature vector is extracted from the real-time data set, and input into the trained model to output an anomaly score. A dynamic threshold is set, and when the anomaly score exceeds the threshold, it is determined that the system has an abnormal type identified by the machine learning model. The dynamic threshold is calculated by continuously maintaining a sliding time window of anomaly scores within the system, and the specified high quantile of the set is used as the current threshold value.

4. The multi-stage security early warning and linkage handling system of the electric-hydrogen complementary energy station according to claim 1, characterized in that: The multi-dimensional analysis method is based on the rule priority principle. The rule-based detection result will immediately generate a preliminary early warning signal. The machine learning model is used for confirmation to check whether the machine learning model also gives an abnormal score in the same or related area. When the rule engine triggers a warning, if the score given by the machine learning model exceeds its dynamic threshold, the warning is confirmed to be valid. When the rule engine triggers a warning, if the score given by the machine learning model does not exceed its dynamic threshold, a final warning will not be issued immediately, and a short observation period will be started. If the rule condition persists during the observation period, even if the machine learning model score does not exceed the dynamic threshold, a warning will be issued. If the rule engine does not trigger a warning, but the score given by the machine learning model exceeds its dynamic threshold, an "unknown mode anomaly" warning will be generated.

5. The multi-stage safety pre-warning and linkage disposal system of the electric-hydrogen complementary energy station according to claim 1, characterized in that: The dynamic risk assessment method is to construct a risk assessment model, predefine a risk base value matrix, and use the warning level and abnormal type as the index. State factors are introduced, including operating mode factor, device redundancy factor, chain effect factor, and environmental and personnel factor; The state coefficient is calculated by weighted summation of multiple state factors; The final risk value is obtained by multiplying the state coefficient with the risk base value and state; According to the calculated risk value, it is mapped to the preset risk level.

6. The multi-stage security early warning and linkage handling system of the electric-hydrogen complementary energy station according to claim 1, characterized in that: The manner of triggering the corresponding preplan is that a preplan library is preset, each preplan containing a trigger condition, a sequence of treatment actions, an execution priority and a mutual exclusion; The risk level and the abnormal type are matched in the preplan library, the matching process supports fuzzy matching and accurate matching, once the optimal preplan is matched, the module generates all the control instructions defined in the preplan in sequence.

7. The multi-stage security early warning and linkage handling system of the electric-hydrogen complementary energy station according to claim 1, characterized in that: The manner of re-evaluating after the treatment is completed is that a short effect monitoring window is preset, key parameters directly related to the treatment action are monitored in real time, the change trend of the parameters is analyzed, and it is judged whether the treatment action produces the expected effect; The execution result state and the effect verification data are taken as inputs, a new risk value is calculated by re-running the risk evaluation model, if the risk value significantly decreases and is lower than a threshold value, it is determined that the treatment is successful, if the risk value is high or even increases, it is determined that the treatment fails or the effect is not good.

8. The multi-stage security early warning and linkage handling system of the electric-hydrogen complementary energy station according to claim 1, characterized in that: The manner of feeding back the real-time monitoring result to the system is that if the treatment is successful, the system records all the data of the event, generates a complete report, and restores the system state to the normal monitoring mode, if the treatment fails or the effect is not good, the module immediately generates a new warning and feeds back to the abnormal detection module, triggering the system to upgrade the treatment.

Images