Intelligent power plant illegal behavior detection method based on deep learning

Abstract

The invention relates to the technical field of power plant violation behavior monitoring, in particular to an intelligent power plant violation behavior detection method based on deep learning. Comprising the steps of setting a plurality of monitoring sub-regions according to power plant equipment parameters, and establishing a behavior analysis model; acquiring a monitoring data packet of each monitoring sub-region according to a preset feedback time node; generating an illegal behavior list of each monitoring sub-region according to all the monitoring data packets and the behavior analysis model; generating a correction parameter of the behavior analysis model according to the preset update time node; according to historical operation parameters, all violation behaviors are screened, characteristic parameters of each violation behavior are analyzed based on a deep learning technology, an analysis sub-model of each violation behavior is constructed, detection of various violation behaviors is realized, monitoring data of a monitoring sub-region is preprocessed through a disturbance sub-model, and a disturbance sub-model is established. The detection precision of the violation behaviors is improved, the early warning efficiency of the violation behaviors in the power plant is improved, and safe operation of the power plant is ensured.

Description

Technical Field

[0001] The present application relates to the technical field of monitoring of power plant violations, and particularly to a method for detecting power plant violations based on deep learning for a smart power plant. Background Art

[0002] Coal is the main raw material of thermal power plants, and the coal cost accounts for more than 70% of the production cost of power generation enterprises. In the whole process management of coal fuel, it is mainly monitored through sensor technology, traditional video monitoring technology, and image recognition technology at present.

[0003] In the process of monitoring and judging violations, due to the differences in environmental parameters in different areas of the power plant, it will interfere with the detection results of violations, resulting in some areas unable to give early warnings of violations in a timely manner, affecting the overall management efficiency of the power plant. Summary of the Invention

[0004] The purpose of the present application is: to solve the above technical problems, the present application provides a method for detecting power plant violations based on deep learning for a smart power plant, aiming to improve the detection accuracy of violations and give early warnings of violations in the power plant in a timely manner.

[0005] In some embodiments of the present application, all violations are screened according to historical operation parameters, and the characteristic parameters of each violation are analyzed based on deep learning technology, and an analysis sub-model for each violation is constructed to realize the detection of various violations and give early warnings of violations in the power plant in a timely manner.

[0006] In some embodiments of the present application, multiple monitoring sub-areas are established according to the equipment parameters in the power plant, and a disturbance sub-model for each monitoring sub-area is generated according to the analysis results of the historical environmental parameters of each monitoring sub-area. The monitoring data of the monitoring sub-area is preprocessed through the disturbance sub-model, so that the characteristic parameters related to violations in the processed monitoring data are more prominent, improving the detection accuracy of violations, improving the early warning efficiency of violations in the power plant, and ensuring the safe operation of the power plant.

[0007] In some embodiments of the present application, a method for detecting power plant violations based on deep learning for a smart power plant is provided, including: Setting multiple monitoring sub-areas according to the power plant equipment parameters and establishing a behavior analysis model; Obtaining the monitoring data packets of each monitoring sub-area according to the preset feedback time node; Generating a list of violations in each monitoring sub-area according to all the monitoring data packets and the behavior analysis model; Generating correction parameters for the behavior analysis model according to the preset update time node; Among them, when setting multiple monitoring sub-regions, it includes: Establish a sequence of monitoring sub-regions A, A = (a 1 , a 2 … a i … a n ), where a i is the i-th monitoring sub-region and n is the number of monitoring sub-regions.

[0008] In some embodiments of the present application, establishing a behavior analysis model includes: Establish a sequence of violation behaviors B according to historical violation parameters, B = (b 1 , b 2 … b i … b m ), where b i is the i-th type of violation behavior and m is the number of violation behavior categories; Set b i as the target violation behavior in sequence according to the sequence of violation behaviors B; Generate a training data packet for the target violation behavior; Generate an analysis sub-model for the target violation behavior according to the training data packet; Generate analysis sub-models for each violation behavior in sequence; Establish a sequence of analysis sub-models P, P = (p 1 , p 2 … p i … p m ), where p i is the i-th analysis sub-model.

[0009] In some embodiments of the present application, establishing a behavior analysis model further includes: Set a i as the target monitoring sub-region in sequence according to the sequence of monitoring sub-regions A; Obtain the historical environmental parameters of the target monitoring sub-region; Generate perturbation parameters for each violation behavior in the target monitoring sub-region according to the historical environmental parameters; Generate a perturbation sub-model for the target monitoring sub-region according to all the perturbation parameters; Generate perturbation sub-models for each monitoring sub-region in sequence; Establish a sequence of perturbation sub-models J, J = (j 1 , j 2 … j i … j n ), where j i is the perturbation sub-model of the i-th monitoring sub-region; Generate a behavior analysis model according to the sequence of perturbation sub-models J and the sequence of analysis sub-models P.

[0010] In some embodiments of the present application, generating a list of violation behaviors for each monitored sub-region includes: Obtaining the monitoring data packet of the target monitored sub-region at the current feedback time node; Setting the perturbation sub-model of the target monitored sub-region as the target perturbation sub-model; Preprocessing the monitoring data packet based on the target perturbation sub-model; Generating a violation data packet of the target monitored sub-region according to the preprocessing result; Generating a sequence C of violation probability values at the current feedback time node according to the violation data packet and the analysis sub-model sequence P; C = (c 1 , c 2 … c i … c n ), where ci is the probability of the target monitored sub-region having the i-th type of violation behavior at the current feedback time node; Generating a list of violation behaviors of the target monitored sub-region according to the sequence C of violation probability values; Generating a violation evaluation value f of the target monitored sub-region; Judging whether to generate a warning instruction according to the violation evaluation value f; Sequentially generating a list of violation behaviors of each monitored sub-region at the current feedback time node.

[0011] In some embodiments of the present application, generating a sequence C of violation probability values at the current feedback time node includes: Sequentially setting b i as the target violation behavior according to the sequence B of violation behaviors; Generating a violation probability value c of the target violation behavior; c = U * (d i - d' i ) 2 ; where U is a conversion coefficient; is the number of characteristic indicators of the target violation behavior; di is the reference value of the i-th characteristic indicator generated based on the violation data packet of the target monitored sub-region; d'i is the standard reference value of the i-th characteristic indicator; Sequentially generating the violation probability values of each violation behavior; Establishing a sequence C of violation probability values according to all the violation probability values.

[0012] In some embodiments of the present application, generating a list of violation behaviors of the target monitored sub-region according to the sequence C of violation probability values includes: Presetting a threshold C1 for the violation probability evaluation value; If c i>C1, there is a type-i violation in the target monitoring sub-region at the current feedback time node, and the number of type-i violations is obtained; If c i <C1, there is no type-i violation in the target monitoring sub-region at the current feedback time node; Generate a list of violations in the target monitoring sub-region according to the judgment result.

[0013] In some embodiments of the present application, generating a violation evaluation value f for the target monitoring sub-region includes: f = e1 * Q1 * (η i * k i) + e2 * Q2 * H; Wherein, e1 is a preset first weight coefficient; e2 is a preset second weight coefficient; Q1 is a preset first fixed coefficient; Q2 is a preset second fixed coefficient; η i is the impact factor of the type-i violation; k i is for the type-i violation generated based on the list of violations in the target monitoring sub-region; H is the risk evaluation value generated based on the list of violations in the target monitoring sub-region.

[0014] In some embodiments of the present application, judging whether to generate a warning instruction according to the violation evaluation value f includes: Preset a first violation evaluation value threshold F1 and a second violation evaluation value threshold F2; If fi < F1, no warning instruction is generated at the target monitoring point at the current feedback time node; If F1 < f < F2, a first-level warning instruction is generated at the target monitoring point at the current feedback time node; If f > F2, a second-level warning instruction is generated at the target monitoring point at the current feedback time node.

[0015] In some embodiments of the present application, generating correction parameters for the behavior analysis model according to a preset update time node includes: Set the i-th perturbation sub-model as the perturbation sub-model to be corrected in turn according to the perturbation sub-model sequence J; Obtain the historical monitoring data of the monitoring sub-region corresponding to the perturbation sub-model to be corrected between the current update time node and the previous update time node; Generate an iterative data packet according to the historical monitoring data; Set the optimization strategy of the perturbation sub-model to be corrected according to the iterative data packet; Generate the optimization strategies of each perturbation sub-model in turn; Generate correction parameters for the behavior analysis model according to all the optimization strategies.

[0016] Compared with the prior art, the beneficial effects of the method for detecting illegal behaviors in a smart power plant based on deep learning according to an embodiment of the present application are as follows: All illegal behaviors are screened according to historical operation parameters, and the characteristic parameters of each illegal behavior are analyzed based on deep learning technology to construct an analysis sub-model for each illegal behavior, so as to detect various types of illegal behaviors and timely warn of the illegal behaviors in the power plant.

[0017] Multiple monitoring sub-regions are established according to the equipment parameters in the power plant, and disturbance sub-models for each monitoring sub-region are generated according to the analysis results of the historical environmental parameters of each monitoring sub-region. The monitoring data of the monitoring sub-region is preprocessed through the disturbance sub-model, so that the characteristic parameters related to illegal behaviors in the processed monitoring data are more prominent, improving the detection accuracy of illegal behaviors, enhancing the early warning efficiency of illegal behaviors in the power plant, and ensuring the safe operation of the power plant. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a schematic flow chart of a method for detecting illegal behaviors in a smart power plant based on deep learning in a preferred embodiment of an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] The following further describes in detail the specific embodiments of the present application in conjunction with the drawings and embodiments. The following embodiments are used to illustrate the present application but are not intended to limit the scope of the present application.

[0020] In the description of the present application, it should be understood that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as limiting the present application.

[0021] The terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present application, unless otherwise stated, the meaning of "a plurality" is two or more.

[0022] In the description of this application, it should be noted that unless otherwise clearly specified and defined, the terms "installation", "connection", and "linkage" shall be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, and it can be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific situations.

[0023] As Figure 1 shown, a method for detecting illegal behaviors in a smart power plant based on deep learning according to a preferred embodiment of an embodiment of this application is characterized by including: S101: Set multiple monitoring sub-regions according to the power plant equipment parameters and establish a behavior analysis model; S102: Obtain the monitoring data packets of each monitoring sub-region according to the preset feedback time node; S103: Generate a list of illegal behaviors for each monitoring sub-region according to all the monitoring data packets and the behavior analysis model; S104: Generate correction parameters for the behavior analysis model according to the preset update time node; Among them, when setting multiple monitoring sub-regions, it includes: Establish a sequence of monitoring sub-regions A, A=(a 1 , a 2 …a i …a n ), where a i is the i-th monitoring sub-region and n is the number of monitoring sub-regions.

[0024] Specifically, establish multiple monitoring sub-regions according to the equipment parameters and work processes in the areas involved in the whole-process management of fuel in the power plant, and generate a sequence of monitoring sub-regions.

[0025] Specifically, set corresponding monitoring devices in each monitoring sub-region to collect monitoring data in real time, and the monitoring devices include but are not limited to image acquisition devices and various sensors.

[0026] Specifically, establishing a behavior analysis model includes: Establish a sequence of illegal behaviors B according to historical illegal parameters, B=(b 1 , b 2 …b i …b m ), where b i is the i-th type of illegal behavior and m is the number of illegal behavior categories; Set b i as the target illegal behavior in sequence according to the sequence of illegal behaviors B; Generate a training data packet for the target violation behavior; Generate an analysis sub-model for the target violation behavior according to the training data packet; Generate analysis sub-models for each violation behavior in sequence; Establish a sequence of analysis sub-models P, P = (p 1 , p 2 … p i … p m ), where p i is the i-th analysis sub-model.

[0027] Specifically, screen various violation behaviors according to the historical operation data in the power plant and establish a sequence of violation behaviors. Analyze a single violation behavior to generate the characteristic index parameters of this violation behavior, so as to establish the corresponding behavior sub-model.

[0028] Specifically, the violation behaviors include, but are not limited to, personnel violation behaviors, equipment illegal behaviors, etc. Other personnel illegal behaviors include, but are not limited to, improper operations, operations not in accordance with procedures, illegal entry of staff, etc. Equipment illegal behaviors include, but are not limited to, environmental parameter settings not meeting the regulations, equipment not placed in the specified position, etc.

[0029] Specifically, analyze a single violation behavior based on deep learning technology to generate the characteristic index parameters of this violation behavior, so as to establish the corresponding behavior sub-model.

[0030] Specifically, establishing a behavior analysis model also includes: Set a i as the target monitoring sub-region in sequence according to the sequence of monitoring sub-regions A; Obtain the historical environmental parameters of the target monitoring sub-region; Generate the perturbation parameters of each violation behavior in the target monitoring sub-region according to the historical environmental parameters; Generate a perturbation sub-model for the target monitoring sub-region according to all the perturbation parameters; Generate perturbation sub-models for each monitoring sub-region in sequence; Establish a sequence of perturbation sub-models J, J = (j 1 , j 2 … j i … j n ), where j i is the perturbation sub-model of the i-th monitoring sub-region; Generate a behavior analysis model according to the sequence of perturbation sub-models J and the sequence of analysis sub-models P.

[0031] Specifically, analyze the historical environmental parameters in the target monitoring sub-region to generate the perturbation parameters of the target monitoring sub-region and establish the corresponding perturbation sub-model.

[0032] Specifically, the perturbation parameter means that the significance of the characteristic indicators of each violation behavior in the environmental parameters within the target monitoring sub-region will decrease, resulting in the inability to detect violation behaviors in a timely manner. For example, in an area with a lot of dust, it is more difficult to capture the behavior characteristics of personnel.

[0033] It can be understood that in the above embodiments, all violation behaviors are screened according to historical operation parameters, and the violation data packets of each violation behavior are analyzed based on deep learning technology to construct an analysis sub-model for each violation behavior. Through the cross-combination of multiple perturbation sub-models and multiple analysis sub-models, accurate detection of different violation behaviors in different regions is achieved, and early warnings are given to the violation behaviors in the power plant in a timely manner.

[0034] In the preferred embodiment of the present application, a list of violation behaviors for each monitoring sub-region is generated, including: Obtain the monitoring data packet of the target monitoring sub-region at the current feedback time node; Set the perturbation sub-model of the target monitoring sub-region as the target perturbation sub-model; Preprocess the monitoring data packet based on the target perturbation sub-model; Generate the violation data packet of the target monitoring sub-region according to the preprocessing result; Generate a sequence C of violation probability values at the current feedback time node according to the violation data packet and the sequence P of analysis sub-models; C=(c 1 ,c 2 …c i …c n ), where ci is the probability of the i-th type of violation behavior existing in the target monitoring sub-region at the current feedback time node; Generate a list of violation behaviors for the target monitoring sub-region according to the sequence C of violation probability values; Generate a violation evaluation value f for the target monitoring sub-region; Judge whether to generate a warning instruction according to the violation evaluation value f; Generate a list of violation behaviors for each monitoring sub-region at the current feedback time node in sequence.

[0035] Specifically, the violation data packet includes the position points where violation behaviors may exist in the current monitoring sub-region and the parameter values of the characteristic indicators extracted from each position point.

[0036] Specifically, the characteristic indicator parameters of each position point are analyzed and compared through all analysis sub-models, so as to generate the types of violation behaviors of each position point in the target monitoring sub-region.

[0037] Specifically, the list of violations includes the number of violations, the location points where violations occur, and the categories of violations at those location points.

[0038] Specifically, the larger the violation probability value ci, the greater the likelihood that the ith type of violation exists in the target monitoring sub-region.

[0039] Specifically, generate the sequence of violation probability values C for the current feedback time node, including: Set b in sequence according to the sequence of violation behaviors B i as the target violation behavior; Generate the violation probability value c for the target violation behavior; c = U * (d i - d' i ) 2 ; where U is the conversion coefficient; is the number of characteristic indicators of the target violation behavior; di is the reference value of the ith characteristic indicator generated based on the violation data packets in the target monitoring sub-region; d'i is the standard reference value of the ith characteristic indicator; Generate the violation probability values of each violation behavior in sequence; Establish the sequence of violation probability values C based on all the violation probability values.

[0040] Specifically, set the corresponding characteristic indicators according to the category of the target violation behavior, and generate the value range of each characteristic indicator when the current violation behavior occurs, so as to generate the standard reference value.

[0041] Specifically, generate the value rules of each characteristic indicator according to historical parameters, and generate the real-time values of each characteristic indicator by analyzing the violation data packets.

[0042] Specifically, the characteristic indicators of each violation behavior are different.

[0043] Specifically, generate the list of violation behaviors in the target monitoring sub-region according to the sequence of violation probability values C, including: The preset threshold C1 for the evaluation value of the violation probability; If c i > C1, there is the ith type of violation behavior in the target monitoring sub-region at the current feedback time node, and obtain the number of the ith type of violation behavior; If c i < C1, there is no ith type of violation behavior in the target monitoring sub-region at the current feedback time node; Generate the list of violation behaviors in the target monitoring sub-region according to the judgment result.

[0044] Specifically, a threshold value of the violation probability evaluation value is set according to historical parameters. When the real-time violation evaluation value c i is greater than the threshold value of the violation evaluation value, it indicates that there is a type-i violation behavior in the target monitoring sub-region.

[0045] Specifically, the list of violation behaviors includes the number of violation behaviors, the position points where the violation behaviors occur, and the types of violation behaviors at these position points.

[0046] It can be understood that in the above embodiments, the monitoring data of the monitoring sub-region is preprocessed by the perturbation sub-model, so that the characteristic parameters related to the violation behavior in the processed monitoring data are more prominent, the detection accuracy of the violation behavior is improved, the early warning efficiency of the violation behavior in the power plant is improved, and the safe operation of the power plant is ensured.

[0047] In the preferred embodiment of the present application, generating a violation evaluation value f of the target monitoring sub-region includes: f = e1 * Q1 * (η i * k i) + e2 * Q2 * H; Wherein, e1 is a preset first weight coefficient; e2 is a preset second weight coefficient; Q1 is a preset first fixed coefficient; Q2 is a preset second fixed coefficient; η i is the influence factor of the type-i violation behavior; k i is the type-i violation behavior generated based on the list of violation behaviors in the target monitoring sub-region; H is the risk evaluation value generated based on the list of violation behaviors in the target monitoring sub-region.

[0048] Specifically, all parameters in the model are normalized by presetting the first fixed coefficient and the second fixed coefficient, so that each parameter is within the same value range.

[0049] Specifically, judging whether to generate a warning instruction according to the violation evaluation value f includes: Presetting a first violation evaluation value threshold F1 and a second violation evaluation value threshold F2; If fi < F1, no warning instruction is generated at the current feedback time node for the target monitoring point; If F1 < f < F2, a first-level warning instruction is generated at the current feedback time node for the target monitoring point; If f > F2, a second-level warning instruction is generated at the current feedback time node for the target monitoring point.

[0050] Specifically, a first-level warning instruction means that there are violations within the target monitoring sub-region, but it will not have a significant impact on the operation of the power plant. Reminder information needs to be sent in a timely manner to correct the corresponding violations. A second-level warning instruction means that there are serious violations within the target monitoring sub-region, and immediate rectification and prevention are required.

[0051] In the preferred embodiment of the present application, the correction parameters of the behavior analysis model are generated according to the preset update time node, including: Set the i-th disturbance sub-model as the disturbance sub-model to be corrected according to the disturbance sub-model sequence J in turn; Obtain the historical monitoring data of the monitoring sub-region corresponding to the disturbance sub-model to be corrected between the current update time node and the previous update time node; Generate an iterative data packet according to the historical monitoring data; Set the optimization strategy of the disturbance sub-model to be corrected according to the iterative data packet; Generate the optimization strategies of each disturbance sub-model in turn; Generate the correction parameters of the behavior analysis model according to all the optimization strategies.

[0052] Specifically, by periodically processing the historical monitoring data of each monitoring sub-region, iterative data packets of each monitoring sub-region are generated, and each disturbance sub-model is iteratively optimized periodically, so as to improve the processing efficiency of each disturbance sub-model for monitoring data and improve the detection and warning ability for violations.

[0053] According to the first concept of the present application, all violations are screened according to historical operation parameters, and the characteristic parameters of each violation are analyzed based on deep learning technology to construct an analysis sub-model for each violation, so as to realize the detection of various violations and timely warn of the violations in the power plant.

[0054] According to the second concept of the present application, multiple monitoring sub-regions are established according to the equipment parameters in the power plant, and the disturbance sub-models of each monitoring sub-region are generated according to the analysis results of the historical environmental parameters of each monitoring sub-region. The monitoring data of the monitoring sub-region is pre-processed through the disturbance sub-model, so that the characteristic parameters related to violations in the processed monitoring data are more prominent, improving the detection accuracy of violations and the warning efficiency of violations in the power plant, and ensuring the safe operation of the power plant.

[0055] The above are only the preferred embodiments of the present application. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present application, several improvements and replacements can be made, and these improvements and replacements should also be regarded as the protection scope of the present application.

Claims

1. A method for detecting illegal behaviors in smart power plants based on deep learning, characterized in that: Including: Set multiple monitoring sub - regions according to the power plant equipment parameters and establish a behavior analysis model; Obtain the monitoring data packets of each monitoring sub - region according to the preset feedback time node; Generate a list of violation behaviors for each monitoring sub - region according to all the monitoring data packets and the behavior analysis model; Generate correction parameters for the behavior analysis model according to the preset update time node; Among them, when setting multiple monitoring sub - regions, it includes: Establish a monitoring sub-area sequence A, A=(a1, a2…a i …a n ), where a i is the ith monitoring sub-area, and n is the number of monitoring sub-areas.

2. The method for detecting violations in a smart power plant based on deep learning according to claim 1, characterized in that: Establishing a behavior analysis model includes: According to the historical violation parameters, a violation behavior sequence B is established, B=(b1,b2…b i …b m ), where b i is the i-th type of violation, m is the number of violation categories; According to the violation sequence B, set b i for targeted violations; Generate training data packets for target violation behaviors; Generate an analysis sub - model for target violation behaviors according to the training data packets; Generate analysis sub - models for each violation behavior in turn; Establish the analysis sub-model series P, P=(p1, p2…p i …p m ), where p i is the i-th analysis sub-model.

3. The method for detecting violations in a smart power plant based on deep learning according to claim 2, characterized in that: Establishing a behavior analysis model also includes: According to the monitoring sub-area sequence A, set a i Monitor sub-areas for the target; Obtain the historical environmental parameters of the target monitoring sub - region; Generate perturbation parameters for each violation behavior in the target monitoring sub - region according to the historical environmental parameters; Generate a perturbation sub - model for the target monitoring sub - region according to all the perturbation parameters; Generate perturbation sub - models for each monitoring sub - region in turn; Establish the perturbation submodel series J, J=(j1, j2…j i …j n ), where j i is the disturbance sub-model of the i-th monitoring sub-area; Generate a behavior analysis model according to the perturbation sub - model sequence J and the analysis sub - model sequence P.

4. The method for detecting violations in a smart power plant based on deep learning according to claim 3 is characterized in that: Generating a list of violation behaviors for each monitoring sub - region includes: Obtain the monitoring data packet of the target monitoring sub - region at the current feedback time node; Set the perturbation sub - model of the target monitoring sub - region as the target perturbation sub - model; Pre - process the monitoring data packet based on the target perturbation sub - model; Generate a violation data packet for the target monitoring sub - region according to the pre - processing result; Generate a sequence C of violation probability values at the current feedback time node according to the violation data packet and the analysis sub - model sequence P; C=(c1,c2…c i …c n ), where ci is the probability that the target monitoring sub-area has the i-th type of violation at the current feedback time node; Generate a list of violation behaviors for the target monitoring sub - region according to the sequence C of violation probability values; Generate a violation evaluation value f for the target monitoring sub - region; Judge whether to generate a warning instruction according to the violation evaluation value f; Generate a list of violation behaviors for each monitoring sub - region at the current feedback time node in turn.

5. The method for detecting violations in a smart power plant based on deep learning according to claim 4, characterized in that: Generating the sequence C of violation probability values at the current feedback time node includes: According to the violation sequence B, set b i for targeted violations; Generate a violation probability value c for the target violation behavior; c=U*[ (d i -d' i ) 2 ]; Wherein, U is the conversion coefficient; is the number of characteristic indicators of the target violation behavior; di is the reference value of the i-th characteristic indicator generated based on the violation data packet of the target monitoring sub-area; d'i is the standard reference value of the i-th characteristic indicator; Generate violation probability values for each violation behavior in turn; Establish a sequence C of violation probability values according to all the violation probability values.

6. The method for detecting violations in a smart power plant based on deep learning according to claim 5, characterized in that: Generating a list of violation behaviors for the target monitoring sub - region according to the sequence C of violation probability values includes: Preset a threshold C1 for the violation probability evaluation value; If c i >C1, there is a violation of type i in the target monitoring sub-area at the current feedback time node, and the number of violations of type i is obtained; If c i <C1, there is no type-i violation in the target monitoring sub-region at the current feedback time node; Generate a list of violation behaviors for the target monitoring sub - region according to the judgment result.

7. The method for detecting violations in a smart power plant based on deep learning according to claim 6, characterized in that: Generating a violation evaluation value f for the target monitoring sub - region includes: f=e1*Q1* (η i *k i) ]+e2*Q2*H; Wherein, e1 is the preset first weight coefficient; e2 is the preset second weight coefficient; Q1 is the preset first fixed coefficient; Q2 is the preset second fixed coefficient; η i is the impact factor of the i-th type of violation; k i is the i-th type of violation generated based on the violation list of the target monitoring sub-area; H is the risk assessment value generated based on the violation list of the target monitoring sub-area.

8. The method for detecting violations in a smart power plant based on deep learning according to claim 7, characterized in that: Judging whether to generate a warning instruction according to the violation evaluation value f includes: Preset a first threshold F1 and a second threshold F2 for the violation evaluation value; If fi < F1, no warning instruction is generated for the target monitoring point at the current feedback time node; If F1 < f < F2, a first - level warning instruction is generated for the target monitoring point at the current feedback time node; If f > F2, a second - level warning instruction is generated for the target monitoring point at the current feedback time node.

9. The method for detecting violations in a smart power plant based on deep learning according to claim 5, characterized in that: Generating correction parameters for the behavior analysis model according to the preset update time node includes: Set the i - th perturbation sub - model as the perturbation sub - model to be corrected in turn according to the perturbation sub - model sequence J; Obtain the historical monitoring data of the monitoring sub - region corresponding to the perturbation sub - model to be corrected between the current update time node and the previous update time node; Generate iterative data packets according to the historical monitoring data; An optimization strategy for the disturbance sub-model to be corrected is set according to an iterative data packet; Generate optimization strategies for each perturbation sub-model in turn; Generate correction parameters of the behavior analysis model based on all optimization strategies.

Images