Self-maintenance system of pollution source monitoring equipment

Abstract

The invention belongs to the technical field of environmental monitoring, and discloses a pollution source monitoring equipment self-maintenance system, which comprises a data acquisition and preprocessing module, an equipment value and redundancy evaluation module, a maintenance strategy simulation module, a multi-objective optimization decision module and an execution and verification feedback module, from the perspective of monitoring network global data inference, objective and quantitative evaluation of contribution importance and repeatability of equipment data is realized, and the technical problems of isolated maintenance decision and unscientific resource allocation caused by lack of a network global perspective in a traditional maintenance mode are fundamentally solved. A core basis is provided for accurate division of maintenance priorities, a candidate maintenance strategy is simulated to repair specific equipment data in a spatio-temporal data matrix, whole-network data is reconstructed by using a matrix completion algorithm, and then reconstruction errors are calculated to quantify pre-evaluation strategy efficiency, so that the system can be used before actual investment of maintenance resources.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of environmental monitoring, in particular to a self-maintenance system of a pollution source monitoring device. BACKGROUND

[0002] The pollution source online monitoring network is the core infrastructure of environmental supervision, and its long-term, stable and high-quality operation is the basis for accurately assessing pollution conditions and implementing supervision responsibilities. With the expansion of the monitoring network and the growth of the service time of the equipment, the traditional operation and maintenance mode which relies on fixed period inspection and post-fault maintenance is increasingly highlighting its limitations. This mode is essentially a passive response, and its decision-making basis is often isolated equipment alarms or simple time threshold, lacking the robustness of the overall data quality of the monitoring network, and the foresight evaluation and global optimization. Therefore, an intelligent self-maintenance system is needed, which can start from the global perspective of the network, actively evaluate the importance of the equipment, predict the maintenance effect and dynamically optimize the decision.

[0003] However, the existing related technologies have the following problems, which cannot meet the above needs: (1) Existing researches mostly focus on using data-driven methods to conduct state early warning or fault diagnosis on a single device, and the optimization target is the reliability of a single device. Such methods fail to evaluate the global impact of the performance decline or data loss of a device on the data integrity and inference ability of the entire monitoring network from the information contribution level, resulting in a lack of scientific basis for network-level allocation of maintenance resources, and possible resource mismatch problems such as "insufficient maintenance of key devices" and "excessive maintenance of secondary devices".

[0004] (2) Although some researches apply information theory (such as value information and mutual information) and matrix completion algorithms to the optimization layout and performance evaluation of sensor networks, their application scenarios are limited to the spatial point design at the initial stage of network construction or the offline evaluation of the static performance of the already deployed network. These methods do not apply the dynamic evaluation of information value and data redundancy, and the pre-evaluation idea of network performance simulation, to the time dimension problem of continuous and dynamic optimization of network operation period maintenance decision-making. Therefore, the current maintenance strategy formulation still heavily relies on historical experience, and cannot quantitatively simulate and compare the network-level effects of different strategies before implementation, lacking the key technical support for upgrading the decision-making from experience-driven to model and simulation-driven. SUMMARY

[0005] In view of the above problems in the prior art, the present application provides a pollution source monitoring device self-maintenance system, which aims to solve the problems of isolated decision-making, unclear effect and lack of predictability in the existing maintenance method.

[0006] To achieve the above purpose, the technical solution adopted by the present application is as follows: The pollution source monitoring equipment self-maintenance system provided by the application comprises a data acquisition and preprocessing module, an equipment value and redundancy evaluation module, a maintenance strategy simulation module, a multi-objective optimization decision module and an execution and verification feedback module.

[0007] Further, the data acquisition and preprocessing module is used for connecting the pollution source monitoring equipment in the monitoring network, collecting and preprocessing the operation data and environmental monitoring data of the monitoring equipment, constructing an original space-time data matrix and providing the original space-time data matrix to the equipment value and redundancy evaluation module. In the data acquisition and preprocessing module, the preprocessing comprises data standardization and coarse noise filtering. The data standardization is to map the operation data and environmental monitoring data of the monitoring equipment to

[0008] Further, the equipment value and redundancy evaluation module is connected to the data acquisition and preprocessing module, calculates the data value index for each monitoring equipment and the data redundancy for each pair of monitoring equipment based on an information theory method, and provides the data value index and the data redundancy to the maintenance strategy simulation module and the multi-objective optimization decision module at the same time.

[0009] Further, the maintenance strategy simulation module receives the candidate maintenance strategies, each of which specifies a set of monitoring equipment to be maintained and a corresponding future maintenance time period, maps each of the candidate maintenance strategies to the original space-time data matrix, the matrix elements of which represent the standardized data of the corresponding monitoring equipment at the corresponding time, marks the data points corresponding to the monitoring equipment to be maintained in the candidate maintenance strategies as high-quality known data points in the future maintenance time period, constructs a known data region, and uses the high-quality known data points in the known data region as the basis to complete and reconstruct the original space-time data matrix by using a non-negative matrix factorization algorithm.

[0010] Further, the multi-objective optimization decision module connects the maintenance strategy simulation module and the equipment value and redundancy evaluation module, establishes and solves a multi-objective optimization model based on the data value index, the data redundancy between equipment, and the pre-evaluation result of efficiency, sets a binary decision variable to represent whether a monitoring equipment is included in the maintenance plan, constructs a three-objective function of maximizing the sum of the data value indexes of the monitoring equipment in maintenance, minimizing the sum of the data redundancy between the monitoring equipment in maintenance, and minimizing the simulation prediction error, sets a constraint condition that the total maintenance cost does not exceed a preset budget, solves the model by using a multi-objective evolutionary algorithm to obtain a Pareto optimal solution set, selects an optimal maintenance plan from the solution set according to the current preset budget, and provides the plan to the execution and verification feedback module.

[0011] Further, the execution and verification feedback module connects the multi-objective optimization decision module and the data collection and preprocessing module, receives and issues the optimal maintenance plan, collects actual data of the monitoring network after maintenance as verification data, takes the verification data of the maintained monitoring equipment as known points, infers the data of the un-maintained monitoring equipment by using the matrix completion algorithm again to obtain an inferred data matrix, calculates the root mean square error between the actual verification data of the un-maintained monitoring equipment and the inferred data matrix as an actual effect score after maintenance, compares the actual effect score with the prediction effect score in the pre-evaluation to obtain deviation information, and if the deviation continuously exceeds a preset threshold, feeds the deviation information back to the equipment value and redundancy evaluation module and the multi-objective optimization decision module; the equipment value and redundancy evaluation module adjusts a smoothing parameter of the probability distribution estimation algorithm when calculating the data value index and the redundancy according to the deviation information, and the multi-objective optimization decision module adjusts the weight coefficient of the objective function in the optimization model according to the deviation information, so that the system realizes closed-loop adaptive learning.

[0012] The application further provides a pollution source monitoring equipment self-maintenance method based on information value and matrix completion, which is applied to the system and includes the following steps: Step S1: collection and construction of monitoring network space-time data; Step S2: equipment data value and redundancy evaluation; Step S3: maintenance strategy simulation and effect pre-evaluation; Step S4: multi-objective optimization maintenance decision generation; Step S5: maintenance execution and closed-loop verification.

[0013] The application has the following beneficial effects by using the above scheme: (1) By calculating the data value index (VOI) of each monitoring device and the data redundancy between devices (TE), from the perspective of global data deducibility of the monitoring network, the objective and quantitative evaluation of the importance of device data contribution and its redundancy is realized, which fundamentally solves the technical problems of isolated maintenance decision and unscientific resource allocation in the traditional maintenance mode due to the lack of network global perspective, and provides a core basis for accurate division of maintenance priority.

[0014] (2) By simulating the candidate maintenance strategy as repairing specific device data in the space-time data matrix, and using the matrix completion algorithm to reconstruct the network data, and then calculating the reconstruction error to quantify the pre-evaluation strategy effectiveness, the system can virtually simulate and compare the network-level repair effect of different schemes before actually investing maintenance resources, which solves the technical problem of long-term reliance on experience and difficulty in predicting the effect of maintenance decision, and realizes the paradigm shift from experience-driven to simulation-optimized.

[0015] (3) By constructing a complete closed-loop architecture of "evaluation, simulation, decision, execution, and verification", and introducing a parameter self-adaptive adjustment mechanism based on effect deviation feedback, using the difference between the actual effect data after maintenance and the pre-evaluation result to dynamically correct the internal parameters of the information theory evaluation model and the multi-objective optimization decision model, the system is given continuous self-learning and self-adaptive ability, which solves the technical problem that static models cannot cope with dynamic factors such as device aging and environmental changes, leading to the gradual invalidation of decision strategies, and ensures the decision accuracy and robustness of the system in long-term operation. BRIEF DESCRIPTION OF DRAWINGS

[0016] The accompanying drawings are used to provide a further understanding of the present scheme, and constitute a part of the specification, together with the embodiments of the present application, to explain the present application, and do not constitute a limitation of the present application.

[0017] Fig. 1 A whole module diagram of a pollution source monitoring device self-maintenance system according to the present application is provided; Fig. 2 A step flow chart of a pollution source monitoring device self-maintenance method based on information value and matrix completion according to the present application is provided; Fig. 3 A detailed flow chart of maintenance strategy simulation and effect pre-evaluation according to the present application is provided. DETAILED DESCRIPTION

[0018] Embodiment one, refer to Figs. 1-3 The pollution source monitoring device self-maintenance system provided by the present application comprises a data acquisition and preprocessing module, a device value and redundancy evaluation module, a maintenance strategy simulation module, a multi-objective optimization decision module, and an execution and verification feedback module. The data acquisition and preprocessing module connects to all pollution source monitoring devices in the monitoring network and is used to collect device operation data and environmental monitoring data to construct the original spatiotemporal data matrix. The equipment value and redundancy assessment module is connected to the data acquisition and preprocessing module. It is used to calculate the data value index of the monitoring equipment and the data redundancy between equipment based on information theory methods, and to generate multiple candidate maintenance strategies based on the data value index and data redundancy. The maintenance strategy simulation module is connected to the equipment value and redundancy assessment module, receives at least one candidate maintenance strategy, and uses a matrix completion algorithm to simulate the received candidate maintenance strategy and pre-evaluate its effect. The multi-objective optimization decision module connects the maintenance strategy simulation module and the equipment value and redundancy assessment module, and is used to generate the optimal maintenance plan through multi-objective optimization. The execution and verification feedback module connects the multi-objective optimization decision-making module and the data acquisition and preprocessing module, and is used to execute the maintenance plan and verify the effect, realizing closed-loop feedback learning.

[0019] Example 2, based on the above examples, involves the data acquisition and preprocessing module acquiring historical and real-time operational data and environmental monitoring data from all devices in the pollution source monitoring network. Preprocessing and feature extraction are performed to construct an original spatiotemporal data matrix arranged by time and device nodes. The device operational data includes calibration history deviation, sensor signal drift rate, and fault alarm frequency. The environmental monitoring data includes pollutant concentration, flow rate, temperature, and humidity. The preprocessing includes data standardization and coarse noise filtering. Data standardization uses minimum-maximum standardization to map all features to... Interval, eliminating dimensional differences, and coarse noise filtering are employed. The criteria remove obviously abnormal data.

[0020] Example 3, based on the above examples, describes an equipment value and redundancy assessment module that uses information theory to calculate the data value index of each monitoring device in the monitoring network. Data redundancy between each pair of monitoring devices Specifically, it includes: Equipment Data Value Index Calculation: Based on the value information method, for any first monitoring device in the monitoring network. Calculate the first monitoring device With each of the second monitoring devices in the monitoring network, except for the first monitoring device. The amount of shared information between data states determines the data redundancy between devices. Finally, the contribution values ​​of all other devices in the network are integrated to obtain the first monitoring device. Global data value index Calculate the first monitoring device For each second monitoring device The formula for calculating value contribution is: ; in, For information collection, For device data status set, Indicates a specific state, Indicates action, For cost matrix elements, For prior probability, For posterior probability, the device Global data value index: ;in, It represents the set of all devices in the network; Data redundancy between devices Calculation: Based on the mutual information method, calculate the first monitoring device. Except for the first monitoring equipment Each of the second monitoring devices outside The amount of shared information between data states is calculated using the following formula: ; in, For computing devices In state And other equipment In state The joint probability, and This represents the marginal probability.

[0021] Example 4, based on the above examples, involves the maintenance strategy simulation module receiving at least one candidate maintenance strategy and using a matrix completion algorithm to quantify and pre-evaluate the effectiveness of each strategy. Specifically, this includes: Candidate maintenance strategy definition: Map each candidate maintenance strategy to the original spatiotemporal data matrix. In the original spatiotemporal data matrix, the number of time points is used as the rows and the number of monitoring device nodes is used as the columns. The matrix elements represent the standardized data of the corresponding monitoring device at the corresponding time. The data points corresponding to the monitoring devices to be maintained specified in the candidate maintenance strategy during the future maintenance period are marked as high-quality known data points. ,in For the number of time points, Number of device nodes Matrix elements represent devices At any moment Standardized data, for a candidate maintenance strategy The set of equipment it plans to maintain is denoted as Construct the known data region corresponding to the candidate maintenance strategy and denote it as... ; Matrix completion simulation reconstruction: using known data regions Based on high-quality known data points, a nonnegative matrix factorization algorithm is used to process the original spatiotemporal data matrix. The process of completing and reconstructing the system is transformed into solving the following optimization problem: ;in, and It is a non-negative decomposition matrix. Given the preset matrix rank, the objective function is... Defined as: ;in, The regularization coefficient is . The Frobenius norm of the matrix is ​​used to calculate the loss only for unknown locations. Solving this optimization problem yields the complete data matrix after completion. Nonnegative matrix factorization is achieved by solving an optimization problem that minimizes an objective function, which is only applicable to the known data region. The reconstruction error is calculated based on the location of unknown data other than the decomposition matrix, and includes a regularization term for the decomposition matrix. Quantitative Pre-evaluation of Strategy Effectiveness: By Calculating the Reconstruction Matrix With the original spacetime matrix Evaluating candidate maintenance strategies based on errors at unknown data locations. The effect is assessed using root mean square error as a pre-evaluation metric: ;in, Represents a set of unknown locations. The number of elements in the set of unknown locations. The smaller the value, the stronger the candidate maintenance strategy. The better the prediction of the overall network data quality, the better.

[0022] Example 5, based on the above examples, describes a multi-objective optimization decision module that establishes and solves a multi-objective optimization model to generate an optimal maintenance plan, specifically including: A multi-objective optimization model is constructed with the goal of minimizing maintenance costs and maximizing the overall data quality of the network. Let binary decision variables be used. This indicates whether the corresponding monitoring equipment has been included in the maintenance plan. =1 indicates the device It has been included in the maintenance plan, and the objective function is as follows: First objective function: Maximize the total value of the data. ; Second objective function: Minimize the total redundancy within the maintenance set. ; Third objective function: Minimize simulation prediction error. ; in, This indicates that the above binary decision variables The constructed decision vector Pre-assessment of the corresponding maintenance strategy The value, constrained by the total maintenance cost. ,in To maintain equipment The cost, For the total budget; Pareto optimal solution set: The non-dominated sorting genetic algorithm with elitist strategy is used to solve the optimization problems of the first objective function, the second objective function and the third objective function to obtain the Pareto optimal solution set; Optimal decision generation: based on the current total budget The most suitable solution is selected from the Pareto optimal solution set as the optimal maintenance plan.

[0023] Example 6, based on the above examples, describes an execution and verification feedback module that executes the maintenance plan and verifies its effectiveness to achieve closed-loop feedback. Specifically, it includes: Maintenance instruction execution: Issue and execute the optimal maintenance plan; Post-maintenance effect verification: After maintenance, actual data was collected to obtain a verification data matrix. The verification data matrix Using the data of the maintained equipment as known points, the matrix completion algorithm is used again to infer the data of the unmaintained equipment, resulting in the inference matrix. Calculate the actual effect score:

[0024] in, This is the set of data locations corresponding to unmaintained equipment. Feedback and Adaptive Learning: Scoring Actual Results Compared with the pre-assessed value (The root mean square error of the preliminary evaluation of the candidate maintenance strategy in Example 4) By comparing the two, we can obtain the deviation information between them; if the deviation... Continuously exceeding the threshold The deviation information will then be fed back to the equipment value and redundancy assessment module and the multi-objective optimization decision-making module. The equipment value and redundancy assessment module adjusts the smoothing parameters of the probability distribution estimation algorithm used when calculating the data value index (VOI) and data redundancy (TE) based on the deviation information. The multi-objective optimization decision module adjusts the objective function in its multi-objective optimization model based on the deviation information. (i.e., the weighting coefficients that minimize the prediction error term); Through the above process, the system achieves closed-loop adaptive learning.

[0025] Example 7 describes a self-maintenance method for pollution source monitoring equipment based on information value and matrix completion, using the system described above. This method is applied to the system and includes the following steps: Step S1: Monitoring network spatiotemporal data acquisition and construction; Step S2: Evaluation of equipment data value and redundancy; Step S3: Maintenance strategy simulation and effect pre-evaluation; Step S4: Generation of multi-objective optimization maintenance decisions; Step S5: Maintenance execution and closed-loop verification.

[0026] Through the above embodiments, the present invention has achieved a fundamental transformation in the maintenance of pollution source monitoring equipment, from passive response to proactive prediction, from local optimization to global balancing, and from open-loop execution to closed-loop learning, significantly improving the overall data reliability and operation and maintenance management efficiency of the monitoring network.

[0027] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A self-maintenance system for pollution source monitoring equipment, characterized in that: It includes a data acquisition and preprocessing module, an equipment value and redundancy assessment module, a maintenance strategy simulation module, a multi-objective optimization decision-making module, and an execution and verification feedback module; The data acquisition and preprocessing module acquires operational data from the monitoring equipment and environmental monitoring data, performs preprocessing, and constructs an original spatiotemporal data matrix. The equipment value and redundancy assessment module, based on information theory, uses the original spatiotemporal data matrix to calculate the data value index and data redundancy between monitoring devices for each monitoring device, and generates multiple candidate maintenance strategies. The method for calculating the data value index for each monitoring device includes: based on the value information method, for any first monitoring device in the monitoring network, calculating the value contribution of the data of the first monitoring device to assess the data status of each second monitoring device in the monitoring network other than the first monitoring device; then summing the value contribution values ​​of the first monitoring device to all second monitoring devices to obtain the data value index of the first monitoring device; The method for calculating the data redundancy between monitoring devices is as follows: based on the mutual information method, for any first monitoring device in the monitoring network, the amount of shared information between the data status of the first monitoring device and each second monitoring device in the monitoring network other than the first monitoring device is calculated to obtain the data redundancy between devices; The maintenance strategy simulation module receives candidate maintenance strategies, maps them to the data regions to be repaired in the original spatiotemporal data matrix, and uses a matrix completion algorithm to simulate the network data reconstruction effect after maintenance, thereby obtaining the complete data matrix after completion; and calculates the performance pre-evaluation results. The multi-objective optimization decision module establishes and solves a multi-objective optimization model based on the data value index, data redundancy between devices, and performance pre-assessment results, generates the optimal maintenance plan, and provides the plan to the execution and verification feedback module. The execution and verification feedback module is used to execute the optimal maintenance plan, verify the recovery effect of network data quality after maintenance, and obtain the verification results.

2. The self-maintenance system for pollution source monitoring equipment according to claim 1, characterized in that: In the data acquisition and preprocessing module, the preprocessing includes data standardization and coarse noise filtering.

3. The self-maintenance system for pollution source monitoring equipment according to claim 2, characterized in that: The maintenance strategy simulation module is specifically used to perform the following operations: Receive candidate maintenance strategies, each of which specifies the set of monitoring devices to be maintained and their corresponding future maintenance period; Each of the candidate maintenance strategies is mapped to the original spatiotemporal data matrix, where the matrix elements represent the standardized data of the corresponding monitoring device at the corresponding time. The data points corresponding to the monitoring equipment to be maintained specified in the candidate maintenance strategy during the future maintenance period are marked as high-quality known data points, forming a known data area. Data outside the known data area is unknown data. Based on the high-quality known data points in the known data region, the original spatiotemporal data matrix is ​​completed and reconstructed using a non-negative matrix factorization algorithm; The effectiveness of the candidate maintenance strategy is quantitatively pre-evaluated by calculating the root mean square error between the complete data matrix obtained after completion and reconstruction and the original spatiotemporal data matrix at the unknown data locations.

4. The self-maintenance system for pollution source monitoring equipment according to claim 3, characterized in that: The multi-objective optimization decision module is specifically used for: Establish a multi-objective optimization model with the goal of minimizing maintenance costs and maximizing the overall data quality of the network; Define binary decision variables and the objective function of the model. Define the constraints of the multi-objective optimization model, wherein the total maintenance cost of all monitoring devices selected for the maintenance plan does not exceed the preset budget; A multi-objective evolutionary algorithm was used to solve the multi-objective optimization model, and a set of Pareto optimal solutions were obtained. Based on the current preset budget, the final optimal maintenance plan is selected from the Pareto optimal solutions.

5. A self-maintenance system for pollution source monitoring equipment according to claim 4, characterized in that: The execution and verification feedback module is specifically used for: The optimal maintenance plan generated by the multi-objective optimization decision module is then deployed and executed. After the maintenance task is completed, the actual data of the monitoring network is collected as verification data. The data of the maintained monitoring equipment in the verification data is used as known data points. The matrix completion algorithm is used again to infer the data of the unmaintained equipment to obtain the inferred data matrix. The actual root mean square error between the inferred data matrix and the data corresponding to the unmaintained equipment in the verification data is calculated and used as the score of the actual effect after maintenance. The actual performance score is compared with the predicted performance score obtained by the maintenance strategy simulation module in its pre-evaluation of the optimal maintenance plan to obtain the deviation information between the two. If the deviation between the actual effect score and the predicted effect score continues to exceed a preset threshold, the deviation information will be fed back to the equipment value and redundancy assessment module and the multi-objective optimization decision module. The equipment value and redundancy assessment module adjusts the smoothing parameters of the probability distribution estimation algorithm when calculating the data value index and redundancy based on the deviation information. The multi-objective optimization decision module adjusts the weight coefficients of the objective function in the optimization model based on the deviation information.

Images