Remote Operation and Maintenance Management and Fault Early Warning System and Method for Environmental Protection Equipment Clusters

Abstract

This invention discloses a remote operation and maintenance management and fault early warning system and method for environmental protection equipment clusters, specifically relating to the fields of industrial automation and environmental protection technology. It includes a multi-dimensional equipment data acquisition module, a multi-dimensional equipment health model construction module, an equipment cluster collaborative fault reasoning module, an equipment predictive operation and maintenance module, an equipment adaptive early warning threshold dynamic adjustment module, and an equipment early warning closed-loop verification and optimization module. The multi-dimensional equipment data acquisition module uses sensor technology to collect performance, loss, and operating condition data from each piece of equipment, obtaining a multi-dimensional equipment parameter set. This invention establishes process relationships between equipment based on knowledge graph technology, collaboratively analyzes the correlation parameters of upstream and downstream equipment, accurately locates the root cause, and provides handling suggestions. Through adaptive early warning threshold dynamic adjustment, sufficient advance intervention time is provided, making early warnings more accurate and reducing false alarms and missed alarms.

Description

Technical Field

[0001] This invention relates to the fields of industrial automation and environmental protection technology, specifically to a remote operation and maintenance management and fault early warning system and method for environmental governance equipment clusters. Background Technology

[0002] In the fields of industrial production and environmental protection, environmental protection equipment plays a crucial role. It is mainly used to treat various pollutants, including waste gas, wastewater and solid waste. Through remote operation and maintenance management, technicians can adjust parameters and update programs without being on-site, thus achieving one-stop management.

[0003] Existing remote operation and maintenance management and fault early warning methods include independent monitoring mode, simple alarm judgment and fixed periodic maintenance; traditional operation and maintenance methods usually monitor exhaust gas, wastewater and solid waste treatment equipment independently, and operation and maintenance personnel need to operate and monitor various types of equipment in different central control rooms; in terms of equipment status judgment, they mainly rely on simple "normal / fault" alarm mechanism; equipment maintenance and spare parts replacement cycles are usually determined based on fixed time or rough experience.

[0004] There are many pain points in the operation and maintenance of existing environmental protection equipment: independent monitoring of waste gas, wastewater, and solid waste treatment equipment makes it impossible to form a unified perspective, resulting in low management efficiency; equipment status is judged only by simple "normal / fault" alarms, which cannot quantify the progressive wear and tear of key components such as activated carbon saturation, membrane fouling degree, and pump and valve wear, resulting in insufficient preventive maintenance; alarms are mostly threshold-triggered, and when an alarm occurs, the fault has often already occurred or is about to occur, which cannot provide enough time for early intervention, leading to unplanned downtime or excessive emissions; reliance on manual experience: equipment maintenance and spare parts replacement cycles are based on fixed time or rough experience, which is not precise enough and may lead to "over-maintenance" causing waste or "under-maintenance" causing faults. Therefore, there is a need for an intelligent method for remote centralized monitoring of environmental protection equipment clusters, multi-dimensional health assessment, predictive maintenance, and collaborative fault early warning. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a remote operation and maintenance management and fault early warning system and method for environmental protection equipment clusters, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a remote operation and maintenance management and fault early warning system for environmental protection equipment clusters, comprising:

[0007] Multi-dimensional equipment data acquisition module: Through sensor technology, it collects performance data, loss data and operating condition data of each device, and obtains a multi-dimensional equipment parameter set after preprocessing;

[0008] Multi-dimensional equipment health model construction module: Based on a multi-dimensional equipment parameter set, calculate the performance dimension score, loss dimension score and operating condition dimension score of each equipment respectively, and then perform comprehensive analysis to construct a multi-dimensional health index model for each equipment. The multi-dimensional health index of each equipment output by the model is then transmitted to the equipment cluster collaborative fault reasoning module.

[0009] Equipment cluster collaborative fault reasoning module: Based on knowledge graph technology, it defines the entity relationships between equipment, constructs an environmental protection process knowledge graph, performs collaborative analysis on fault warning signals issued by each equipment through multi-dimensional health indices, and transmits the collaborative analysis results to the equipment predictive operation and maintenance module;

[0010] Predictive Maintenance Module: Based on the collaborative analysis results of multi-dimensional health indices of each piece of equipment and environmental protection process knowledge graph, the health status of the equipment is predicted, and the predictive maintenance results are obtained based on the prediction results.

[0011] The device adaptive early warning threshold dynamic adjustment module: Based on the load of each device, the basic early warning threshold of the multi-dimensional health index of each device is adaptively and dynamically adjusted, and the obtained dynamic early warning threshold is transmitted to the device predictive operation and maintenance module.

[0012] Equipment early warning closed-loop verification and optimization module: Verifies the equipment health prediction results, optimizes the multi-dimensional health index model of each device based on the verification results, and feeds back the optimized model parameters to the multi-dimensional equipment health model construction module.

[0013] A preferred method for remote operation and maintenance management and fault early warning of environmental protection equipment clusters includes the following steps:

[0014] S1: Using sensor technology, collect performance data, loss data, and operating condition data of each device, and obtain a multi-dimensional set of device parameters after preprocessing.

[0015] S2: Based on a multi-dimensional set of equipment parameters, calculate the performance dimension score, loss dimension score and operating condition dimension score of each equipment respectively, and then conduct a comprehensive analysis to construct a multi-dimensional health index model for each equipment to obtain the multi-dimensional health index of each equipment.

[0016] S3: Based on knowledge graph technology, define the entity relationships between equipment, construct an environmental protection process knowledge graph, and conduct collaborative analysis on the fault warning signals issued by the equipment through the environmental protection process knowledge graph to obtain the collaborative analysis results of the environmental protection process knowledge graph;

[0017] S4: Based on the collaborative analysis results of the multi-dimensional health index of each device and the knowledge graph of environmental protection process, the health of the device is predicted, and the predictive operation and maintenance results of the device are obtained based on the prediction results.

[0018] S5: Based on the load of each device, the basic early warning threshold of the multi-dimensional health index of each device is adaptively and dynamically adjusted, and the obtained dynamic early warning threshold is fed back to the prediction process in S4.

[0019] S6: Verify the equipment health prediction results, optimize the multi-dimensional health index model of each device based on the verification results, and feed the optimized model parameters back to the multi-dimensional equipment health model.

[0020] The technical effects and advantages of this invention are as follows:

[0021] 1. This invention constructs a multi-dimensional equipment health (PHM) model: innovatively constructing a model that can quantitatively assess the health status of environmental protection equipment from three dimensions: processing performance, material loss, and equipment operating conditions, and calculating the multi-dimensional health index (HI) of each piece of equipment in real time, thereby realizing the transformation from "fault alarm" to "health warning".

[0022] 2. This invention establishes process relationships between equipment based on knowledge graph technology. When a unit issues an early warning, the system can collaboratively analyze the correlation parameters of upstream and downstream equipment, accurately locate the root cause, provide handling suggestions, and automatically generate purchase orders and maintenance work orders to achieve precise management of spare parts inventory.

[0023] 3. This invention dynamically adjusts the early warning threshold based on the current load of the equipment, making the early warning more accurate, reducing false alarms and missed alarms, providing sufficient time for early intervention, and enabling predictive maintenance to reduce over-maintenance and failure losses, thereby improving the intelligence of operation and maintenance management and fault early warning of environmental protection equipment. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the overall process of the present invention.

[0025] Figure 2 This is a schematic diagram of the method flow of the present invention.

[0026] Figure 3 This is a flowchart for constructing the environmental protection process knowledge graph of the present invention. Detailed Implementation

[0027] 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. Example 1:

[0028] Please see Figure 1 As shown, the present invention provides a remote operation and maintenance management and fault early warning system for environmental protection equipment clusters, including a multi-dimensional equipment data acquisition module, a multi-dimensional equipment health model construction module, an equipment cluster collaborative fault reasoning module, an equipment predictive operation and maintenance module, an equipment adaptive early warning threshold dynamic adjustment module, and an equipment early warning closed-loop verification and optimization module.

[0029] The multi-dimensional device data acquisition module is connected to the multi-dimensional device health model construction module. The device cluster collaborative fault reasoning module is connected to the device predictive operation and maintenance module and the multi-dimensional device health model construction module, respectively. The device adaptive early warning threshold dynamic adjustment module is connected to the device predictive operation and maintenance module and the device early warning closed-loop verification and optimization module, respectively. The device early warning closed-loop verification and optimization module is connected to the multi-dimensional device health model construction module.

[0030] Multi-dimensional equipment data acquisition module: Utilizing sensor technology, this module collects performance, loss, and operating condition data for each piece of equipment. After preprocessing, it obtains a multi-dimensional equipment parameter set, including the following steps:

[0031] The data collection involves: collecting the inlet and outlet values ​​of all target pollutants to be processed by each device and the processing volume Q of all target pollutants per unit time through device performance dimension sensors; and collecting the key consumables dataset D to be monitored by each device and the cumulative operating time h of each device body through device wear dimension sensors, where D=[d1,d2,...,d...]. i ,...,d n ], d i For each device, there is the i-th key consumable data, where n is the number of classes of key consumable data; the operating status dataset D1, where D1=[a1,a2,...,a3], is collected from sensors along the operating condition dimension to monitor the operating status of each device. i ,...,a n1 ], a i Let n1 be the i-th operating status data of each device, and n1 be the number of operating status data.

[0032] This embodiment requires specific explanation of the various devices, including those for waste gas, wastewater, and solid waste. Waste gas pollutants include particulate matter, SOx / NOx, and VOCs, and waste gas equipment includes, for example, scrubbing towers, activated carbon adsorption, and RTO. Wastewater pollutants include suspended solids, COD / BOD, nitrogen and phosphorus, and heavy metals, and wastewater equipment includes, for example, sedimentation tanks and aeration tanks. Solid waste pollutants include general solid waste, hazardous waste, and domestic waste, and solid waste equipment includes, for example, crushers, sorting machines, incinerators, and solidification equipment. In this invention, waste gas treatment capacity refers to the volume of gas treated, wastewater treatment capacity refers to the volume of water treated, and solid waste pollutant treatment capacity refers to the weight of solids treated.

[0033] This embodiment requires specific explanation of the key data to be monitored, which can be obtained through industry management standards, environmental protection equipment design requirements, etc., including key consumable data and equipment operating status data. Data is collected through various equipment performance sensors, including but not limited to collecting water / gas pollutant concentration data (e.g., VOCs concentration, COD concentration, SO2 concentration, etc.) at the inlet and outlet of the environmental protection equipment through online water / gas analyzers, and collecting pollutant treatment flow rate Q through flow meters. Data is collected through equipment wear sensors, including but not limited to collecting the gas pressure difference between the inlet and outlet of the activated carbon bed through activated carbon bed differential pressure sensors, and monitoring reagent consumption through metering pump stroke counters. Data is collected through operating condition sensors, including but not limited to collecting equipment temperature through temperature sensors, collecting equipment operating current and voltage through current and voltage sensors, collecting valve opening through valve positioners, and collecting vibration amplitude through vibration sensors.

[0034] The preprocessing of the collected multi-dimensional data includes data cleaning and deduplication standardization. The data cleaning package removes invalid and outlier values, and the data deduplication refers to deduplication based on the primary key. For example, a unique primary key (such as device ID + timestamp) is generated for each data record, and records with duplicate primary keys are deleted to obtain a multi-dimensional device parameter set.

[0035] Multi-dimensional equipment health model construction module: Based on a multi-dimensional equipment parameter set, calculate the performance dimension score, loss dimension score and operating condition dimension score of each equipment respectively, and then perform comprehensive analysis to construct a multi-dimensional health index model for each equipment. The multi-dimensional health index of each equipment output by the model is then transmitted to the equipment cluster collaborative fault reasoning module.

[0036] The performance evaluation of each device is as follows: Based on the inlet and outlet values ​​of all target pollutants to be treated by each device, the ratio of the actual removal rate of each target pollutant to the corresponding design removal rate is obtained to obtain the removal rate index wr of each target pollutant. All target pollutants of each device are traversed, and the removal rate indices wr of each target pollutant are added together with equal weights to obtain the removal rate index WR of all target pollutants of each device, with a weighted sum of 1. Based on the processing volume Q of all target pollutants per unit time of each device, the ratio of the processing volume Q of each target pollutant per unit time to the corresponding design processing volume is obtained to obtain the treatment efficiency index of each target pollutant. All target pollutants of each device are traversed, and the treatment efficiency index wx of each target pollutant is added together with equal weights to obtain the treatment efficiency index WX of all target pollutants of each device, with a weighted sum of 1. Finally, the performance score SC1 of each device is calculated, SC1 = a1 × WR + (1 - a1) × WX, where a1 is the weight of the removal rate index WR of all target pollutants of each device, with a value range of 0.6 to 0.8, determined by training with historical data.

[0037] The equipment loss dimension scoring is as follows: Based on the key consumable data set D required for monitoring each equipment, the ratio of the i-th key consumable data to the corresponding maximum allowable consumable data is obtained to obtain the loss index si of the i-th key consumable data. All key consumable data are traversed, and the loss indices si of each key consumable data are added together with equal weights to obtain the loss index SI of all key consumable data of each equipment, with a weight sum of 1. Based on the cumulative operating time h of each equipment body, the ratio of the cumulative operating time t of each equipment body to the corresponding design life time is obtained to obtain the operating loss index YI of each equipment. Finally, the loss dimension score SC2 of each equipment is calculated, SC2=a2×(1-SI)+(1-a2)×(1-YI), where a2 is the weight of the loss index of all key consumable data of each equipment, with a value range of 0.7 to 0.9, determined by training with historical data.

[0038] The scoring for each equipment operating condition dimension is as follows: Based on the operating status dataset D1 that needs to be monitored for each equipment, it is divided into a negative indicator operating status dataset and a deviation indicator operating status dataset. For the negative indicator operating status data, the positive score for each negative indicator operating status data is obtained through the expression max[0,1-(actual value / maximum allowed value)]. For the deviation indicator operating status data, the positive score for each deviation indicator operating status data is obtained through the expression 1-(|actual value-set value| / range). Finally, all operating status data are traversed and added together with equal weights to obtain the operating condition dimension score SC3 for each equipment, with a weight sum of 1.

[0039] In this embodiment, it should be specifically noted that the larger the negative index value, the worse the health condition of the equipment and the greater the negative impact on the overall health (HI). For example, the larger the vibration amplitude, the greater the vibration, indicating that the mechanical equipment is less balanced. The further the deviation index value deviates from its ideal or set value (whether it is too large or too small), the worse the health condition. There is an optimal operating range. For example, the current and voltage need to be stable near the rated value. Too high or too low values ​​will affect the equipment's lifespan and performance.

[0040] The multi-dimensional health index model for each device is as follows: The performance score SC1, the wear dimension score SC2, and the operating condition dimension score SC3 of each device are weighted and fused to construct a multi-dimensional health (PHM) model HI for each device, where HI = b1×SC1 + b2×SC2 + b3×SC3, HI is the multi-dimensional health index of each device, and b1, b2, and b3 are the corresponding weights, for example, b1 = 0.4, b2 = 0.35, and b3 = 0.25.

[0041] Equipment cluster collaborative fault reasoning module: Based on knowledge graph technology, it defines the entity relationships between equipment, constructs an environmental protection process knowledge graph, performs collaborative analysis on fault warning signals issued by each equipment through multi-dimensional health indices, and transmits the collaborative analysis results to the equipment predictive operation and maintenance module;

[0042] Please see Figure 3 As shown, the environmental protection process knowledge graph includes:

[0043] A1: Environmental Protection Process Knowledge Graph Entity Construction: Based on the design data of environmental protection equipment clusters, five core entities are defined for the environmental protection process knowledge graph, serving as nodes in the knowledge graph, including the equipment entity set E. de Process entity set E pr Parameter entity collection E pa Faulty entity set E fa and the consumables entity set E ma E de =[e d 1 ,e d 2 ,...,e d m ], e d m Let m be the m-th entity, where m is the number of environmental protection equipment cluster types. Define the attributes of each equipment entity, including equipment ID, equipment name, model, installation location, operating status (start, stop, and fault), and operating design parameters (rated current, rated voltage, valve opening setting, etc.). For example, taking an activated carbon adsorption box as an example, its equipment ID can be defined as "AC-101", equipment name as "activated carbon adsorption equipment", model as "XZ-500", operating design parameters include "processing air volume 5000m³ / h, activated carbon filling amount 2000kg, rated voltage, rated current", location as "Area 3 of the waste gas treatment workshop", and operating status as "starting". pr =[e p 1 ,e p 2 ,...,e p L ], e p LLet L be the Lth process entity, where L is the number of process entity types. Define the attributes of each process entity, including process ID, process name, design treatment standard, upstream process ID, downstream process ID, and maximum allowable load. For example, the process ID of "VOCs adsorption process" is set to "VAP-001", the process name is "VOCs adsorption process", the upstream is "waste gas collection process ID", the downstream is "waste gas emission process ID", and the treatment standard is "outlet VOCs concentration ≤ 50 mg / m³". pa =[e pa 1 ,e pa 2 ,...,e pa o ], e pa o Let O be the 0th parameter entity, where O is the number of parameter entity types. Define the attributes of each parameter entity, including parameter ID, parameter name, parameter unit, data source (sensor ID), sampling frequency, design upper limit, design lower limit, and warning threshold. Parameter entities include not only equipment operating status data but also parameters from environmental protection processes. For example, a parameter entity like "Imported VOCs Concentration" might have the parameter ID "IVC-001," the parameter name "Concentration Parameter," the unit "mg / m³," the design range "0-1000mg / m³," the sampling frequency "1 time / minute," and an ultra-high concentration warning of 600mg / m³. fa =[e f 1 ,e f 2 ,...,e f p ], e pa p Let p be the p-th fault entity, where p is the number of fault entity types. Define the attributes of each fault entity, including fault ID, fault name, scope of impact, possible causes, handling suggestions, and average repair time. For example, the fault entity "Adsorption Saturation Fault" has the fault ID "ASF-001", the fault name "Activated Carbon Adsorption Saturation", the scope of impact is "Activated carbon adsorption box cannot effectively adsorb VOCs, resulting in excessive outlet concentration", the possible causes are "Upstream concentration is too high, activated carbon was not replaced in time", the handling suggestion is "Replace activated carbon", and the average repair time is 4 hours. ma =[e m 1 ,e m 2 ,...,e m q ], e m qLet q be the q-th consumable entity, and p be the number of consumable entity types. Define the attributes of each consumable entity, including consumable ID, consumable name, specifications, unit, average consumption rate, current inventory, and safety stock threshold. For example, the consumable entity name is "Activated Carbon", the ID is "AC-001", the specifications are 100×100×100mm, and the unit is m. 3 The average consumption rate is 0.5m. 3 / month, current inventory 2.5m 3 Safety stock threshold 1m 3 The entire set of entities in the knowledge graph is E, and E = E de ∪E pr ∪E pa ∪E fa ∪E ma ];

[0044] This embodiment requires specific explanation of the design documents for, for example, environmental protection equipment clusters, which include information such as equipment layout and functional design; equipment operation manuals, which clearly define the equipment's operating specifications and performance parameters; historical operation and maintenance records, which reflect the equipment's status changes, fault conditions, and maintenance operations during actual operation; unification of different descriptions of the same entity (e.g., unifying "activated carbon adsorption box" as "activated carbon adsorption box"); and assigning a unique ID to each entity (e.g., equipment ID rule: equipment type abbreviation + area number + serial number).

[0045] A2: Construction of Entity Relationships in Environmental Protection Process Knowledge Graph: Based on the process flow diagrams and text descriptions in the design documents of environmental protection equipment clusters, nine core entity relationships R are defined as edges of the knowledge graph, including equipment-process association, upstream process-downstream process connection, equipment-parameter monitoring, parameter-fault triggering, equipment-consumable data, fault equipment association, process-parameter requirements, auxiliary equipment-main equipment collaboration, and fault solution association. Each entity relationship is associated with a subject-verb-object SPO triple (text recognition is achieved through the BERT model of natural language processing technology). The relationship direction is the forward and backward direction of each entity relationship. For example, equipment-process association: activated carbon adsorption box-execution-VOCs adsorption process triple, the relationship direction is equipment→process, upstream process→downstream process, auxiliary equipment→main equipment, fan drives activated carbon adsorption box, parameter-fault triggering is activated carbon adsorption box triggering adsorption saturation;

[0046] This embodiment specifically explains how, by analyzing the process flow diagrams and text descriptions in the design documents, the direct process relationships such as the connection sequence between equipment and the direction of material flow are determined. For example, in the wastewater treatment process, the sequential treatment relationship between the "bar" and the "grit chamber" is clarified.

[0047] A3: Construction and Storage of Environmental Protection Process Knowledge Graph: Based on the complete entity set E of the entire knowledge graph and the core entity relations R, construct the environmental protection process knowledge graph G. sc G sc =[E,R]; A hybrid storage architecture of graph database and relational database is adopted. The graph database (e.g., Neo4j) stores SPO triples and entity attributes, supporting efficient association queries and path analysis. The relational database (e.g., MySQL) stores structured data in the environmental protection process, such as parameter time series data, fault history records, etc., and is associated with the graph database through entity ID;

[0048] A4: Environmental Technology Knowledge Graph Quality Assessment: Entity Accuracy (AC) is obtained based on the ratio of correctly linked entities to the total number of entities. E The entity relation accuracy AC is obtained by comparing the ratio of the correctly defined number of entity relations to the total number of entity relations. R Based on the total number of entities |E| and the number of entity relation triples |S|, the knowledge graph density DE is obtained, DE=|S| / [|E|×(|E|-1)], which is used to measure the connectivity of the knowledge graph. When any indicator is lower than the corresponding preset threshold, for abnormal entity accuracy, the entity text matching optimization process is triggered; for abnormal entity relation accuracy, the triple data correction optimization process is triggered; and for abnormal knowledge graph density, the missing entity relation completion optimization process is triggered.

[0049] This embodiment requires a detailed explanation of the calculation of entity accuracy and relation accuracy. It is based on a benchmark dataset manually labeled and corrected by domain experts. This dataset is randomly sampled from the full graph (e.g., 1000 entities and 2000 relations), covering various entities and their relations, such as equipment, processes, parameters, and faults. Using an entity alignment algorithm, the system automatically constructs a set of entities to be evaluated, matching them with entities in the benchmark dataset. The number of entities whose key attributes match is counted as the number of correctly linked entities. Based on entity alignment, relation triples are further matched, and the number of triples whose subject, predicate, and object all match the benchmark dataset is counted as the number of correctly defined entity relations.

[0050] A5: Environmental Protection Process Knowledge Graph Update: When a new entity or new entity relationship New_Data is detected. t At that time, the knowledge graph update function G is triggered. t+1 G t+1 =G t ∪f ex (New_Data t )\f de (Obsolete_Data t ), G t+1 and Gt Obsolete_Data represents the knowledge graph state at time points t and t+1, i.e., the set of all correct triples at that time. ∪ is the union operator, indicating the addition of new knowledge, and \ is the difference operator, indicating the deletion of old knowledge. t (The triplets marked as obsolete originate from the scrap list in the equipment management system, status monitoring results such as reports of equipment being offline for extended periods from the monitoring system, and process change notifications), f ex For the knowledge extraction function, a direct mapping method based on environmental protection process knowledge graph rules is used for structured data to extract new triples. For unstructured text data, entity relations are extracted using a BERT model. de As a knowledge obsolescence function, it uses instructions from the equipment management system, instructions from the monitoring system, and process change notifications to generate a set of triples to be obsolescence;

[0051] The results of the collaborative analysis include:

[0052] B1: If the target device's multi-dimensional health index is greater than or equal to the corresponding threshold HI th (Based on historical data) This indicates that the health of each piece of equipment is good. Conversely, if the equipment issues a fault warning signal, the environmental protection process knowledge graph is retrieved to determine whether the fault is triggered by a single abnormal upstream parameter or by the combined influence of multiple upstream parameters. The correlation influence coefficient η1 for a single upstream parameter and the correlation influence coefficient η2 for multiple upstream parameters are calculated respectively. X j and X j0 These represent the actual values ​​and design processing standards of the upstream parameters, respectively. j The weight of the j-th parameter is obtained from historical fault data. J represents the number of upstream parameters affecting downstream faults, and η1 j Let be the single-parameter correlation influence coefficient of the j-th parameter;

[0053] B2: Based on path analysis using knowledge graphs, calculate the failure probability P for each potential root cause parameter. j , L represents the number of associated paths from the potential root cause parameter j to the target fault, and ω represents the number of paths. k η is the weight of the length of the k-th associated path. jk The path length is the correlation coefficient of the j-th root cause parameter on the k-th path. The path length refers to the number of relational edges traversed from the potential root cause parameter j to the target fault. The shorter the path, the greater the weight. For example, the direct correlation path ω k The indirect association path ω is 1. kThe probability was 0.6; finally, the root cause was identified as the one with the highest probability, and treatment suggestions were given. The results of the collaborative analysis of the environmental protection process knowledge graph were obtained, including the parameter correlation influence coefficient, the failure probability of each potential root cause parameter, the root cause, and treatment suggestions.

[0054] B3: Influence Weight Optimization: Minimize the predicted failure probability P using gradient descent with the number of historical failure samples N. I,预测 And the actual failure probability P I,实际 Deviation minLoss, The weights w are optimized by taking the derivative of the loss function and considering the influence of the parameters. j ;

[0055] Predictive Maintenance Module: Based on the collaborative analysis results of multi-dimensional health indices of each piece of equipment and environmental protection process knowledge graph, the health status of the equipment is predicted, and the predictive maintenance results are obtained based on the prediction results.

[0056] The HI index based on the multi-dimensional health index of each device now The health status of the equipment is predicted, and the initial health status decline time prediction T is obtained. HI th The basic early warning threshold, K, representing the HI (hygiene degradation) rate, is fitted from historical equipment operating data. The correlation coefficient η from the collaborative analysis results of the environmental process knowledge graph is used to obtain the corrected health degradation time T. now T now =T×(1-η), where η includes the correlation coefficient η1 of a single upstream parameter and the correlation coefficient η2 of multiple upstream parameters; T is adjusted based on the time of health decline. now By combining the root causes and treatment suggestions from the environmental protection process knowledge graph, predictive maintenance results for equipment are obtained. For example, if the model predicts that the HI will drop to the warning line of 60% after 120 hours, the outlet concentration may exceed the standard. The system retrieves the correlation coefficient between the upstream exhaust gas concentration data and the equipment, automatically corrects the predicted replacement time to 90 hours later, and suggests arranging replacement. Combined with the inventory, it automatically sends an activated carbon purchase suggestion to the procurement system.

[0057] The device adaptive early warning threshold dynamic adjustment module: Based on the load of each device, the basic early warning threshold of the multi-dimensional health index of each device is adaptively and dynamically adjusted, and the obtained dynamic early warning threshold is transmitted to the device predictive operation and maintenance module.

[0058] Based on the load of each device, the basic early warning thresholds of the multi-dimensional health index of each device are adaptively adjusted to obtain the dynamic early warning threshold HI. d HI d =HI th×[1-α×(current load / maximum allowable load)], where α is the load impact coefficient (generally 0.1-0.2), and the load of each device refers to the amount of pollutants it handles during the current actual operation. The dynamic early warning threshold is then transmitted to the device predictive maintenance module.

[0059] Equipment early warning closed-loop verification and optimization module: Verifies the equipment health prediction results, optimizes the multi-dimensional health index model of each device based on the verification results, and feeds back the optimized model parameters to the multi-dimensional equipment health model construction module;

[0060] To verify the equipment health prediction results, the prediction time deviation rate ε is first defined. T now and T 实际 Let Loss1 be the predicted time to reach the threshold and the actual time to reach the threshold, respectively. The number of historical fault samples is N; then, the partial derivative of the loss function Loss1 with respect to the weight b1 is obtained. Update weight b1 along the gradient direction now b1 now =b1-λ×(∂Loss1 / ∂b1), where λ is the learning rate (controlling the step size of weight updates, usually taken as 0.01~0.1). Δb1 is the increment of b1 (this value is extremely small, for example, 10). -5 The logic for obtaining the partial derivative of weight b2 is the same as that for weight b2, thus obtaining the updated weight b2. now The updated weights of b3 now We obtained the optimized model parameters for the multi-dimensional health index model of each device. Example 2:

[0061] For parts not described in detail in this embodiment, please refer to the description in Embodiment 1. The method for providing remote operation and maintenance management and fault early warning of environmental protection equipment clusters includes the following steps:

[0062] S1: Using sensor technology, collect performance data, loss data, and operating condition data of each device, and obtain a multi-dimensional set of device parameters after preprocessing.

[0063] S2: Based on a multi-dimensional set of equipment parameters, calculate the performance dimension score, loss dimension score and operating condition dimension score of each equipment respectively, and then conduct a comprehensive analysis to construct a multi-dimensional health index model for each equipment to obtain the multi-dimensional health index of each equipment.

[0064] S3: Based on knowledge graph technology, define the entity relationships between equipment, construct an environmental protection process knowledge graph, and conduct collaborative analysis on the fault warning signals issued by the equipment through the environmental protection process knowledge graph to obtain the collaborative analysis results of the environmental protection process knowledge graph;

[0065] S4: Based on the collaborative analysis results of the multi-dimensional health index of each device and the knowledge graph of environmental protection process, the health of the device is predicted, and the predictive operation and maintenance results of the device are obtained based on the prediction results.

[0066] S5: Based on the load of each device, the basic early warning threshold of the multi-dimensional health index of each device is adaptively and dynamically adjusted, and the obtained dynamic early warning threshold is fed back to the prediction process in S4.

[0067] S6: Verify the equipment health prediction results, optimize the multi-dimensional health index model of each device based on the verification results, and feed the optimized model parameters back to the multi-dimensional equipment health model.

[0068] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0069] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A remote operation and maintenance management and fault early warning system for a cluster of environmental governance equipment, characterized in that: The method comprises the following steps: A multi-dimensional equipment data acquisition module: through sensor technology, the performance dimension data, loss dimension data and working condition dimension data of each equipment are collected respectively, and after preprocessing, a multi-dimensional equipment parameter set is obtained; A multi-dimensional equipment health model construction module: based on the multi-dimensional equipment parameter set, the performance dimension score, loss dimension score and working condition dimension score of each equipment are calculated, and then a comprehensive analysis is performed to construct a multi-dimensional health index model of each equipment, and the multi-dimensional health index of each equipment output by the model is transmitted to the equipment cluster collaborative fault reasoning module; An equipment cluster collaborative fault reasoning module: based on knowledge graph technology, the entity relationship between equipment is defined, an environmental protection process knowledge graph is constructed, the fault warning signals sent through the multi-dimensional health index of each equipment are analyzed collaboratively, and the collaborative analysis results are transmitted to the equipment predictive operation and maintenance module; The collaborative analysis results include: B1: If the target device multi-dimensional health index is greater than or equal to the corresponding threshold value HI th , it indicates that the health of each device is good, otherwise, a fault warning signal is sent to the device, the environmental protection process knowledge graph is called, it is determined whether the device fault is triggered by a single upstream parameter abnormality or jointly affected by multiple upstream parameters, a single upstream parameter correlation influence coefficient η1 and a multiple upstream parameter correlation influence coefficient η2 are calculated respectively, , X j and X j0 are the actual values and design processing standards of the upstream parameters respectively, w j is the jth parameter influence weight, J is the number of upstream parameters affecting the downstream fault, η1 j is the single parameter correlation influence coefficient of the jth parameter; B2: Path analysis based on knowledge graph, calculate the failure probability P of each potential root cause parameter j ; the one with the highest final positioning probability is the root cause, and treatment suggestions are given, and the environmental protection process knowledge graph collaborative analysis result is obtained, including parameter correlation influence coefficient, failure probability of each potential root cause parameter, root cause and treatment suggestions; B3: Impact weight optimization: minimize prediction failure probability P by historical failure sample number N using gradient descent method I,预测 and actual failure probability P I,实际 bias minLoss, , optimize parameter impact weight w by derivation of Loss function j ; An equipment predictive operation and maintenance module: based on the multi-dimensional health index of each equipment and the collaborative analysis results of the environmental protection process knowledge graph, the health of the equipment is predicted, and the predictive operation and maintenance results of the equipment are obtained according to the prediction results; The device predictive operation and maintenance result is based on a multi-dimensional health index HI of each device now , the initial health degradation time prediction T is obtained by predicting the health of the device, , HI th is a basic warning threshold, K is a HI degradation rate; an associated influence coefficient η in a collaborative analysis result of an environmental protection process knowledge graph, a health degradation time correction T is obtained now , T now =T×(1-η), η includes a single upstream parameter associated influence coefficient η1 and a plurality of upstream parameter associated influence coefficients η2; based on the health degradation time correction T now and the root cause and treatment suggestion in the collaborative analysis result of the environmental protection process knowledge graph, a device predictive operation and maintenance result is obtained; An equipment self-adaptive early warning threshold dynamic adjustment module: based on the load of each equipment, the basic early warning threshold of the multi-dimensional health index of each equipment is adaptively and dynamically adjusted, and the dynamic early warning threshold obtained is transmitted to the equipment predictive operation and maintenance module; The dynamic early warning threshold is based on the load of each device, and the basic early warning threshold of the multi-dimensional health index of each device is adaptively adjusted to obtain a dynamic early warning threshold HI d , HI d =HI th ×[1-α×(current load / max allowable load)], HI th is the basic early warning threshold, α is the load influence coefficient, and the load of each device refers to the amount of pollutants borne in the current actual operation process. An equipment early warning closed-loop verification and optimization module: the prediction results of the health of the equipment are verified, the multi-dimensional health index model of each equipment is optimized according to the verification results, and the optimized model parameters are fed back to the multi-dimensional equipment health model construction module. 2.The remote operation and maintenance management and fault early warning system of the environmental protection governance device cluster of claim 1, characterized in that: The performance dimension score of each equipment: based on the inlet and outlet values of all target pollutants required to be treated by each equipment, the ratio of the actual removal rate of each target pollutant to the corresponding design removal rate is obtained to obtain a removal rate index wr of each target pollutant, the removal rate indexes wr of all target pollutants of each equipment are added with equal weights to obtain a removal rate index WR of all target pollutants of each equipment, and the weight sum is 1; based on the processing capacity Q of all target pollutants of each equipment per unit time, the ratio of the processing capacity Q per unit time of each target pollutant to the corresponding design processing capacity is obtained to obtain a processing efficiency index of each target pollutant, the processing efficiency indexes wx of all target pollutants of each equipment are added with equal weights to obtain a processing efficiency index WX of all target pollutants of each equipment, and the weight sum is 1; finally, the performance score SC1 of each equipment is calculated, SC1=a1×WR+(1-a1)×WX, and a1 is the weight of the removal rate index WR of all target pollutants of each equipment. 3.The remote operation and maintenance management and fault early warning system of the environmental protection governance device cluster of claim 1, characterized in that: The loss dimension score of each device: based on the key consumable data set D required to be monitored by each device, the ratio of the ith key consumable data to the corresponding allowed maximum consumable data is obtained, the loss index si of the ith key consumable data is obtained, all key consumable data loss indexes si of each device are added with equal weight to obtain the loss index SI of all key consumable data of each device, and the weight sum is 1; based on the cumulative running time h of the body of each device, the ratio of the cumulative running time t of the body of each device to the corresponding design life time is obtained, the running loss index YI of each device is obtained; finally, the loss dimension score SC2 of each device is calculated, SC2=a2×(1-SI)+(1-a2)×(1-YI), a2 is the weight of the loss index of all key consumable data of each device.

4. The remote operation and maintenance management and fault early warning system of the environmental protection governance device cluster according to claim 1, characterized in that: The working condition dimension score of each device: based on the running state data set D1 required to be monitored by each device, it is divided into negative index running state data set and deviation type index running state data set, for the negative index running state data, the positive score of each negative index running state data is obtained through the expression max[0,1-(actual value / allowed maximum value)], for the deviation type index running state data, the positive score of each deviation type index running state data is obtained through the expression 1-(|actual value-set value| / range); finally, all running state data are traversed, and the working condition dimension score SC3 of each device is obtained by adding with equal weight, and the weight sum is 1.

5. The remote operation and maintenance management and fault early warning system of the environmental protection governance device cluster according to claim 1, characterized in that: The multi-dimensional health index model of each device: the performance score SC1 of each device, the loss dimension score SC2 of each device and the working condition dimension score SC3 of each device are weighted and fused to construct the multi-dimensional health degree model HI of each device, HI=b1×SC1+b2×SC2+b3×SC3, HI is the multi-dimensional health degree index of each device, b1, b2 and b3 are the corresponding weights.

6. The remote operation management and fault early warning method of the environmental governance equipment cluster, used for the remote operation management and fault early warning system of the environmental governance equipment cluster according to any one of claims 1-5, characterized in that, The method comprises the following steps: S1: collecting each device performance dimension data, loss dimension data and working condition dimension data through sensor technology, and obtaining multi-dimensional device parameter set after preprocessing; S2: based on the multi-dimensional device parameter set, the performance dimension score, the loss dimension score and the working condition dimension score of each device are calculated, and the multi-dimensional health index model of each device is constructed by comprehensive analysis to obtain the multi-dimensional health index of each device; S3: based on the knowledge graph technology, the entity relationship between devices is defined, the environmental protection process knowledge graph is constructed, the fault warning signal issued by the device is analyzed through the environmental protection process knowledge graph, and the environmental protection process knowledge graph collaborative analysis result is obtained; S4: based on the multi-dimensional health degree index of each device and the environmental protection process knowledge graph collaborative analysis result, the health degree of the device is predicted, and the predictive operation and maintenance result of the device is obtained according to the prediction result; S5: based on the load of each device, the adaptive dynamic adjustment of the basic warning threshold of the multi-dimensional health degree index of each device is carried out, and the dynamic warning threshold obtained is fed back to the prediction process in S4. S6: verifying the equipment health degree prediction result, optimizing the multi-dimensional health degree index model of each equipment according to the verification result, and feeding back the optimized model parameters to the multi-dimensional equipment health model.

Images