Production and living ecological water safety risk intelligent evaluation method for reservoir type water source

Abstract

The invention discloses a production and living ecological water safety risk intelligent evaluation method for a reservoir type water source, and belongs to the technical field of water source safety evaluation. The method comprises the following steps: establishing a five-dimensional safety index system of water quantity safety, water quality safety, behavior safety, facility safety and management safety to obtain original data of each dimension; according to the method, intelligent means such as an intelligent weight determination method, an artificial intelligence identification method and an abnormal alarm discrimination method are adopted to generate five-dimensional safety indexes respectively, a comprehensive index is calculated by combining a comprehensive evaluation weight and a one-ticket override rule, and a safety level is determined and dynamic early warning information and disposal suggestions are generated based on the comprehensive index. According to the invention, a multi-dimensional comprehensive evaluation system is established, intelligent dynamic evaluation is realized, the water supply safety condition of the reservoir type water source can be comprehensively evaluated, the safety risk can be found and early warned in time, and powerful support is provided for fine management and scientific decision making of the water source.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of water source safety evaluation, in particular to a production and life and ecological water safety risk intelligent evaluation method for reservoir-type water sources, and is especially suitable for providing quantitative evaluation basis for the supply water safety risk grading control of large centralized reservoir-type drinking water sources. BACKGROUND

[0002] Reservoir-type water sources are important water supply sources for cities and regions, and their water supply safety is directly related to the safety of drinking water for residents, the guarantee of industrial and agricultural water use, and the satisfaction of ecological water demand. With the rapid development of the economy and society and the intensification of the impact of climate change, reservoir-type water sources are facing challenges in water quantity security, water quality safety, facility operation, behavior supervision, and management standardization, and it is urgent to establish a scientific, systematic, and intelligent safety risk evaluation method to provide technical support for the precise control and risk early warning of water sources.

[0003] The existing water source safety risk evaluation methods mainly have the following shortcomings:

[0004] First, the evaluation dimension is single. For example, patent document CN105654236A discloses a groundwater-type drinking water source pollution risk evaluation method, which mainly evaluates the pollution risk of groundwater-type water sources, couples the inherent vulnerability evaluation results of groundwater with the pollution source hazard level to evaluate the risk, and completes the risk evaluation through characteristic pollutant migration simulation. Although this method can evaluate the pollution risk of groundwater-type water sources from the time and space dimensions, it has obvious shortcomings: the evaluation object is limited to groundwater-type water sources, and does not consider the particularity of reservoir-type water sources; the evaluation content mainly focuses on pollution risk, and does not consider other important dimensions such as water quantity safety, facility safety, behavior safety, and management safety; the evaluation method mainly uses traditional vulnerability evaluation and pollution source hazard evaluation, lacks intelligent evaluation means, and is difficult to realize dynamic real-time evaluation and accurate early warning.

[0005] Second, it lacks pertinence. Most existing methods use a general risk evaluation framework and do not fully consider the unique characteristics of reservoir-type water sources. Reservoir-type water sources differ significantly from groundwater-type water sources in terms of water source characteristics, water supply methods, and risk types. Reservoir-type water sources have characteristics such as adjustable water quantity, water quality easily affected by upstream pollution, complex facilities and equipment, and high management requirements, and therefore require the establishment of an evaluation index system and evaluation method specifically for the characteristics of reservoir-type water sources.

[0006] Third, the evaluation system is imperfect. Existing evaluation methods often only focus on one or several aspects of risk, such as only evaluating water quality risk or only evaluating water quantity risk, and fail to establish a comprehensive evaluation system covering multiple dimensions such as water quantity, water quality, behavior, facilities, and management. For reservoir-type water sources, water supply safety is a system engineering, and any problem in one dimension may affect the overall water supply safety, so a multi-dimensional comprehensive evaluation system needs to be established to comprehensively assess the safety of water sources.

[0007] Fourth, the degree of intelligence is low. With the rapid development of Internet of Things, big data and artificial intelligence technology, water source monitoring generates a large amount of real-time monitoring data, but existing evaluation methods mostly use traditional statistical analysis methods, which are difficult to fully tap the value of massive monitoring data and cannot realize intelligent evaluation and dynamic early warning based on data. At the same time, the existing method lacks the ability to intelligently identify monitoring data anomalies, making it difficult to discover potential safety risks in a timely manner.

[0008] Fifth, lack of dynamic early warning mechanism. Existing evaluation methods are mostly static or periodic, making it difficult to achieve real-time monitoring and dynamic early warning of water source safety risks. The safety of water sources is affected by many factors and is in a state of dynamic change, and a dynamic evaluation and early warning mechanism based on real-time monitoring data needs to be established to timely discover risk trends and provide timely early warning information and disposal suggestions for decision-makers.

[0009] Sixth, lack of grading control basis. Although existing evaluation methods can provide risk evaluation results, they often lack clear grading standards and quantitative basis, making it difficult to provide effective support for risk grading and control. Water source management departments need to take differentiated control measures according to different risk levels, so the evaluation method needs to provide scientific, clear and operable grading standards.

[0010] Therefore, it is urgent to develop a safety risk evaluation method for reservoir-type water sources that is multi-dimensional, comprehensive and intelligent, to meet the needs of fine management and scientific decision-making of water sources in the new era. SUMMARY

[0011] The purpose of the present application is to provide an intelligent evaluation method for the production, life and ecological water safety risk of reservoir-type water sources, to solve the technical problems of single evaluation dimension, lack of pertinence, imperfect evaluation system, low degree of intelligence, lack of dynamic early warning mechanism and insufficient grading control basis in the prior art.

[0012] To achieve the above object, the application provides a reservoir-type water source production and living ecological water safety risk intelligent evaluation method, which forms a comprehensive safety risk evaluation system for reservoir-type water source production, living and ecological water by establishing a five-dimensional safety index system of water quantity safety, water quality safety, behavior safety, facility safety and management safety. The method first obtains multi-dimensional safety original data of the reservoir-type water source, including water quantity safety original data, water quality safety original data, behavior safety original data, facility safety original data and management safety original data. Then, based on the evaluation index system and intelligent evaluation method of each dimension, water quantity safety index, water quality safety index, behavior safety index, management safety index and facility safety index are generated respectively. Finally, the comprehensive index calculation method is used to generate the water supply safety comprehensive index of the reservoir-type water source by combining the comprehensive evaluation weight and the one-vote veto rule, and the safety level is determined according to the comprehensive index, and the dynamic early warning information and disposal suggestions are generated.

[0013] The technical scheme of the application has the following characteristics and advantages:

[0014] First, a five-dimensional comprehensive evaluation system is established. The application establishes an evaluation index system from five dimensions of water quantity safety, water quality safety, behavior safety, facility safety and management safety, which comprehensively covers the main factors affecting the water supply safety of the reservoir-type water source. The water quantity safety evaluation focuses on the water supply guarantee capacity of the water source, the water quality safety evaluation focuses on the water quality standard of the water source, the behavior safety evaluation focuses on the illegal and irregular behavior in the protection area, the facility safety evaluation focuses on the operation state of the key facilities such as dam, and the management safety evaluation focuses on the implementation of the management system. The five-dimensional evaluation system is mutually related and supported, forming a complete safety evaluation framework, avoiding the one-sidedness of single-dimensional evaluation.

[0015] Second, intelligent evaluation method is adopted. The application adopts corresponding intelligent evaluation technology according to the characteristics of different dimensions. In the water quality safety evaluation, the intelligent weight determination method combining information entropy and machine learning is adopted, which can dynamically adjust the weight according to the information amount of water quality monitoring data and the influence degree on water supply safety, improving the scientificity and accuracy of evaluation. In the behavior safety evaluation, the target detection and behavior recognition technology based on deep learning is adopted, which can automatically identify illegal behavior in video monitoring, greatly improving the supervision efficiency and recognition accuracy. In the facility safety evaluation, the abnormal detection method based on time series analysis is adopted, which can timely find the abnormal change of monitoring data, realizing intelligent early warning of facility safety hidden danger. These intelligent evaluation methods make full use of modern information technology and artificial intelligence technology, significantly improving the automation and intelligence level of evaluation.

[0016] Third, dynamic real-time evaluation is realized. The application is based on real-time monitoring data for evaluation, which can timely reflect the safety condition changes of the water source. By establishing a time series database and a time series prediction model, not only the current safety condition can be evaluated, but also the future risk change trend can be predicted, realizing the change from static evaluation to dynamic evaluation, from post-evaluation to pre-warning. Dynamic real-time evaluation provides more sufficient response time for risk prevention and control, which helps to eliminate the risk in the embryonic state.

[0017] Fourth, quantitative grading standards are provided. The application establishes clear five-level evaluation standards, divides the comprehensive index into five levels of excellent, good, qualified, basically qualified and unqualified, and each level corresponds to a clear score interval and color identification. This quantitative grading standard provides a scientific basis for risk grading and control, and it is convenient for management departments to take differentiated control measures according to different risk levels. At the same time, the application also sets a veto rule, sets a bottom line requirement for key indicators, and ensures that key risks are effectively controlled.

[0018] Fifth, obstacle factor identification and disposal suggestion generation functions are integrated. The application not only provides evaluation results, but also identifies the main obstacle factors affecting safety, and generates targeted disposal suggestions based on the disposal strategy knowledge base. The obstacle factor identification uses the obstacle degree model, which can quantitatively calculate the constraint degree of each factor on the comprehensive index, providing a basis for the priority ranking of disposal measures. The disposal suggestion generation module can match the corresponding disposal strategy from the knowledge base according to the characteristics of the obstacle factor, and optimize and adjust it combined with the historical disposal effect, improving the pertinence and effectiveness of the disposal suggestion.

[0019] Sixth, it has self-adaptive optimization capability. The application designs a self-adaptive dynamic evaluation module, which can dynamically adjust the evaluation parameters and monitoring strategies according to external environmental factors such as climate change and pollution risk changes. For example, adjust the evaluation weight of water quantity safety according to climate change prediction, and optimize the frequency and point of water quality monitoring according to pollution risk prediction. This adaptive mechanism enables the evaluation method to adapt to the changing external environment, maintaining the advancement and applicability of the evaluation method.

[0020] The method of the application has the characteristics of comprehensiveness, intelligence, dynamics and scientificity, and can provide strong technical support for the fine management and risk prevention and control of reservoir-type water source, and has important application value and popularization significance. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 The figure is the overall flowchart of the intelligent evaluation method for safety risk of reservoir-type water source.

[0022] Figure 2 The figure is the five-dimensional safety index calculation flowchart of the application.

[0023] Figure 3 This is a schematic diagram of the intelligent evaluation process for water quality safety index of the present invention.

[0024] Figure 4 This is a schematic diagram of the intelligent recognition process for behavioral safety index of the present invention.

[0025] Figure 5 This is a schematic diagram of the dynamic evaluation process for the facility safety index of this invention.

[0026] Figure 6 This is a schematic diagram of the comprehensive index calculation and risk warning process of this invention.

[0027] Figure 7 This is a schematic diagram of the obstacle factor identification and handling suggestion generation process of the present invention. Detailed Implementation

[0028] Please refer to Figures 1-7 The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments to make the technical solution of the present invention clearer. It should be understood that these embodiments are only for illustrating the present invention and are not intended to limit the scope of the present invention.

[0029] Reference Figure 1 The intelligent assessment method for the safety risks of production, living and ecological water use in reservoir-type water source areas of the present invention includes key links such as data acquisition, five-dimensional safety index calculation, comprehensive index generation and dynamic early warning and disposal, forming a complete intelligent assessment system.

[0030] Data acquisition module 1 is responsible for collecting multi-dimensional raw safety data from reservoir-type water sources. Specifically, data acquisition module 1 obtains raw data from five aspects from various monitoring devices, management systems, and information platforms at the reservoir-type water source.

[0031] Raw data for water security includes real-time water level data, inflow data, outflow data, evaporation data, rainfall data, water supply data, water demand data, and reservoir capacity curve data. This data is used to assess the reservoir's water security capacity and supply-demand balance. In one embodiment, water level data is collected in real-time using water level gauges installed in the reservoir, with a collection frequency of once every 15 minutes; the reservoir capacity curve data is obtained from the reservoir's design data or measured data, reflecting the reservoir's storage capacity at different water levels.

[0032] The water quality safety raw data includes spatial distribution data of the water quality monitoring station network, water quality monitoring index data of each monitoring station, and key water quality monitoring parameter data. The water quality monitoring index data covers regular indicators and characteristic pollutant indicators. The regular indicators include pH, dissolved oxygen, permanganate index, ammonia nitrogen, total phosphorus, total nitrogen, etc. The characteristic pollutant indicators are determined according to the characteristics of the pollution sources upstream of the water source, and may include heavy metals, organic pollutants, pesticide residues, etc. In an embodiment, eight fixed monitoring stations and three mobile monitoring points are set up in the water source. The fixed stations are monitored once a day, and the monitoring frequency of the mobile monitoring points is dynamically adjusted according to the risk assessment results.

[0033] The behavior safety raw data includes video image data collected by the video monitoring system around the water source, infrared monitoring data, vehicle information collected by the vehicle recognition system, and personnel access information recorded by the access control system, etc. In an embodiment, thirty high-definition cameras are set up around the primary protection zone of the water source to achieve 24-hour all-weather monitoring, with a coverage rate of more than 95%. Vehicle recognition systems are set up at the entrances of major roads to automatically identify vehicle types and license plate information.

[0034] The facility safety raw data includes deformation monitoring data, seepage monitoring data, seepage pressure monitoring data, stress and strain monitoring data collected by the dam safety monitoring system, and rainfall data and water level data collected by the hydrological and meteorological monitoring system. Dam deformation monitoring is carried out through monitoring points arranged on the dam body and dam foundation, and the monitoring content includes horizontal displacement, vertical displacement and inclination deformation. Seepage monitoring includes seepage quantity monitoring and seepage pressure monitoring. Seepage quantity is measured by a water measuring weir or a catch basin, and seepage pressure is measured by a seepage pressure gauge buried in the dam body and dam foundation. In an embodiment, forty-five deformation monitoring points, thirty-two seepage pressure monitoring points and eight seepage flow monitoring points are arranged in the dam, and the monitoring data is transmitted in real time to the monitoring center through an automatic collection system.

[0035] The management safety raw data includes water source management system files, patrol system requirements, patrol record data, emergency plan files, emergency drill records, training records, etc. The patrol record data details the patrol personnel, patrol time, patrol route, discovered problems and disposal. In an embodiment, a hierarchical patrol system is implemented in the water source. The primary protection zone is patrolled once a day, and the secondary protection zone is patrolled twice a week. The patrol personnel upload patrol records and on-site photos in real time through a handheld terminal.

[0036] The data acquisition module 1 pre-processes the collected raw data, including data cleaning, outlier removal, missing value supplementation, etc., to ensure that the data quality meets the requirements of subsequent evaluation and analysis. The data pre-processing adopts the 3σ criterion for outlier identification. For the identified outliers, it is determined whether they are real outliers or data errors through manual verification. For missing values, linear interpolation or spatial interpolation based on adjacent stations is used for supplementation.

[0037] Referring to Figure 2 , the water quantity safety index calculation module 2 determines the water quantity safety grade division standard based on the water quantity safety raw data through the rating method, and further generates the water quantity safety index.

[0038] The water quantity safety index calculation module 2 first constructs a water balance model, which comprehensively considers various water quantity factors such as reservoir inflow, water consumption and loss. The basic equation of water balance is: change of storage volume is equal to inflow minus outflow minus evaporation and seepage loss. Based on historical hydrological data and water supply and consumption records, the water balance model is used to simulate the water supply guarantee capacity of the reservoir under different inflow conditions.

[0039] In one embodiment, the hydrological data of typical wet years, normal years and dry years in the past 10 years are selected to simulate and calculate the water supply capacity of the reservoir under different guarantee rates. When the water supply guarantee rate is 95%, the reservoir can meet the design water supply demand, which is defined as water quantity safety; when the water supply guarantee rate is lower than 85%, the reservoir may not be able to meet the basic water supply demand, which is defined as water quantity insecurity; and between 85% and 95% is the basic water quantity safety.

[0040] The water quantity safety index calculation module 2 also considers the influence of the reservoir capacity curve characteristics on water quantity safety. By analyzing the reservoir capacity curve, the characteristic water levels of the reservoir such as dead water level, beneficial water level, design flood level, etc. and the corresponding reservoir capacity are identified. When the real-time water level is higher than the beneficial water level, the reservoir has sufficient available water quantity, and the water quantity safety is high; when the real-time water level is between the beneficial water level and the dead water level, the available water quantity is limited, and the dispatching management needs to be strengthened; when the real-time water level is close to the dead water level, the water quantity safety is seriously threatened.

[0041] Based on the water balance simulation results and the reservoir capacity curve analysis, the water quantity safety index calculation module 2 determines the water quantity safety grade division standard. In this embodiment, the water quantity safety index adopts the percentage system, which is divided into 5 grades: excellent (85-100 points) indicates that the water supply guarantee rate is higher than 98% and the real-time water level is higher than 90% of the beneficial water level; good (70-85 points) indicates that the water supply guarantee rate is between 95%-98%; qualified (55-70 points) indicates that the water supply guarantee rate is between 90%-95%; basically qualified (40-55 points) indicates that the water supply guarantee rate is between 85%-90%; unqualified (0-40 points) indicates that the water supply guarantee rate is lower than 85% or the real-time water level is close to the dead water level.

[0042] The water quantity safety index is calculated by considering both the current water quantity condition and the future water supply risk. The current water quantity condition is evaluated based on the real-time water level and the available water supply, and the future water supply risk is evaluated based on the weather forecast and the water demand prediction. In one embodiment, the seasonal autoregressive moving average model is used to predict the rainfall in the next month, and combined with the growth trend of water demand, the future water balance condition is predicted. When the prediction shows that there may be a water supply shortage in the future, the current water quantity safety index is appropriately reduced to remind the management department to take measures in advance.

[0043] The water quantity safety index output by the water quantity safety index calculation module 2 not only includes the comprehensive score, but also includes the scores of the sub-indicators, such as the water supply guarantee rate score, the water level condition score, and the supply-demand balance score, so as to facilitate the identification of the main factors affecting the water quantity safety.

[0044] Referring to Figure 3 , the water quality safety index calculation module 3 generates the water quality safety evaluation criteria based on the water quality safety raw data, and then generates the water quality safety index through the intelligent weight determination method.

[0045] The core innovation of the water quality safety index calculation module 3 is the use of the intelligent weight determination method, which combines objective weighting and machine learning prediction to achieve dynamic adjustment of the weights. The specific implementation steps are as follows:

[0046] First, according to the monitoring data of each water quality monitoring indicator, the entropy weight method is used to calculate the objective weight. The basic idea of the entropy weight method is to determine the weight according to the variation degree of the numerical value of each indicator, and the greater the variation degree, the greater the weight of the indicator. For the jth water quality monitoring indicator, the objective weight is calculated using the information entropy theory. The monitoring data is normalized and the information entropy value of each indicator is calculated, and then the weight is determined according to the information entropy value. In this embodiment, 15 key water quality monitoring indicators are selected, including pH, dissolved oxygen, permanganate index, chemical oxygen demand, biochemical oxygen demand, ammonia nitrogen, total phosphorus, total nitrogen, fluoride, cyanide, arsenic, mercury, lead, hexavalent chromium, and volatile phenol.

[0047] Second, based on the historical water quality data and the water supply safety event records, a machine learning model is constructed to predict the influence degree of each water quality indicator on the water supply safety. In one embodiment, the random forest algorithm is used to establish the correlation model between the water quality indicators and the water supply safety events. The input features of the model are the concentration values of each water quality indicator, and the output is the occurrence probability of the water supply safety event. Through model training, the feature importance score of each water quality indicator can be obtained, which reflects the influence degree of each indicator on the water supply safety.

[0048] Thirdly, the comprehensive weight of each water quality index is determined by combining the objective weight and the feature importance score using a combined weighting method. The combined weighting method uses a multiplication synthesis method to multiply the objective weight and the feature importance score and normalize the result. This method takes into account both the variability of the data itself and the actual impact of the index on water supply safety, making the weight determination more scientific and reasonable.

[0049] The water quality safety index calculation module 3 also takes into account the key water supply purposes of the water source. For water sources that mainly supply drinking water, the weight of indicators related to the drinking water health standard is increased; for water sources that also supply industrial water, the water quality requirements for industrial water are appropriately considered; for water sources that bear the function of ecological water replenishment, ecological water quality indicators are focused on. In this embodiment, the target reservoir mainly bears the task of supplying urban drinking water, so the indicators specified in the "Drinking Water Health Standard" (GB 5749-2022) are given higher weights in the weight determination.

[0050] Based on the determined comprehensive weight, the water quality safety index calculation module 3 calculates the water quality safety index using a combination of the single-factor index method and the comprehensive index method. The single-factor index method evaluates each monitoring index. When the index concentration exceeds the standard limit, the single-factor index is greater than 1, indicating that the index has an exceeding risk. The comprehensive index method is to weight and sum all the single-factor indexes of the indicators according to the comprehensive weight to obtain the comprehensive water quality safety index.

[0051] In one embodiment, the calculation of the water quality safety index also introduces a time weighting factor. Higher weights are given to recent monitoring data, and lower weights are given to earlier historical data, to better reflect the latest trends in water quality. An exponential decay function is used as the time weighting factor, and the decay coefficient is determined according to the time scale of water quality changes. A larger decay coefficient is used for indicators that change quickly, and a smaller decay coefficient is used for indicators that change slowly.

[0052] The water quality safety index output by the water quality safety index calculation module 3 is divided into five levels: excellent (85-100 points) indicates that all monitoring indicators meet the standard and have a large margin; good (70-85 points) indicates that all indicators meet the standard; qualified (55-70 points) indicates that individual indicators are close to the standard limit; basically qualified (40-55 points) indicates that individual indicators occasionally appear slightly over standard; unqualified (0-40 points) indicates that key indicators are seriously over standard or multiple indicators are over standard at the same time.

[0053] Referring to Figure 4 , the behavior safety index calculation module 4 generates behavior recognition results based on the behavior safety raw data through artificial intelligence recognition methods, and further generates a behavior safety index.

[0054] The core innovation of the behavior safety index calculation module 4 lies in the adoption of deep learning target detection technology and behavior recognition technology, realizing automatic recognition and intelligent evaluation of illegal behaviors around the water source.

[0055] The video preprocessing unit first preprocesses the collected video monitoring data. The preprocessing includes video frame extraction, image enhancement, size normalization, etc. In this embodiment, 2 frames of images per second are extracted from the video stream for analysis, which ensures the real-time nature of the analysis and avoids excessive consumption of computing resources. Image enhancement operations include brightness adjustment, contrast enhancement, noise reduction processing, etc., to improve the accuracy of subsequent target detection. Size normalization uniformly adjusts the image to 640x480 pixels, meeting the input requirements of the target detection model.

[0056] The target detection unit uses an improved YOLO deep learning model to detect targets in video frames. The YOLO model is a single-stage target detection algorithm with the advantages of fast detection speed and high accuracy, suitable for real-time video analysis scenarios. In this embodiment, the YOLO model is improved and trained to accurately identify personnel, vehicles, ships, buildings, and other target objects around the water source. The model training uses more than 10,000 labeled images of water source scenes, and the trained model has an average detection accuracy of 92.5% on the test set.

[0057] The illegal behavior recognition unit further identifies whether there is illegal behavior based on the detected target objects. The illegal behavior recognition adopts a combination of rule-based and learning-based methods. The rule-based method sets judgment rules for clearly defined illegal behaviors, such as: the appearance of unauthorized personnel or vehicles in the primary protection zone is judged as illegal entry; the appearance of swimming, fishing, etc. in the protection zone is judged as illegal activity; the appearance of new buildings or structures in the protection zone is judged as illegal construction. The learning-based method trains a behavior recognition model to recognize more complex illegal behavior patterns. In this embodiment, a time convolution network is used to analyze video sequences to recognize the behavior trajectory and activity pattern of personnel and determine whether there is illegal behavior.

[0058] The four disorder identification unit is specially designed for identifying the "four disorders" of rivers and lakes, including disorderly occupation, disorderly mining, disorderly stacking, and disorderly construction. The disorderly occupation identification analyzes the land use changes within the protected area to identify whether there is illegal occupation of water areas or shorelines. The disorderly mining identification detects whether there are sand mining ships, excavators, and other equipment, as well as whether there are changes in the shape of the river, to identify illegal sand mining behavior. The disorderly stacking identification detects whether there are slag, garbage, and other stacking materials within the protected area to identify disorderly stacking behavior. The disorderly construction identification compares historical images and current images to detect whether there are new buildings or structures to identify illegal construction behavior. In this embodiment, the four disorder identification unit combines remote sensing image analysis and video monitoring analysis, with an identification accuracy of 89.3%.

[0059] The hazardous chemical vehicle identification unit identifies hazardous chemical transport vehicles through a vehicle identification system. The vehicle identification system includes license plate recognition and vehicle type identification functions. The license plate recognition uses deep learning character recognition technology to accurately identify license plate numbers with an accuracy of 99.5%. The vehicle type identification analyzes the appearance characteristics of the vehicle to identify whether it is a hazardous chemical transport vehicle, with features including vehicle body identification, vehicle shape, and vehicle body color. For identified hazardous chemical vehicles, the system automatically queries their driving route and destination to determine whether there are violations of entering protected areas or illegal parking.

[0060] The behavior safety index calculation module 4 calculates the behavior safety index based on the number, type, and severity of the identified illegal behaviors. In this embodiment, a violation scoring table is established, with different types of illegal behaviors assigned different deduction values. For example, unauthorized personnel entering a first-class protected area deducts 5 points, hazardous chemical vehicles violating regulations deducts 10 points, "four disorder" problems deducts 8 points, and illegal construction deducts 15 points. The behavior safety base score is 100 points, and the score is deducted based on the identified illegal behaviors. The behavior safety index is obtained by accumulating the deductions. When the behavior safety index is less than 60 points, it is considered unqualified and immediate corrective measures are required.

[0061] The behavior safety index calculation module 4 also has a learning optimization function. By accumulating illegal behavior samples and manual review results, the target detection model and behavior recognition model are continuously optimized to improve the accuracy of identification. At the same time, the identification threshold is adjusted according to seasonal changes, weather conditions, and other factors to reduce false positives and false negatives.

[0062] The management safety index calculation module 5 generates the management safety index based on the management safety raw data through the compliance rate calculation method.

[0063] The management safety index calculation module 5 first identifies the various system requirements for the water source management, including the patrol system, the emergency management system, the training and education system, the archive management system, etc. For each system requirement, a specific standard is set. For example, the patrol system requires that the first-class protected area be patrolled no less than once a day, the second-class protected area be patrolled no less than twice a week, the patrol record be complete and accurate, and the problems found be disposed of in time and recorded; the emergency management system requires that an emergency drill be carried out at least once a year, the emergency plan be revised and updated regularly, and the emergency supplies be fully equipped and checked regularly; the training and education system requires that the management personnel receive training no less than 40 hours a year, and the new employees receive pre-service training and pass the examination.

[0064] The management safety index calculation module 5 compares the actual execution with the standard, and calculates the compliance rate of each system requirement. The compliance rate is equal to the number of requirements actually met divided by the total number of requirements evaluated. In this embodiment, 25 management system requirements are set, and whether each requirement is met is checked by checking the patrol record, the emergency drill record, the training record, the archive data, etc. For example, the patrol record of a month is checked to determine whether the actual patrol times meet the requirements, the patrol record is complete, the problems found are disposed of in time, and whether the patrol system meets the requirements is determined according to the checking result.

[0065] The management safety index is a percentage system, and the percentage of the compliance rate is directly taken as the management safety index. For example, if 22 of the 25 management system requirements are met, the compliance rate is 88%, and the management safety index is 88 points. The management safety index is divided into five levels: excellent (90-100 points) indicates that the management system is sound and the execution is in place; good (80-90 points) indicates that the management system is basically sound and the execution is good; qualified (70-80 points) indicates that the management system is relatively perfect and the execution is basically in place; basically qualified (60-70 points) indicates that the management system is not perfect or the execution is not in place; unqualified (0-60 points) indicates that the management system is missing or the execution is seriously out of place.

[0066] The management safety index calculation module 5 also identifies the weak links of the management safety. For the system requirements that are not met, the reasons are analyzed and improvement suggestions are put forward. In this embodiment, it is found that the patrol system is not executed in place, which is the main factor affecting the management safety index. It is found through analysis that the lack of patrol personnel and the unreasonable patrol route are the main reasons, and it is suggested that the patrol personnel be increased and the patrol route be optimized.

[0067] Referring to Figure 5 , the facility safety index calculation module 6 generates the facility safety index based on the facility safety raw data through the abnormal alarm judgment method and the closed-loop disposal evaluation method.

[0068] The core innovation of the facility safety index calculation module 6 lies in the adoption of an abnormal alarm discrimination method based on time series analysis and a closed-loop treatment evaluation method, thereby realizing dynamic monitoring and whole-process management of the facility safety state.

[0069] The time series data establishment unit first arranges the dam safety monitoring data in time series, and establishes time series data sequences of each monitoring index. The time series data sequences record the change history of the monitoring index over time. In this embodiment, eight key monitoring indexes are selected to establish time series sequences, including dam crest horizontal displacement, dam body maximum horizontal displacement, dam foundation vertical displacement, dam body maximum seepage pressure, total seepage flow, reservoir water level, rainfall and air temperature. The sampling interval of the time series data is determined according to the characteristics of the monitoring index, and the deformation monitoring data is collected once a day, the seepage monitoring data is collected once every 12 hours, and the reservoir water level and weather data are collected once every hour.

[0070] The normal range modeling unit establishes a normal change range model of each monitoring index based on historical monitoring data. The normal change range model describes the change law and fluctuation range of the monitoring index under normal conditions. In this embodiment, a statistical analysis method is used to establish the normal change range model, and the specific steps are as follows:

[0071] Firstly, collect the historical monitoring data in the past five years, eliminate abnormal data and missing data, and form an effective data set.

[0072] Secondly, analyze the correlation between each monitoring index and environmental factors such as reservoir water level, rainfall and air temperature, and establish a multiple regression model. For example, the dam body horizontal displacement is mainly affected by the reservoir water level, and the regression model uses a linear relationship to describe the relationship between displacement and water level; the seepage flow is jointly affected by the reservoir water level and rainfall, and the regression model uses a multiple linear relationship.

[0073] Thirdly, calculate the residual sequence of the regression model, and the residual reflects the deviation between the monitoring value and the model predicted value. Statistical analysis is performed on the residual sequence, and the mean and standard deviation of the residual are calculated. The normal change range is defined as the model predicted value plus or minus three times the standard deviation, which covers 99.7% of the normal observation values.

[0074] Finally, establish a hierarchical alarm threshold. When the monitoring value exceeds the normal range but does not exceed the model predicted value plus or minus five times the standard deviation, a yellow warning is issued; when the monitoring value exceeds the model predicted value plus or minus five times the standard deviation, an orange warning is issued; and when the monitoring value continuously exceeds the normal range or exceeds the model predicted value plus or minus seven times the standard deviation, a red warning is issued.

[0075] The real-time monitoring unit analyzes the time series monitoring data in real time to determine whether there is an abnormality. The real-time monitoring unit compares the new monitoring data with the normal change range model as soon as the new monitoring data is received. When the monitoring value exceeds the normal range, an abnormal alarm signal is generated. The abnormal alarm signal includes information such as the abnormal index name, the monitoring point position, the abnormal value size, the exceeding degree, the occurrence time, and the like.

[0076] In one embodiment, the real-time monitoring unit also uses a trend analysis method to determine whether the change trend of the monitoring index is abnormal. Even if the monitoring value has not exceeded the normal range, if the change rate is abnormal or the change trend does not conform to the historical law, a pre-warning signal is issued. For example, the displacement of the dam body usually presents a slow change with the rise and fall of the reservoir water level. If the displacement suddenly accelerates, even if it has not exceeded the normal range, it may indicate a safety hazard and needs to be paid attention to.

[0077] The closed-loop treatment evaluation unit tracks and evaluates the entire process of the treatment of the abnormal alarm. The closed-loop treatment evaluation includes response timeliness evaluation and treatment effectiveness evaluation.

[0078] The response timeliness evaluation is evaluated according to the time interval from the issuance of the abnormal alarm to the start of the emergency response. In this embodiment, it is stipulated that the yellow pre-warning should start the response within 2 hours, the orange pre-warning should start the response within 1 hour, and the red pre-warning should start the response immediately. If the actual response time meets the requirements, the response timeliness evaluation is qualified; if the stipulated time is exceeded, the response timeliness evaluation is unqualified.

[0079] The treatment effectiveness evaluation is evaluated according to whether the abnormal state is eliminated after the implementation of the treatment measures. The closed-loop treatment evaluation unit continuously tracks the monitoring data of the abnormal index to determine whether the monitoring value returns to the normal range after the treatment and whether the abnormal change trend is suppressed. If the treatment measures effectively eliminate the abnormal state, the treatment effectiveness evaluation is effective; if the abnormal state is not eliminated or worsens, the treatment effectiveness evaluation is ineffective, and further measures need to be taken.

[0080] The facility safety index is calculated according to the number of abnormal alarms, the severity of the abnormality, and the closed-loop treatment. The facility safety base score is 100 points, and the score is deducted according to the abnormal alarm. The yellow pre-warning is deducted by 2 points each time, the orange pre-warning is deducted by 5 points each time, and the red pre-warning is deducted by 10 points each time. If the response to the abnormal alarm is not timely, 3 points are additionally deducted; if the treatment is ineffective, 5 points are additionally deducted. In this embodiment, 3 yellow pre-warnings and 1 orange pre-warning occur in a month, all the pre-warnings are responded in time and the treatment is effective, and therefore the facility safety index is 100-3x2-1x5=89 points, which is evaluated as a good level.

[0081] The facility safety index is divided into five levels: excellent (90-100 points) indicates that the facility is running normally, with no abnormal alarms or only a small number of yellow warnings and timely and effective disposal; good (80-90 points) indicates that the facility is running normally, with a small number of warnings and timely and effective disposal; qualified (70-80 points) indicates that the facility has certain hidden dangers, with a large number of warnings but effective disposal; basically qualified (60-70 points) indicates that the facility has more hidden dangers, with some warnings not disposed of in time or not effectively; unqualified (0-60 points) indicates that the facility has major safety hazards, with red warnings not disposed of effectively or cumulative points exceeding 40 points.

[0082] Referring to Figure 6 , the comprehensive index calculation and risk warning module 7 generates a comprehensive index of water supply safety of the reservoir-type water source based on the five-dimensional safety index, combined with the comprehensive evaluation weight and the one-vote veto rule, and generates dynamic warning information and disposal suggestions according to the comprehensive index.

[0083] The determination of the comprehensive evaluation weight comprehensively considers the importance of each dimension safety index to the water supply safety. In this embodiment, the weight of each dimension is determined by combining expert scoring method and analytic hierarchy process. Ten experts in the fields of water source management, water conservancy engineering, and environmental monitoring are invited to score the relative importance of the five dimensions, a judgment matrix is constructed, and the weight of each dimension is calculated by analytic hierarchy process. The calculated weight is: water quality safety weight 0.30, water quantity safety weight 0.25, facility safety weight 0.20, behavior safety weight 0.15, and management safety weight 0.10. The weight distribution reflects that water quality and water quantity are the basis and core of water supply safety, facilities are the material basis for guaranteeing water supply safety, and behavior supervision and standardized management are important means to maintain water supply safety.

[0084] The one-vote veto rule is set for key risk indicators. When the safety index of a certain dimension is lower than the preset veto threshold, the comprehensive index is directly determined as unqualified regardless of the index of other dimensions. In this embodiment, the set one-vote veto rule includes: one-vote veto when the water quality safety index is lower than 40 points, because serious non-compliance of water quality directly endangers drinking water safety; one-vote veto when the water quantity safety index is lower than 30 points, because serious water shortage cannot guarantee basic water supply demand; one-vote veto when the facility safety index is lower than 50 points, because the facility has major safety hazards that may lead to disastrous consequences such as dam collapse. No one-vote veto is set for behavior safety index and management safety index, but when both are lower than 50 points, a special attention mechanism is triggered, requiring rectification within a limited period.

[0085] The comprehensive index calculation unit first determines whether the one-vote veto rule is triggered. If the one-vote veto is triggered, the comprehensive index is directly set to 0 points, the qualified level is evaluated, and a red warning is immediately issued. If the one-vote veto is not triggered, the comprehensive index is calculated by the weighted summation method.

[0086] The calculation formula of the comprehensive index is:

[0087]

[0088] wherein, S is the comprehensive index of water supply safety of the reservoir water source, S is the weight of water quality safety, S is the weight of water quantity safety, S is the weight of facility safety, S is the weight of behavior safety, and S is the weight of management safety, and the values are 0.30, 0.25, 0.20, 0.15, and 0.10 respectively, S is the water quality safety index, S is the water quantity safety index, S is the facility safety index, S is the behavior safety index, and S is the management safety index. All indexes are in percentage system, and the value range is 0 to 100.

[0089] In one embodiment, the five-dimensional safety indexes of a reservoir water source are as follows: the water quality safety index is 85 points, the water quantity safety index is 78 points, the facility safety index is 89 points, the behavior safety index is 72 points, and the management safety index is 88 points. It is judged that the one-vote veto rule is not triggered. The comprehensive index is calculated as:

[0090] The comprehensive index is divided into five levels: excellent (80-100 points) corresponds to blue identification, indicating that the water supply safety condition is excellent, and all indexes meet the requirements and have a larger margin; good (60-80 points) corresponds to green identification, indicating that the water supply safety condition is good, and individual indexes may have slight deficiencies but do not affect the overall safety; qualified (40-60 points) corresponds to yellow identification, indicating that the water supply safety is basically guaranteed, but there are certain risk hidden dangers that need attention; basically qualified (20-40 points) corresponds to orange identification, indicating that the water supply safety has a greater risk, and measures need to be taken to strengthen control; unqualified (0-20 points) corresponds to red identification, indicating that the water supply safety is facing serious threats, and immediate measures must be taken to respond.

[0091] ​​​​​​​​​​​​​The dynamic early warning unit generates dynamic early warning information based on the real-time changes and prediction trends of the comprehensive index. The dynamic early warning unit first establishes a time series database of the comprehensive index, recording the comprehensive index of each period in history. Then, a time series prediction model is used to predict the future comprehensive index. In this embodiment, a long short-term memory network is used to predict the trend of the comprehensive index in the next month. The long short-term memory network is a special recurrent neural network that can learn the long-term dependencies of time series data and is suitable for time series prediction tasks. The input of the model is the sequence of comprehensive indexes in the past 6 months and related meteorological data, water supply data and other influencing factors, and the output is the predicted value of the comprehensive index in the next month.

[0092] When the prediction shows that the comprehensive index will decrease below the early warning threshold, a risk increase early warning is generated. The early warning threshold is determined according to the safety level limit, and a yellow early warning is issued when the predicted index decreases to below 60 points, an orange early warning is issued when it decreases to below 40 points, and a red early warning is issued when it decreases to below 20 points. The risk increase early warning prompts decision makers to take preventive measures in advance to avoid further deterioration of the risk.

[0093] When the real-time comprehensive index is below the early warning threshold, a current risk early warning is generated. The current risk early warning determines the early warning level according to the interval of the real-time comprehensive index, and the qualified level corresponds to a yellow early warning, the basically qualified level corresponds to an orange early warning, and the unqualified level corresponds to a red early warning.

[0094] The dynamic early warning information is presented to decision makers and management personnel through a visual interface. The visual interface uses a dashboard form to intuitively display the current comprehensive index, five-dimensional safety index, early warning level, risk change trend and other information. The early warning level is indicated by warning lights of different colors, with blue indicating excellent, green indicating good, yellow indicating qualified and needing attention, orange indicating basically qualified and needing vigilance, and red indicating unqualified and needing immediate disposal. When an early warning occurs, the system automatically sends early warning SMS and emails to the relevant responsible persons to ensure that the early warning information is timely conveyed.

[0095] Reference Figure 7 The obstacle factor identification and disposal suggestion generation module 8 identifies the main obstacle factors affecting the comprehensive index based on the five-dimensional safety index and generates targeted disposal suggestions.

[0096] The obstacle factor identification unit identifies the obstacle effect of each dimension and each index on the comprehensive index using the obstacle degree model. The obstacle degree reflects the degree of hindering the achievement of the target, and the greater the obstacle degree, the more the factor is a main limiting factor affecting the achievement of the target.

[0097] The calculation formula of the obstacle degree is as follows:

[0098] ,

[0099] wherein, is the obstacle degree of the i-th factor, is the weight of the i-th factor, is the index score of the i-th factor, is the index score of the i-th factor, is the index score of the i-th factor, is the total number of factors, and is 5 (corresponding to five dimensions). The obstacle degree ranges from 0% to 100%, and the greater the obstacle degree, the stronger the constraint of the factor on the comprehensive index. In the foregoing embodiment, the five-dimensional safety index is respectively: water quality safety index 85 points (weight 0.30), water quantity safety index 78 points (weight 0.25), facility safety index 89 points (weight 0.20), behavior safety index 72 points (weight 0.15), and management safety index 88 points (weight 0.10). The obstacle degrees of each dimension are calculated as follows:

[0100] Water quality safety obstacle degree:

[0101] ; Water quantity safety obstacle degree:

[0102] ; Facility safety obstacle degree:

[0103] ; Behavior safety obstacle degree:

[0104] ; Management safety obstacle degree:

[0105] ; According to the obstacle degree, the water quantity safety is the most important obstacle factor (obstacle degree 31.7%), followed by the water quality safety (obstacle degree 25.9%) and the behavior safety (obstacle degree 24.2%), and the obstacle degrees of the facility safety and the management safety are relatively small.

[0106] The obstacle factor identification unit further analyzes the obstacle degrees of specific indicators in each dimension to identify specific constraint factors. For example, for the water quantity safety dimension, the obstacle degrees of the water supply guarantee rate, water level condition, and supply-demand balance are analyzed to identify whether the water quantity safety problem is caused by insufficient water supply capacity or increasing water demand.

[0107]

[0108] ​The treatment suggestion generation unit matches corresponding treatment suggestions from a treatment strategy knowledge base based on the identified obstacle factors. The treatment strategy knowledge base is a pre-constructed expert knowledge base that contains treatment strategies and measures for various obstacle factors. For example, for water quantity safety obstacles, treatment suggestions include: strengthening water resource scheduling, optimizing water supply plans; implementing water-saving measures, controlling water demand growth; conducting artificial rain enhancement operations to increase reservoir inflow; starting emergency backup water sources to disperse water supply risks, etc. For water quality safety obstacles, treatment suggestions include: strengthening upstream pollution source control to reduce pollutant inflow into the reservoir; increasing water quality monitoring frequency to timely detect water quality abnormalities; optimizing water treatment processes to improve purification effects; implementing reservoir ecological restoration to enhance water self-purification capacity, etc. For behavior safety obstacles, treatment suggestions include: strengthening patrol supervision to timely detect and stop illegal activities; carrying out special rectification actions to clean up illegal facilities and activities; improving video monitoring systems to improve intelligent identification capabilities; strengthening publicity and education to improve public protection awareness, etc.

[0109] The treatment suggestion generation unit also prioritizes treatment suggestions according to the urgency and impact of obstacle factors. Obstacle factors with high urgency and large impact are prioritized for treatment; obstacle factors with low urgency or small impact are placed at the end and can be appropriately delayed. In this embodiment, water quantity safety and water quality safety have high obstacle degrees and are directly related to water supply safety, so they are placed at the top of the treatment suggestion priority.

[0110] The treatment suggestion optimization unit optimizes and adjusts treatment suggestions based on historical treatment records and treatment effect data. The treatment suggestion optimization unit establishes a treatment effect evaluation database that records treatment measures and treatment effects for obstacle factors. By analyzing historical data, it identifies which treatment measures have significant effects and which treatment measures have poor effects, thereby optimizing the content and priority of treatment suggestions. For example, through historical data analysis, it is found that for water quantity safety issues, optimizing water supply plans and implementing water-saving measures have good effects, while artificial rain enhancement is affected by weather conditions and is not stable enough, so the recommendation priority of the first two measures is increased and the recommendation priority of artificial rain enhancement is decreased.

[0111] The treatment suggestion report output by the treatment suggestion generation module 8 includes obstacle factor list, obstacle degree ranking, specific treatment measures, expected effects, implementation subject, and completion time limit, etc., providing detailed decision-making reference for decision-makers.

[0112] The method of the present application also includes an adaptive dynamic evaluation module 9 that dynamically adjusts evaluation parameters and evaluation strategies based on external environmental changes and system operation feedback, achieving adaptive optimization of the evaluation method.

[0113] The climate change adaptation unit dynamically adjusts the evaluation weight of the water quantity safety index based on the future rainfall pattern and evaporation change predicted by a climate change model. The influence of climate change on the water source mainly reflects in the change of rainfall and evaporation, which further affects the reservoir inflow and water surface evaporation loss. The climate change adaptation unit adopts a regional climate model, combined with global climate change scenarios, to predict the rainfall and evaporation change trend in the next 10 to 30 years. When the prediction shows that the future rainfall will significantly decrease, it indicates that the water quantity safety is facing greater challenges, and the weight of water quantity safety in the comprehensive evaluation should be appropriately increased; when the prediction shows that the rainfall change is not large or increases, the weight of water quantity safety remains unchanged.

[0114] The pollution risk prediction unit predicts the spatio-temporal distribution of water quality safety risk based on the upstream pollution source distribution information and hydrological connectivity data through a pollutant migration simulation model. The pollution risk prediction unit first investigates and identifies the pollution sources upstream of the water source, including industrial enterprises, livestock and poultry breeding, agricultural non-point source, urban life and other pollution source types, and obtains the information such as the location, scale and emission characteristics of the pollution sources. Then, based on the hydrological model and the pollutant migration model, the influence of different pollution sources on the water quality of the water source under different hydrological conditions is simulated. In this embodiment, the SWAT model is used for hydrological and water quality simulation at the watershed scale, which can comprehensively consider the processes of rainfall runoff, soil erosion, nutrient migration and transformation, and predict the generation, migration and entry process of pollutants.

[0115] The monitoring optimization unit dynamically adjusts the water quality monitoring frequency and monitoring point based on the predicted spatio-temporal distribution of water quality safety risk. When the pollution risk of a certain area or a certain period is high, the monitoring frequency of the area or the period is increased, or temporary monitoring points are added in the high-risk area to strengthen the monitoring and early warning. When the pollution risk is low, the monitoring frequency can be appropriately reduced to optimize the allocation of monitoring resources. In this embodiment, through pollution risk prediction, it is identified that agricultural non-point source pollution has a greater impact on the water source during the rainy season, so the monitoring frequency of the upstream agricultural area is increased from once a week to once every 3 days, and 2 temporary monitoring points are added, effectively strengthening the monitoring and early warning of agricultural non-point source pollution.

[0116] The evaluation method updating unit regularly updates the parameters and models of the evaluation method according to accumulation of monitoring data and summary of evaluation experience. The evaluation method updating includes adjustment of the evaluation index system, correction of the weight coefficients, optimization of the model parameters, and the like. In the embodiment, a comprehensive review and update of the evaluation method is performed once a year, according to the operation in the past year, the applicability and accuracy of the evaluation method are analyzed, aspects that need to be improved are identified, and an update scheme is proposed. For example, it is found through one year of operation that the original water quantity safety evaluation model does not fully consider the seasonal variation of evaporation loss, resulting in a high water quantity safety index in summer, so an evaporation correction coefficient is introduced in the model update, the calculation of evaporation loss is dynamically adjusted according to the season, and the accuracy of the evaluation is improved.

[0117] The introduction of the adaptive dynamic evaluation module 9 enables the evaluation method to automatically adjust and optimize according to changes in the external environment and internal operation, maintains the advancement and applicability of the evaluation method, and realizes continuous improvement and intelligent evolution of the evaluation method.

[0118] To verify the effectiveness of the method, a large reservoir water source is selected for a one-year application test. The water source is the main drinking water source for the city, with a design water supply capacity of 1 million m³ / d, a service population of 2 million people, a total reservoir capacity of 250 million m³, and a catchment area of 1,200 km².

[0119] During the application test, a complete five-dimensional safety evaluation system is established according to the method, and automatic monitoring equipment and intelligent analysis systems are deployed. The water quality monitoring stations are increased from the original 5 to 8, with 3 new monitoring points added at key sections; the video monitoring cameras are increased from the original 18 to 30, achieving full coverage of the primary protection zone; the dam safety monitoring system is upgraded, with the addition of automatic data acquisition and transmission functions.

[0120] The application test results show that:

[0121] First, the comprehensive evaluation is significantly improved. Compared with the original single evaluation method that only focuses on water quality, the five-dimensional evaluation system established by the method can comprehensively reflect the safety status of the water source. During the test, 3 water quantity safety risks, 2 water quality safety risks, 15 behavior safety risks, 1 facility safety risk, and several management safety risks are identified, while the original method can only identify water quality safety risks. Through multi-dimensional evaluation, the problem of missing other risks due to focusing on a single dimension is avoided.

[0122] Second, the intelligent level is greatly improved. By introducing machine learning, deep learning and other artificial intelligence technologies, intelligent determination of water quality index weight, automatic identification of illegal behavior, intelligent detection of facility abnormalities and other functions are realized. During the test, the intelligent identification system automatically identified 92 illegal behaviors, with an accuracy rate of 89.1%, which improved the discovery efficiency and identification accuracy compared with manual patrol. The water quality anomaly detection system warned of an upstream pollution event 3 days in advance, gaining valuable time for emergency disposal.

[0123] Third, the dynamic early warning capability is significantly enhanced. The method is based on real-time monitoring data for dynamic evaluation, and predicts the risk change trend through a time series prediction model, realizing early warning. During the test, 5 risk increase warnings were issued, all 1 to 2 weeks before the actual risk occurred, with a warning accuracy rate of 100%. Compared with the original periodic evaluation method, the dynamic early warning capability is significantly enhanced, providing more time for risk prevention and control.

[0124] Fourth, the decision support effect is significant. The method not only provides evaluation results, but also identifies obstacles and generates disposal suggestions, providing strong support for decision-making. During the test, according to the obstacle factor analysis and disposal suggestions, the water source management department focused on strengthening water quantity scheduling management and behavior supervision, and implemented 8 targeted improvement measures, which improved the water quantity safety index from an average of 72 to 81, and the behavior safety index from an average of 68 to 79, significantly improving the overall safety situation.

[0125] Fifth, the economic benefits are good. By accurately identifying risks and optimizing management measures, unnecessary monitoring investment and management costs are reduced. At the same time, through early warning and timely disposal, multiple potential water pollution incidents are avoided, reducing emergency disposal costs and social impact. Preliminary estimates show that after applying the method, the economic benefits of water source management have increased by about 20%, with significant social benefits.

[0126] In summary, the intelligent evaluation method for production, life and ecological water safety risks of reservoir-type water sources provided by the present application effectively solves the problems of single evaluation dimension, low intelligence level and lack of dynamic early warning in the prior art by establishing a five-dimensional safety index system, using intelligent evaluation methods, and realizing dynamic early warning and obstacle factor identification, providing strong technical support for fine management and scientific decision-making of reservoir-type water sources, and having good application value and promotion prospects.

[0127] The above only describes the preferred embodiments of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation using the content of the present application specification, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for intelligent evaluation of safety risks of production and living ecological water of a reservoir water source, characterized in that ,including: obtaining water quantity safety original data, water quality safety original data, behavior safety original data, facility safety original data and management safety original data of the reservoir water source, wherein the water quantity safety original data includes reservoir water supply data and reservoir capacity curve characteristic data, the water quality safety original data includes water quality monitoring station network arrangement data, water quality monitoring index data and key water quality monitoring parameter data, the behavior safety original data includes video monitoring data, illegal behavior identification data and dangerous chemical vehicle monitoring data, the facility safety original data includes dam deformation data, seepage data, seepage pressure data, rainfall data and water quantity monitoring data, and the management safety original data includes patrol requirement data and up-to-standard patrol data; based on a preset water quantity safety evaluation index system, combining the reservoir water supply data and the reservoir capacity curve characteristic data, determining the water quantity safety grade division standard through a rating method, and generating a water quantity safety index according to the water quantity safety grade division standard; based on a preset water quality safety evaluation index system, combining the water quality monitoring station network arrangement data, the water quality monitoring index data and the key water quality monitoring parameter data, generating a water quality safety evaluation standard through an intelligent weight determination method, and generating a water quality safety index according to the water quality safety evaluation standard; based on a preset behavior safety evaluation index system, combining the video monitoring data, the illegal behavior identification data and the dangerous chemical vehicle monitoring data, generating a behavior identification result through an artificial intelligence identification method, and generating a behavior safety index according to the behavior identification result; based on a preset management safety evaluation index system, combining the patrol requirement data and the up-to-standard patrol data, generating a management safety index through an up-to-standard rate calculation method; based on a preset facility safety evaluation index system, combining the dam deformation data, the seepage data, the seepage pressure data, the rainfall data and the water quantity monitoring data, generating a facility safety index through an abnormal alarm discrimination method and a closed-loop disposal evaluation method; based on the water quantity safety index, the water quality safety index, the behavior safety index, the management safety index and the facility safety index, combining a preset comprehensive evaluation weight and a veto rule, generating a reservoir water source water supply safety comprehensive index through a comprehensive index calculation method; determining a safety grade of the reservoir water source according to the comprehensive index, wherein the safety grade includes five grades of excellent, good, qualified, basically qualified and unqualified, and generating dynamic early warning information and disposal suggestions based on the safety grade.

2. The method according to claim 1, characterized in that , the intelligent weight determination method includes: calculating information entropy of each water quality monitoring index according to the water quality monitoring index data; determining objective weights of each water quality monitoring index based on the information entropy; obtaining historical water quality data of the water source and key water supply purpose information; based on the historical water quality data and the key water supply purpose information, predicting the influence degree of each water quality monitoring index on water supply safety through a machine learning model; based on the influence degree and the objective weights, generating a comprehensive weight of each water quality monitoring index through a combination weighting method; generating the water quality safety evaluation standard according to the comprehensive weight.

3. The method according to claim 1, characterized in that , the artificial intelligence identification method includes: preprocessing the video monitoring data to generate a standardized video frame sequence; The standardized video frame sequence is analyzed based on a deep learning target detection model to identify personnel activities, vehicle passing and facility status around the water source; The personnel activities and vehicle passing identified are determined for illegal behavior based on a pre-trained illegal behavior identification model; The surrounding environment of the water source is identified for illegal occupation, illegal mining, illegal stacking and illegal construction based on a four-chaos identification model; The vehicle passing is determined for dangerous chemical vehicle based on a dangerous chemical vehicle identification model; The behavior identification result is generated based on the illegal behavior determination result, the four-chaos identification result and the dangerous chemical vehicle determination result.

4. The method according to claim 1, characterized in that The abnormal alarm identification method comprises: A time series monitoring data sequence is established based on dam deformation data, leakage data, seepage pressure data, rainfall data and water quantity monitoring data; A normal change range model of the monitoring index is established based on historical monitoring data through a time series analysis method; Real-time monitoring is performed on the time series monitoring data sequence based on the normal change range model, and an abnormal alarm signal is generated when the monitoring value exceeds the normal change range; An abnormal level is determined based on the type, duration and severity of the abnormal alarm signal.

5. The method according to claim 4, characterized in that The closed-loop treatment evaluation method comprises: Treatment record data corresponding to the abnormal alarm signal is obtained; It is judged whether the abnormal alarm is responded in time based on the treatment record data; It is judged whether the treatment measures effectively eliminate the abnormal state based on the treatment record data; A closed-loop treatment evaluation result is generated based on the response timeliness and the treatment effectiveness; The facility safety index is adjusted according to the closed-loop treatment evaluation result.

6. The method according to claim 1, characterized in that The comprehensive index calculation method comprises: It is judged whether there is an index triggering the veto rule among the water quantity safety index, the water quality safety index, the behavior safety index, the management safety index and the facility safety index; When there is an index triggering the veto rule, the water supply safety comprehensive index of the reservoir-type water source is directly set to the minimum score corresponding to the unqualified level; When there is no index triggering the veto rule, the water supply safety comprehensive index of the reservoir-type water source is calculated by weighted summation method based on the water quantity safety index, the water quality safety index, the behavior safety index, the management safety index and the facility safety index, combined with the respective comprehensive evaluation weights.

7. The method according to claim 1, characterized in that The determination of the water quantity safety level division standard comprises: The design water supply capacity, the actual water supply quantity and the reservoir capacity curve data of the reservoir are obtained; The available water quantity of the reservoir at different water levels is determined based on the reservoir capacity curve data; A water balance model is established based on historical water supply and consumption data and rainfall data; The water quantity safety risk in future different time periods is predicted based on the water balance model; The critical threshold for water quantity safety level division is determined according to the predicted water quantity safety risk; The water quantity safety level division standard is generated based on the critical threshold. 8.The method of claim 1, wherein the method is characterized by The generation of dynamic early warning information comprises: The historical change trend data of the water supply safety comprehensive index of the reservoir-type water source is obtained; The change trend of the comprehensive index in future time periods is predicted based on the historical change trend data through a time series prediction model; When the predicted change trend of the comprehensive index shows that the comprehensive index will decrease below a preset warning threshold, risk rising warning information is generated; When the real-time comprehensive index is lower than the preset warning threshold, current risk warning information is generated. Determine the warning level based on the severity of the risk rising early warning information or the current risk early warning information; Generate dynamic early warning information with different color features according to the warning level. 9.The method of claim 8, wherein the method is characterized by The generation of the treatment suggestion includes: Identify the main obstacle factors affecting the comprehensive index based on the water quantity safety index, the water quality safety index, the behavior safety index, the management safety index and the facility safety index; Determine the priority treatment obstacle factor based on the contribution degree ranking of the obstacle factor; Generate targeted treatment suggestions based on the preset treatment strategy knowledge base and the priority treatment obstacle factor; Optimize and adjust the treatment suggestions according to the historical treatment records and treatment effect data of the water source. 10.The method of claim 1, wherein the method is characterized by The method further includes: Predict the future rainfall pattern and evaporation change through a climate change model based on the geographical location information and meteorological data of the reservoir type water source; Adjust the evaluation weight of the water quantity safety index based on the predicted rainfall pattern and evaporation change; Predict the spatio-temporal distribution of water quality safety risk through a pollutant migration simulation model based on the upstream pollution source distribution information and hydrological connectivity data; Dynamically adjust the water quality monitoring frequency and monitoring point based on the predicted spatio-temporal distribution of water quality safety risk; Update the water quality safety index based on the adjusted monitoring data to realize the adaptive dynamic evaluation of the safety risk of the reservoir type water source.

Images