Water ecological environment monitoring system based on artificial intelligence

Abstract

The invention is based on an artificial intelligence-based water ecological environment monitoring system. The data detection module detects water quality parameters and hydrological characteristics, enters the data encoding module, performs quality assessment on the detection data from four dimensions of accuracy, completeness, timeliness and consistency, then performs data cleaning, and finally performs encoding: quantization, compression, and encryption processing. The transmission module transmits the encoded data to the monitoring platform using GPS, and configures a buffer, a network performance detector, a locator, and a main control. The buffer is triggered when there is a burst of data volume, uncoordinated access speed, and congestion, so as to reduce the load pressure of the data decoding module of the back-end monitoring platform. The data decoding module receives the encoded data transmitted from the communication network, decodes the detection data and the positioning data, and enters the data summary module. The data summary module constructs a model for analysis, early warning, and tracing of the received data, and intuitively displays the ecological environment status of the water area in the form of a map, thereby realizing intelligent management and control.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence and aquatic ecological environment, and in particular relates to an aquatic ecological environment monitoring system based on artificial intelligence. Background Art

[0002] Due to human production and life activities, substances or energy are directly or indirectly introduced into the river environment. At present, the ecological environment pollution in my country's waters is becoming more and more serious. Strengthening water resource management and ensuring water resource security have become major issues concerning the national economy and people's livelihood.

[0003] Intelligent supervision of water ecological environment uses artificial intelligence technology to analyze and process various data on water quality and hydrological characteristics to achieve intelligent management and optimization. The existing intelligent management uses fixed-point detection instruments and sensors to detect parameters, which are transmitted to the monitoring platform through the configured main control and communication module. The monitoring platform intelligently analyzes, predicts and processes the data. It is inevitable that there is data loss, delay and high repeatability, which leads to poor quality of data received by the monitoring platform, inaccurate analysis and processing results, and data quality assessment and prediction rely heavily on the adaptive optimization strategy of parameters of the set model, which leads to overfitting or underfitting within limited computing resources, and each parameter is predicted and processed independently, making it difficult to provide high-quality solutions. Summary of the invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an artificial intelligence-based water ecological environment supervision system, which effectively solves the problem of low quality of water ecological environment supervision caused by poor data quality, poor model parameter fitting, and independence of various parameters in the prior art.

[0005] The technical solution includes:

[0006] Data detection module: a sampler placed in the water area is used to obtain the water sample to be tested. The sensor configured in the sampler detects the water quality parameters. The sensor outside the sampler placed in the water area detects the hydrological characteristics and enters the data encoding module;

[0007] The data encoding module first evaluates the data quality, then cleans the data, and finally encodes it: quantization, compression, and encryption;

[0008] The transmission module transmits the encoded data via a communication network and is configured with a buffer, a network performance detector, and a locator;

[0009] The data decoding module receives the encoded data transmitted from the communication network, decodes the detection data and positioning data, and enters the data aggregation module;

[0010] The data aggregation module analyzes, warns and traces the received data, and intuitively displays the ecological environment of the water area in the form of a map.

[0011] Preferably, the steps of evaluating data quality and then performing data cleaning are:

[0012] Step 1: Determine the evaluation criteria: including accuracy Z, completeness W, timeliness S, and consistency Y;

[0013] Step 2: Build the evaluation model, P=Z×35%+W×14%+S×26%+Y×25%;

[0014] Step 3, obtain real-time data and historical data;

[0015] Step 4, compare the deviation between the historical data prediction data and the real-time data to obtain the accuracy Z;

[0016] By judging the degree of missingness of the sampled data, the completeness W is obtained;

[0017] By comparing the sampling frequency and the time from sampling to receiving data with the reference time, the timeliness S is obtained;

[0018] The consistency Y is obtained by multiple detections of the same sensor at different time points, or simultaneous detection by different sensors;

[0019] Step 5: Construct a data quality assessment formula and output the data quality assessment value SZ: SZ = Pi ×80%+ Pi -1×12%+ Pi- 2×8%;

[0020] Step 6: Determine whether the data quality assessment value SZ meets the requirements. If yes, perform encoding; if no, perform data cleaning;

[0021] Step 7, firstly, adopt the regression method to interpolate according to the completeness of the data;

[0022] Secondly, identify abnormal data based on accuracy and consistency;

[0023] Finally, for abnormal data, we take timeliness into consideration and take measures to correct or delete outliers.

[0024] Specifically, the above evaluation model also includes regular evaluation, adjustment and optimization of weights, the steps are:

[0025] S1, set the initial weights to 35%, 14%, 26%, and 25% respectively;

[0026] S2, using the hill climbing algorithm to fine-tune the current weight upward and downward, and calculate the data quality assessment value SZ after the weight is modified;

[0027] S3, select the maximum value of data quality assessment value SZ;

[0028] S4, the number of upward and downward fine-tuning quantization is set to 3, the step size is set to 1%, and the total weight is 1.

[0029] Preferably, the data aggregation module analyzes, warns and traces the received data, and intuitively displays the ecological environment of the water area in the form of a map. The specific steps are:

[0030] A1, divide the data into data groups according to type and measurement point;

[0031] A2, extract abnormal features within the data group according to the divided data groups;

[0032] A3, consistency analysis of abnormal feature-related neighboring points, yes, normal data, no, median compensation;

[0033] A4, predict water quality conditions based on water quality test data, historical water quality data, and change rate;

[0034] Predict hydrological characteristics based on hydrological characteristics data and historical hydrological characteristics data;

[0035] A5, detect water quality anomalies based on water quality standards and predicted water quality conditions;

[0036] Detect hydrological characteristic anomalies based on hydrological characteristic standards and predicted hydrological characteristic conditions;

[0037] A6, use tracing technology to track pollution sources and locate the pollution sources;

[0038] A7, combined with Geographic Information System (GIS), analyze the transmission path, impact range and diffusion speed of pollutants;

[0039] A8, use GIS technology to draw a water ecological environment map, mark the location of monitoring points and water quality conditions, and use different colors to represent different water quality levels and pollution levels. For example, transparent, yellow, and black are used to represent pollution-free areas, areas exceeding the standard, and high-risk areas;

[0040] A9. Initiate emergency monitoring mechanisms and disposal measures based on the area ratio of exceeded areas and high-risk areas.

[0041] The beneficial effects of the present invention are as follows: 1. The water quality parameters and hydrological characteristic data are detected and transmitted through ZIGBEE, and the quality of the detected data is evaluated from four dimensions: accuracy, completeness, timeliness, and consistency. Errors, inconsistencies, and missing values ​​in the data are identified, and then data cleaning is performed to correct errors, duplications, inconsistencies, and missing values ​​in the data, and the parameters of the data quality evaluation model are adaptively adjusted to ensure data quality from the source. Finally, encoding is performed, that is, quantization, compression, and encryption processing are performed. The transmission module uses GPS to transmit the encoded data to the monitoring platform, and configures a buffer, a network performance detector, a locator, and a main control. The buffer is triggered when there is a burst of data volume, access speed uncoordinated, or congestion, so as to reduce network transmission data loss and delay, detect the transmission performance of the GPS network, and the load of the data decoding module. Triple buffering is used for transmission and reception to reduce the load pressure of the data decoding module of the back-end monitoring platform, and can prevent the leakage of sensitive data;

[0042] The monitoring platform uses a data decoding module to receive the encoded data transmitted from the communication network, first decrypt and decompress it, and decode the detection data and positioning data, and then perform analysis, early warning and trace the source. It can intuitively display the ecological environment of the water area in the form of a map, and can intelligently conduct analysis, early warning and trace the source, predict the early warning model parameters and perform adaptive optimization strategies for the parameters, thereby avoiding overfitting or underfitting. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a module diagram of the present invention.

[0044] Figure 2 It is a flow chart of data quality assessment and data cleaning steps of the present invention.

[0045] Figure 3 It is a flow chart of the working steps of the data summary module of the present invention. DETAILED DESCRIPTION

[0046] The technical solutions of various embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0047] The following is in conjunction with the instruction manual Figure 1 , the specific implementation modes of the present invention are further described in detail.

[0048] The water ecological environment supervision system based on artificial intelligence includes: data detection module: using a pair of samplers (also called sampling bottles, such as the common LB-8000K water quality automatic sampler, placed in the water area, set a single sampling volume of 500ml sampling) deployed in the water area to obtain the water sample to be tested, and the sensor configured in the sampler detects water quality parameters. The water quality parameters include turbidity sensor, dissolved oxygen sensor, CDD sensor, water temperature sensor and conductivity sensor, ammonia nitrogen sensor, heavy metal water quality sensor, and the sensor outside the sampler deployed in the water area detects hydrological characteristics, including water level detected by water level meter and flow rate, and flow rate detected by ultrasonic flow meter, and enters the data encoding module through the transmission module ZIGBEE;

[0049] The data encoding module first evaluates the data quality, which is used to pre-process the data to reduce interference in the later analysis process. Specifically, the detection data is evaluated from the four dimensions of accuracy, completeness, timeliness, and consistency, and errors, inconsistencies, and missing values ​​in the data are identified. Then, data cleaning is performed to correct errors, duplications, inconsistencies, and missing values ​​in the data, and the parameters of the data quality assessment model are adaptively adjusted to ensure data quality from the source. Finally, encoding is performed to facilitate subsequent data classification statistics and analysis: that is, quantization, compression, and encryption processing. For example, quantization can be binary quantization;

[0050] Compression uses arithmetic coding to reduce the space occupied by data storage, the time occupied by transmission, and reduce data latency;

[0051] Asymmetric ECC encryption is used to ensure data confidentiality, integrity and availability, and reduce data loss;

[0052] The transmission module uses the communication network to transmit the encoded data, and is equipped with a buffer, a network performance detector, a locator and a main control. The main control is a single-chip microcomputer, which triggers the front-end data detection module and the data encoding module to work according to the timing sequence, and the back-end transmission module transmits the GPS to the monitoring platform;

[0053] The monitoring platform uses a data decoding module to receive the encoded data transmitted from the communication network, first decrypt and decompress it, decode the detection data and positioning data, and enter the data summary module;

[0054] The data aggregation module analyzes, warns and traces the received data, and intuitively displays the ecological environment of the water area in the form of a map. It can intelligently and jointly perform analysis, warning and trace, predict the parameters of the warning model and perform adaptive optimization strategies for the parameters, thereby avoiding overfitting or underfitting.

[0055] Based on the above technical solution, the steps of evaluating data quality and then cleaning data are as follows:

[0056] Step 1: Determine the evaluation criteria: including accuracy Z, completeness W, timeliness S, and consistency Y;

[0057] Step 2: Build an evaluation model: P = Z × 35% + W × 14% + S × 26% + Y × 25%, where 35%, 14%, 26%, and 25% are the initial weights. P represents the comprehensive score of data quality assessment, which is an evaluation value that combines accuracy, completeness, timeliness, and consistency.

[0058] Step 3, obtain real-time data and historical data;

[0059] Step 4, compare the deviation between the data predicted by historical data (that is, the data at this moment predicted by historical data, for example, obtained by superimposing the historical data on the change rate of historical data) and the real-time data to obtain the accuracy Z;

[0060] By judging the degree of missingness of the sampled data, the completeness W is obtained;

[0061] By comparing the sampling frequency and the time from sampling to receiving data with the reference time, the timeliness S is obtained;

[0062] The consistency Y is obtained by multiple detections of the same sensor at different time points, or simultaneous detections of different adjacent sensors;

[0063] Step 5, construct a data quality assessment formula and output the data quality assessment value SZ: SZ = Pi ×80%+ Pi -1×12%+ Pi- 2×8%, where Pi represents the data quality assessment score of the current data batch or the current sampling point, "i" here represents the current or specific data batch or sampling point, Pi-1 represents the data quality assessment score of the previous data batch or the previous sampling point, "i-1" here represents the previous or previous data batch or sampling point, Pi-2 represents the data quality assessment scores of the previous two data batches or the previous two sampling points, "i-2" here represents the previous two or previous data batches or sampling points;

[0064] Step 6: Determine whether the data quality assessment value SZ meets the requirements. If yes, perform encoding; if no, perform data cleaning;

[0065] Step 7, firstly, adopt the regression method to interpolate according to the completeness of the data;

[0066] Secondly, identify abnormal data based on accuracy and consistency;

[0067] Finally, for abnormal data, we take timeliness into consideration and take measures to correct or delete outliers.

[0068] On the basis of the above scheme, the evaluation model also includes regular evaluation, adjustment and optimization of weights, the steps are as follows:

[0069] S1, set the initial weights to 35%, 14%, 26%, and 25% respectively;

[0070] S2, uses the hill climbing algorithm to fine-tune the current weight upward and downward, and calculates the data quality assessment value SZ after the weight is modified, which can satisfy the self-adjustment of the weight parameter;

[0071] S3, select the maximum value of data quality assessment value SZ;

[0072] S4, the number of upward and downward fine-tuning quantization is set to 3, the step size is set to 1%, and the total weight is 1. This setting determines the weight of a limited number of times and step size to avoid overfitting and underfitting.

[0073] On the basis of the above technical solution, the transmission module transmits the encoded data using a communication network, and is equipped with a buffer, a network performance detector, a locator, and a master control;

[0074] The communication network includes a ZIGBEE for short-distance transmission, wherein the sensor detection data is transmitted to the data encoding module of the main control side through the ZIGBEE;

[0075] The communication network includes a GPS for long-distance transmission and positioning, and the GPS also includes a timestamp. The master controller transmits the encoded data, positioning data, and timestamp data to the monitoring platform via the GPS;

[0076] The network performance detector detects the transmission performance of the ZIGBEE network, the access speed of the master control and the degree of congestion, and triggers the buffer cache when the burst data volume, access speed uncoordinated, and congestion occur, so as to reduce the loss and delay of network transmission data, detect the transmission performance of the GPS network, and the load of the data decoding module, which can be quantified and evaluated by the bandwidth, network delay, packet loss rate, and throughput using an evaluation model, and adopts triple buffering transmission and reception to reduce the load pressure of the data decoding module of the back-end monitoring platform and prevent the leakage of sensitive data.

[0077] On the basis of the above technical solution, the data aggregation module analyzes, warns and traces the received data, and intuitively displays the ecological environment of the water area in the form of a map. The specific steps are as follows:

[0078] A1, divide the data into data groups according to the type and measurement point. For example, the K-means clustering algorithm can be used to select all the measurement point data to divide the data groups as needed. Furthermore, according to the measurement points and the amount of data, the number of clusters to be grouped in the data group is determined to further divide the sub-data groups;

[0079] A2, according to the divided data groups, extract abnormal features in the data groups, abnormal features refer to abnormal values ​​and outlier values ​​of the data;

[0080] A3, consistency analysis of abnormal feature-related neighboring points, yes, normal data, no, median compensation;

[0081] A4, predict water quality conditions based on water quality test data, historical water quality data, and change rate;

[0082] Predict hydrological characteristics based on hydrological characteristics data and historical hydrological characteristics data;

[0083] The following is an example of a model and steps for describing prediction:

[0084] Step A41, calculating the difference between the water quality test data and the water quality historical data, and obtaining the change value ⊿φ;

[0085] Step A42, calculating the change rate b, b=⊿φ / ⊿t, where ⊿t is time;

[0086] Step A43, using the exponential extrapolation method Y=abt to predict the changing trend of water quality conditions, where Y represents the trend value, a represents the intercept, b represents the rate of change, and t represents the time, substitute the data to generate the predicted water quality condition value;

[0087] Step A44, the change rate b is corrected by the historical predicted change rates bi-1 and bi-2, namely, the change rate b×70%+bi-1×20%+Pi-2×10%. Specifically, bi-1 represents the change rate of the previous time period (or the previous data point), and bi-2 represents the change rate of the previous two time periods (or the previous two data points);

[0088] Step A45, using a hill climbing algorithm to fine-tune and quantize the current weight upward or downward, and calculating the predicted data after the modified weight is close to the true value of the data;

[0089] Step A46, selecting the weight closest to the true value of the data;

[0090] Step A47, the number of upward and downward fine-tuning quantization is set to 6, the step size is set to 0.8%, and the total weight is 1, so as to set and determine the weight of the limited number of times and step size to avoid overfitting and underfitting;

[0091] A5, detect water quality anomalies based on water quality standards and predicted water quality conditions;

[0092] Detect hydrological characteristic anomalies based on hydrological characteristic standards and predicted hydrological characteristic conditions;

[0093] A6, use tracing technology to track pollution sources and locate the location of pollution sources. For example, the Gaussian model can be used to track pollution sources based on the speed of water flow in the water area and the flow direction caused by the topography.

[0094] A7, combined with the geographic information system (GIS), the speed of water flow in the water area, the flow direction caused by the landform, and the degradation situation, the propagation path, impact range, and diffusion speed of the pollutants are analyzed, and the Gaussian model is constructed to track and locate the pollution source. After the pollution source is determined, the impact range and diffusion speed are calculated in turn. The following is an embodiment describing the steps of locating the pollution source of the Gaussian model:

[0095] Step A71, model assumption, assuming that the pollution source concentration distribution of the pollution source follows a Gaussian distribution, that is, the pollutant concentration of the pollution source decreases exponentially with increasing distance from the source;

[0096] Step A72, parameter setting, pollutant concentration (C) at the test point, pollution source concentration (q), distance from the pollution source (x, y), water flow velocity (u), flow direction (θ) and diffusion coefficient (σ);

[0097] Step A73, set the formula for the two-dimensional water area to consider only the lateral diffusion, C(x, y) = (q / 2πθuσy) exp [-1 / 2 (y*y) / (σy*σy)], where σy represents the lateral diffusion coefficient, σy = f(x), where C(x, y) represents the pollutant concentration at the position (x, y), and y represents the distance of the test point relative to the pollution source in the lateral direction (i.e., the y direction);

[0098] Step A74, bringing the detected water quality data and hydrological characteristic data into the formula, and simulating and calculating the distribution diagram of pollutant concentration at different locations and different time points;

[0099] Step A75, calculating the distance (q) between the measurement point where the pollutant concentration (C) of the measured point is close to zero and the pollution source concentration, and obtaining the influence range;

[0100] Step A76, optimize the Gaussian model parameters, use the hill climbing algorithm to fine-tune and quantize the current parameters upward and downward, and calculate the adaptability and prediction accuracy of the model after the parameters are modified.

[0101] A8, use GIS technology to draw a water ecological environment map, mark the location of monitoring points and water quality conditions, and use different colors to represent different water quality levels and pollution levels. For example, transparent, yellow, and black are used to represent pollution-free areas, areas exceeding the standard, and high-risk areas;

[0102] A9. Initiate emergency monitoring mechanisms and disposal measures based on the area ratio of exceeded areas and high-risk areas.

[0103] When the present invention is used specifically, water quality parameters and hydrological characteristic data are detected and transmitted through ZIGBEE, and the quality of the detected data is evaluated from four dimensions: accuracy, completeness, timeliness and consistency. The process is: determine the evaluation standard; build an evaluation model; obtain real-time data and historical data; compare the deviation between the historical data predicted data and the real-time data to obtain the accuracy Z; determine the degree of missing data by judging the completeness W; determine the timeliness S by judging the sampling frequency and the time from sampling to receiving data and the reference time; determine the consistency Y by multiple detections at different time points of the same sensor, or simultaneous detections by different adjacent sensors; construct a data quality evaluation formula, and output the data quality evaluation value SZ: SZ = Pi × 80% + Pi -1 × 12% + Pi- 2 × 8%; determine whether the data quality evaluation value SZ meets the requirements, if yes, encode; if not, perform data cleaning; first, adopt the regression method to interpolate according to the completeness of the data;

[0104] Secondly, identify abnormal data based on accuracy and consistency;

[0105] Finally, for abnormal data, we take timeliness into consideration and take measures to correct or delete abnormal values;

[0106] It also includes regular evaluation, adjustment and optimization of weights, including setting initial weights; using hill climbing algorithm to fine-tune the current weights upward and downward; selecting the maximum value of data quality assessment value; setting the number of upward and downward fine-tuning quantization to 3, the step size to 1%, and the total weight to 1, so as to set a limited number of times and step size weights to avoid overfitting and underfitting; thereby ensuring data quality from the source, and finally encoding, that is, quantization, compression, and encryption processing. The transmission module uses GPS to transmit the encoded data to the monitoring platform, and configures the cache, network performance detector, locator and master control, and triggers the cache when there is burst data volume, access speed uncoordinated, or congestion, which is used to reduce network transmission data loss and delay, and detect GPS In order to improve the transmission performance of the S network and the load of the data decoding module, triple buffering is adopted for transmission and reception to reduce the load pressure of the data decoding module of the back-end monitoring platform and prevent the leakage of sensitive data; the monitoring platform adopts a data decoding module to receive the encoded data transmitted from the communication network, first decrypt and decompress, decode the detection data and positioning data, and then analyze, warn and trace the source, and intuitively display the ecological environment of the water area in the form of a map. It can intelligently and jointly analyze, warn and trace the source, predict the parameters of the warning model and perform adaptive optimization strategies for the parameters. The process is to divide the data into data groups according to type and measurement point; according to the divided data groups, extract abnormal features in the data groups, and the abnormal features refer to the abnormal values ​​and outlier values ​​of the data;

[0107] Abnormal feature association neighbor point consistency analysis, yes, it is normal data, no, median compensation;

[0108] Predict water quality conditions based on water quality test data, historical water quality data, and change rate;

[0109] Predict hydrological characteristics based on hydrological characteristics data and historical hydrological characteristics data;

[0110] Detect water quality anomalies based on water quality standards and predicted water quality conditions;

[0111] Detect hydrological characteristic anomalies based on hydrological characteristic standards and predicted hydrological characteristic conditions;

[0112] Use tracing technology to track pollution sources and locate the location of pollution sources. For example, a Gaussian model can be used to track pollution sources based on the speed of water flow in the water area and the flow direction caused by the topography.

[0113] Combined with the Geographic Information System (GIS), the speed of water flow in the water area, the flow direction caused by the landform, and the degradation situation, the propagation path, impact range, and diffusion speed of pollutants are analyzed. By constructing a Gaussian model to track and locate, the impact range and diffusion speed are calculated in turn after the pollution source is determined;

[0114] Use GIS technology to draw a water ecological environment map, mark the location of monitoring points and water quality conditions, and use different colors to represent different water quality levels and pollution levels. For example, transparent, yellow, and black are used to represent non-polluted areas, areas exceeding the standard, and high-risk areas;

[0115] Based on the area ratio of exceeded areas and high-risk areas, emergency monitoring mechanisms and disposal measures were activated to achieve intelligent management and control of the water ecological environment.

Claims

1. The water ecological environment supervision system based on artificial intelligence is characterized by: include: Data detection module: a sampler placed in the water area is used to obtain the water sample to be tested. The sensor configured in the sampler detects the water quality parameters. The sensor outside the sampler placed in the water area detects the hydrological characteristics and enters the data encoding module; The data encoding module first evaluates the data quality, then cleans the data, and finally encodes it: quantization, compression, and encryption; The transmission module transmits the encoded data via a communication network and is configured with a buffer, a network performance detector, and a locator; The data decoding module receives the encoded data transmitted from the communication network, decodes the detection data and positioning data, and enters the data aggregation module; The data summary module analyzes, warns and traces the received data, and intuitively displays the ecological environment of the water area in the form of a map; The steps of evaluating data quality and then cleaning data are as follows: Step 1: Determine the evaluation criteria: including accuracy Z, completeness W, timeliness S, and consistency Y; Step 2: Build an evaluation model: P = Z × 35% + W × 14% + S × 26% + Y × 25%, where P represents the comprehensive score of data quality assessment; Step 3, obtain real-time data and historical data; Step 4, compare the deviation between the historical data prediction data and the real-time data to obtain the accuracy Z; By judging the degree of missingness of the sampled data, the completeness W is obtained; By comparing the sampling frequency and the time from sampling to receiving data with the reference time, the timeliness S is obtained; The consistency Y is obtained by multiple detections of the same sensor at different time points, or simultaneous detection by different sensors; Step 5: Construct a data quality assessment formula and output the data quality assessment value SZ: SZ = P i ×80%+P i -1 ×12%+P i- 2 ×8%, of which P i Represents the data quality assessment score of the current data batch or current sampling point, P i-1 and P i-2 Represent the data quality assessment scores of the previous and previous two data batches or sampling points respectively; Step 6: Determine whether the data quality assessment value SZ meets the requirements. If yes, perform encoding; if no, perform data cleaning; Step 7: Data cleaning: First, based on the completeness of the data, regression method is used for interpolation; Secondly, identify abnormal data based on accuracy and consistency; Finally, for abnormal data, we take timeliness into consideration and take measures to correct or delete abnormal values; The evaluation model also includes regular evaluation, adjustment and optimization of weights, the steps are: S1, set the initial weights to 35%, 14%, 26%, and 25% respectively; S2, using the hill climbing algorithm to fine-tune the current weight upward and downward, and calculate the data quality assessment value SZ after the weight is modified; S3, filter out the maximum value of data quality assessment value SZ; S4, the number of upward and downward fine-tuning quantization is set to 3, the step size is set to 1%, and the total weight is 1.

2. The water ecological environment monitoring system based on artificial intelligence according to claim 1 is characterized in that: The water quality parameters include turbidity sensor, dissolved oxygen sensor, CDD sensor, water temperature sensor and conductivity sensor, ammonia nitrogen sensor, and heavy metal water quality sensor; The hydrological characteristics include the water level and flow rate detected by a water level meter, and the flow rate detected by an ultrasonic flow meter.

3. The water ecological environment monitoring system based on artificial intelligence according to claim 1 is characterized in that: The transmission module transmits the encoded data through a communication network and is configured with a buffer, a network performance detector, a locator, and a master control; The communication network includes ZIGBEE for short-range transmission; The communication network includes GPS for long-distance transmission and positioning, and the GPS also includes a timestamp; The network performance detector detects the transmission performance of the ZIGBEE network, the access speed of the master control and the congestion level, triggers the buffer cache when the burst data volume, access speed uncoordinated, and congestion occur, detects the transmission performance of the GPS network, the load of the data decoding module, and adopts triple buffering transmission and reception.

4. The water ecological environment monitoring system based on artificial intelligence according to claim 1 is characterized in that: The data aggregation module analyzes, warns and traces the received data, and intuitively displays the ecological environment of the water area in the form of a map. The specific steps are as follows: A1, divide the data into data groups according to type and measurement point; A2, extract abnormal features within the data group according to the divided data groups; A3, consistency analysis of abnormal feature-related neighboring points, yes, normal data, no, median compensation; A4, predict water quality conditions based on water quality test data, historical water quality data, and change rate; Predict hydrological characteristics based on hydrological characteristics data and historical hydrological characteristics data; A5, detect water quality anomalies based on water quality standards and predicted water quality conditions; Detect hydrological characteristic anomalies based on hydrological characteristic standards and predicted hydrological characteristic conditions; A6, using tracing technology to track pollution sources, locate the location of pollution sources, and use Gaussian models to track according to the speed of water flow in the water area and the flow direction caused by the topography; A7, combined with Geographic Information System (GIS), to analyze the transmission path, impact range and diffusion speed of pollutants; A8, use GIS technology to draw a water ecological environment map, mark the location of monitoring points and water quality conditions, and use different colors to represent different water quality levels and pollution levels; A9. Initiate emergency monitoring mechanisms and disposal measures based on the area ratio of exceeded areas and high-risk areas.

Images