Reservoir environment monitoring method and system

Abstract

The invention provides a reservoir environment monitoring method and system, and belongs to the field of reservoir water quality monitoring. The method comprises the following steps: acquiring vertical temperature time sequence data and a plurality of layered fluctuation coefficients under a plurality of depth intervals; setting a water quality sensor monitoring frequency, and obtaining a first water quality parameter of the first depth interval; configuring a prediction resource based on the first layered fluctuation coefficient, predicting a water quality parameter of a second depth interval according to the first water quality parameter and the first layered fluctuation coefficient, and verifying to obtain a prediction error coefficient; and performing layer-by-layer prediction according to the prediction error coefficient and the layered fluctuation coefficient until the water quality parameter of the target depth and the reservoir environment monitoring result are obtained. The technical problems of inaccurate target depth water quality monitoring and high cost caused by the fact that a water quality sensor cannot realize full-depth coverage in the prior art are solved, and the technical effects of accurately monitoring the target depth water quality and reducing the deployment cost of the sensor are achieved through temperature stratified fluctuation analysis and layer-by-layer prediction error compensation.

Description

Technical Field

[0001] This invention relates to the field of reservoir water quality monitoring, and in particular to a reservoir environmental monitoring method and system. Background Technology

[0002] As important water resource storage facilities, reservoirs play a crucial role in ensuring water supply security through water quality monitoring. Due to the significant depth of reservoirs, water quality parameters at different depths often vary considerably, necessitating effective monitoring of multiple depth layers.

[0003] Currently, traditional reservoir water quality monitoring methods primarily employ real-time monitoring by installing water quality sensors at different depths. To achieve full-depth coverage monitoring, corresponding water quality sensors need to be configured at each depth layer, resulting in a large number of sensors and extremely high equipment investment and maintenance costs. However, due to cost constraints, existing technologies often only allow for sensor installation at a limited number of depths, failing to achieve continuous monitoring across all depths, particularly lacking accuracy in monitoring water quality parameters at the target depth. Summary of the Invention

[0004] This invention addresses the technical problem in existing technologies where water quality sensors cannot provide full-depth coverage, leading to inaccurate water quality monitoring at target depths and high costs. It provides a method and system for reservoir environmental monitoring to solve this problem.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0006] In a first aspect, the present invention provides a reservoir environment monitoring method, comprising: acquiring vertical temperature time-series data in multiple depth intervals within the reservoir, obtaining multiple interval temperature sequences, calculating multiple interval temperature change rates, and obtaining multiple stratified fluctuation coefficients; setting multiple monitoring frequencies of water quality sensors configured in multiple depth intervals based on the multiple stratified fluctuation coefficients, and acquiring a first water quality parameter of a first water quality sensor in a first depth interval; configuring a first prediction resource based on the first stratified fluctuation coefficient of the first depth interval, predicting a second water quality parameter in a second depth interval based on the first water quality parameter and the first stratified fluctuation coefficient, verifying it with the second water quality parameter monitored by the second water quality sensor, and obtaining a first prediction error coefficient; configuring a second prediction resource based on the first prediction error coefficient and the second stratified fluctuation coefficient of the second depth interval, predicting and acquiring a third water quality parameter in a third depth interval, until a target water quality parameter and a final prediction error coefficient are predicted at a target depth, calculating water quality impact parameters, and obtaining the water quality environment monitoring results of the reservoir.

[0007] Secondly, the present invention provides a reservoir environment monitoring system, comprising: a temperature stratification monitoring module, used to test and acquire vertical temperature time-series data in multiple depth intervals within the reservoir, obtain multiple interval temperature sequences, calculate multiple interval temperature change rates, and obtain multiple stratification fluctuation coefficients; a water quality parameter acquisition module, used to set multiple monitoring frequencies of water quality sensors configured in multiple depth intervals according to the multiple stratification fluctuation coefficients, and acquire a first water quality parameter of a first water quality sensor in a first depth interval; a prediction verification module, used to configure a first prediction resource based on the first stratification fluctuation coefficient of the first depth interval, predict a predicted second water quality parameter in a second depth interval according to the first water quality parameter and the first stratification fluctuation coefficient, verify it with the second water quality parameter monitored by the second water quality sensor, and obtain a first prediction error coefficient; and a water quality assessment module, used to configure a second prediction resource according to the first prediction error coefficient and the second stratification fluctuation coefficient of the second depth interval, predict and acquire a third water quality parameter in a third depth interval, until a target water quality parameter and a final prediction error coefficient are predicted at a target depth, calculate water quality impact parameters, and obtain the water quality environment monitoring results of the reservoir.

[0008] The beneficial effects of this invention are:

[0009] The test acquired vertical temperature time-series data across multiple depth ranges within the reservoir, obtaining temperature sequences for each range and calculating temperature change rates and stratified fluctuation coefficients. Analyzing temperature fluctuations within these depth ranges reflects the diffusion of water pollutants, providing a basis for subsequent monitoring strategies. Based on these stratified fluctuation coefficients, multiple monitoring frequencies were set for water quality sensors deployed across the various depth ranges. The first water quality parameter of the first sensor in the first depth range was acquired. A dynamic monitoring configuration was implemented by using a strategy of increasing the monitoring frequency for larger fluctuations, starting water quality parameter acquisition from the shallowest depth range. The first stratified fluctuation coefficient of the first depth range was used to configure the first... The system predicts water quality parameters for the second depth range based on the first water quality parameter and the first stratified fluctuation coefficient. These parameters are then verified against the second water quality parameter monitored by the second water quality sensor to obtain the first prediction error coefficient. This process extends the depth range and establishes an error feedback mechanism. Based on the first prediction error coefficient and the second stratified fluctuation coefficient of the second depth range, the system configures the second prediction resource to predict and obtain the third water quality parameter for the third depth range. This process continues until the target water quality parameter and the final prediction error coefficient are obtained at the target depth. The system then calculates the water quality impact parameters and obtains the water quality environment monitoring results for the reservoir. By predicting layer by layer and adjusting the prediction strategy using error compensation, the system achieves accurate monitoring of water quality at the target depth.

[0010] The above technical solution utilizes the temperature stratification fluctuation characteristics to guide water quality monitoring strategies. By adopting a layer-by-layer prediction and error compensation approach, accurate monitoring of water quality at the target depth is achieved without the need for full-depth sensor deployment. This ensures monitoring accuracy while effectively reducing sensor deployment costs. Attached Figure Description

[0011] Figure 1 A schematic flowchart of a reservoir environmental monitoring method provided by the present invention;

[0012] Figure 2 This is a schematic diagram of the structure of a reservoir environmental monitoring system provided by the present invention.

[0013] In the attached diagram, the components represented by each number are as follows:

[0014] Temperature stratification monitoring module 11, water quality parameter acquisition module 12, prediction and verification module 13, water quality assessment module 14. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0017] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0018] Example 1, as Figure 1 As shown in the figure, an embodiment of the present invention provides a method for monitoring the environment of a reservoir, including:

[0019] S1. Test and obtain vertical temperature time series data in multiple depth ranges within the reservoir, obtain multiple temperature series in multiple ranges, calculate the temperature change rate in multiple ranges, and obtain multiple stratified fluctuation coefficients.

[0020] Specifically, firstly, multiple depth intervals are defined vertically within the reservoir, each representing a layer of the water body. Temperature sensors deployed in each depth interval continuously collect temperature data from each depth layer within a preset time range, forming vertical temperature time-series data. These vertical temperature time-series data are arranged chronologically to constitute the temperature time-series dataset corresponding to each depth interval, i.e., multiple interval temperature sequences.

[0021] Subsequently, the temperature sequences of each interval were analyzed and processed. By calculating the average variation range of temperature data for each depth interval per unit time, the temperature fluctuation characteristics of each water layer were quantified, and multiple interval temperature change rates were obtained. Furthermore, the temperature change rates of each interval were compared with the historical maximum temperature change rate, and the ratio was calculated to obtain standardized stratified fluctuation coefficients, resulting in multiple stratified fluctuation coefficients corresponding to multiple depth intervals. The stratified fluctuation coefficients reflect the thermodynamic stability and mixing degree of the water body at each depth interval. A larger stratified fluctuation coefficient indicates more drastic temperature changes, more active water mixing, and a higher probability of pollutant diffusion and transport at that depth. Obtaining multiple stratified fluctuation coefficients provides a basis for subsequent optimization of water quality monitoring frequency and prediction of water quality parameters, and lays the foundation for analyzing the diffusion behavior of water pollutants.

[0022] S2. Based on multiple stratified fluctuation coefficients, set multiple monitoring frequencies of water quality sensors configured in multiple depth ranges, and obtain the first water quality parameters of the first water quality sensor in the first depth range.

[0023] Specifically, firstly, the preset monitoring frequency of the water quality sensor is obtained as a baseline parameter. Then, based on the stratification fluctuation coefficient corresponding to each depth interval, the preset monitoring frequency is corrected. The correction principle is: for depth intervals with larger stratification fluctuation coefficients, the corresponding water quality sensor monitoring frequency is higher to ensure more intensive monitoring of actively mixing areas of the water body; conversely, for depth intervals with smaller stratification fluctuation coefficients, the monitoring frequency can be appropriately reduced to achieve a reasonable allocation of monitoring resources. Through the above correction calculations, multiple monitoring frequencies corresponding one-to-one with each depth interval are obtained.

[0024] Subsequently, based on the calculated monitoring frequencies, the water quality sensors configured in multiple depth ranges were reconfigured and their parameters set, thus configuring the water quality sensors in multiple depth ranges. Then, using the first water quality sensor located in the first depth range, water quality monitoring was performed according to its corresponding monitoring frequency to obtain the first water quality parameter. The first water quality parameter is a water pollution index, characterizing the pollutant concentration level in that depth range; a higher pollutant concentration value indicates a more severe degree of water pollution. The first depth range is the range with the smallest depth value among all depth ranges, i.e., the surface area of ​​the reservoir, and its monitoring data (the first water quality parameter) will serve as the starting input parameter for subsequent layer-by-layer prediction.

[0025] By setting multiple monitoring frequencies based on multiple stratified fluctuation coefficients, the monitoring frequency and hydrodynamic characteristics of the water body were combined. While ensuring monitoring accuracy, the configuration of sensor resources was optimized, and the overall monitoring efficiency was improved.

[0026] S3. Configure the first prediction resource based on the first layer fluctuation coefficient of the first depth range, predict the second water quality parameter of the second depth range according to the first water quality parameter and the first layer fluctuation coefficient, verify it with the second water quality parameter monitored by the second water quality sensor, and obtain the first prediction error coefficient.

[0027] Specifically, firstly, a first prediction resource is configured based on the first stratified fluctuation coefficient of the first depth interval. Specifically, a water quality diffusion prediction network group containing T water quality diffusion prediction networks is pre-constructed, where T is the total number of networks in the group, and is a positive integer. The number of first networks is determined by multiplying the first stratified fluctuation coefficient of the first depth interval with the total number of networks T and then rounding down. Subsequently, water quality diffusion prediction networks corresponding to the number of first networks are randomly selected from the water quality diffusion prediction network group to constitute the first prediction resource. This dynamic configuration mechanism ensures the matching between the prediction resource and the water body fluctuation characteristics: the larger the stratified fluctuation coefficient, the more water quality diffusion prediction networks are selected, and the higher the reliability of the integrated prediction.

[0028] Then, the obtained first water quality parameters and first stratification fluctuation coefficients are simultaneously input into various water quality diffusion prediction networks in the first prediction resource for parallel computation. Each water quality diffusion prediction network outputs a corresponding prediction result based on the input parameters. Subsequently, the arithmetic mean of the output results of all water quality diffusion prediction networks is calculated to obtain the predicted second water quality parameters for the second depth range. Simultaneously, actual monitoring is performed using a second water quality sensor configured within the second depth range to obtain the second water quality parameters for that depth range, i.e., the pollutant concentration level within that depth range. Afterward, the predicted second water quality parameters are compared and analyzed with the measured second water quality parameters to calculate the error magnitude between the two, obtaining the first prediction error coefficient. This first prediction error coefficient quantifies the accuracy of the first prediction resource and provides important feedback information for optimizing the prediction accuracy of subsequent depth ranges.

[0029] By integrating multi-network prediction and real-time verification mechanisms, the robustness of prediction is improved, and an error quantification system is established, laying a reliable foundation for progressively deeper prediction.

[0030] S4. Based on the first prediction error coefficient and the second stratified fluctuation coefficient of the second depth interval, configure the second prediction resource, predict and obtain the third water quality parameter of the third depth interval, until the target water quality parameter and the final prediction error coefficient of the target depth are predicted and obtained, calculate the water quality impact parameter, and obtain the water quality environment monitoring results of the reservoir.

[0031] Specifically, firstly, error compensation is calculated based on the obtained first prediction error coefficient and the second stratified fluctuation coefficient for the second depth range. The first prediction error coefficient and the second stratified fluctuation coefficient are summed to obtain the second compensated fluctuation coefficient, where the maximum value of the compensated fluctuation coefficient is limited to 1 to ensure the standardization characteristics of the parameters. This compensation mechanism effectively integrates the accuracy information of the previous layer's prediction with the hydrodynamic characteristics of the current layer, improving the adaptability of the prediction strategy.

[0032] Then, based on the second compensation fluctuation coefficient and the total number T of the water quality diffusion prediction network group, a multiplication and rounding operation is performed to determine the second number of networks. A corresponding number of prediction networks are then selected from the water quality diffusion prediction network group to form the second prediction resource. The second prediction resource is used to predict water quality diffusion in the third depth range, obtaining the predicted third water quality parameters for the third depth range.

[0033] Following the aforementioned prediction-verification-compensation cyclical mechanism, layer-by-layer water quality prediction continues. Each layer's prediction is dynamically adjusted based on the error information of the previous layer, until the target water quality parameters and final prediction error coefficients at the target depth are obtained. The target depth is the deepest depth interval in a sequence of multiple depth intervals, representing the water quality status of the bottom layer of the reservoir.

[0034] Subsequently, a water quality impact assessment is conducted based on the target water quality parameters and the final prediction error coefficient. Specifically, firstly, the ratio of the target water quality parameter to the rated water quality parameter at the target depth is calculated to obtain the basic water quality impact parameters; then, compensation calculations are performed on the basic water quality impact parameters based on the final prediction error coefficient to obtain the comprehensive water quality impact parameters. These water quality impact parameters comprehensively reflect the deep water quality status of the reservoir and its degree of impact on the overall aquatic environment, forming the reservoir's water quality environmental monitoring results.

[0035] By employing error feedback and layer-by-layer compensation mechanisms, the limitations of traditional monitoring methods in deep water quality detection are overcome, achieving high-precision full-depth water quality prediction and environmental impact assessment. Through the aforementioned layer-by-layer prediction and error compensation, the technical problem of inaccurate and costly water quality monitoring at target depths due to the inability of water quality sensors to cover the entire depth is effectively solved. Compared to traditional dense sensor deployment schemes, only a small number of sensors need to be deployed in key depth layers. Accurate monitoring of water quality at target depths can be achieved through temperature stratification fluctuation analysis and layer-by-layer prediction error compensation, significantly reducing sensor deployment costs while ensuring monitoring accuracy, providing an economical and efficient solution for reservoir environmental monitoring.

[0036] Furthermore, the test acquired vertical temperature time-series data at multiple depth intervals within the reservoir, obtaining multiple interval temperature sequences, and calculated multiple interval temperature change rates, resulting in multiple stratified fluctuation coefficients, including:

[0037] S11. Test and acquire temperature data at multiple times within a preset time range in multiple depth intervals within the reservoir to obtain vertical temperature time series data;

[0038] S12. Sort the temperature data of multiple depth intervals according to time order to obtain multiple interval temperature sequences;

[0039] S13. Based on multiple interval temperature sequences, calculate multiple interval temperature change rates and obtain multiple stratified fluctuation coefficients.

[0040] In one feasible implementation, firstly, multiple depth intervals are set at predetermined intervals on the vertical profile of the reservoir, each depth interval representing a water layer within a specific depth range. Temperature sensors are deployed within each depth interval to continuously collect temperature data at multiple time points within a preset time range (e.g., 24 hours, 48 ​​hours, or longer). The collection frequency can be set according to monitoring needs, such as collecting data every 15 minutes, 30 minutes, or 1 hour. Through the above continuous collection, vertical temperature time-series data encompassing multiple depth intervals is obtained.

[0041] Subsequently, the collected vertical temperature time-series data were processed and categorized. This data includes temperature data from multiple moments within a preset time range across multiple depth intervals. The temperature data for each depth interval was sorted chronologically according to the order of data acquisition, ensuring that the temperature data for each depth interval exhibits a continuous trend along the time axis. After sorting, multiple interval temperature sequences were formed, each corresponding to a specific depth interval. Each interval temperature sequence reflects the temperature evolution pattern of its corresponding depth interval within the preset time range.

[0042] Then, mathematical analysis and feature extraction are performed based on multiple interval temperature sequences. Specifically, based on each interval temperature sequence, the average variation amplitude of temperature data for each depth interval per unit time is calculated, quantifying the intensity of temperature fluctuations in the water at each depth layer, and obtaining multiple interval temperature change rates. Subsequently, the temperature change rate of each interval is compared with the maximum temperature change rate in historical statistical data to obtain multiple standardized stratified fluctuation coefficients. These stratified fluctuation coefficients eliminate the influence of absolute values, facilitating cross-sectional comparative analysis between different depth intervals.

[0043] Through the above steps, multiple stratified fluctuation coefficients for various depth ranges were obtained, providing a solid data foundation for subsequent monitoring frequency configuration and water quality prediction.

[0044] Furthermore, based on multiple interval temperature sequences, multiple interval temperature change rates are calculated, resulting in multiple stratified fluctuation coefficients, including:

[0045] S131. Based on multiple interval temperature sequences, calculate the average change rate of temperature data in multiple depth intervals per unit time to obtain the temperature change rate of multiple intervals.

[0046] S132. Calculate the ratio of the temperature change rate in multiple intervals to the historical maximum temperature change rate to obtain multiple stratified fluctuation coefficients.

[0047] In a preferred embodiment, the temperature sequences of each depth interval are first subjected to quantitative analysis. Specifically, for the temperature sequence of each depth interval, the time period is divided according to a preset unit time interval (e.g., 1 hour, 2 hours, etc.), resulting in multiple temperature data subsets within multiple unit time intervals. For each unit time interval, the difference between the maximum and minimum values ​​of the temperature data within that unit time interval is calculated to obtain the temperature change amplitude corresponding to that unit time interval. Through the above calculation, multiple temperature change amplitudes corresponding to multiple unit time intervals are obtained. Subsequently, the arithmetic mean of the temperature change amplitudes of all unit time intervals is calculated to obtain the average temperature change amplitude of that depth interval, which is used as the interval temperature change rate of that depth interval. The above calculation process is repeated for all depth intervals to obtain multiple interval temperature change rates corresponding one-to-one with each depth interval. The multiple interval temperature change rates quantify the average temperature fluctuation intensity of the water body in each depth interval within a unit time, providing basic data for subsequent stratified fluctuation analysis.

[0048] Subsequently, the historical maximum temperature change rate was introduced as a standardized reference benchmark. This historical maximum temperature change rate was obtained by analyzing the maximum temperature change rate of each depth interval in the reservoir's historical monitoring data, representing the extreme level of temperature fluctuation in the reservoir. The obtained temperature change rate for each interval was divided by the historical maximum temperature change rate to calculate the ratio, resulting in several standardized stratified fluctuation coefficients. The numerical range of the stratified fluctuation coefficients is between 0 and 1, where values ​​close to 1 indicate that the temperature fluctuation in that depth interval is close to the historical extreme level; values ​​close to 0 indicate that the temperature in that depth interval is relatively stable.

[0049] The above processing ensures the comparability and consistency of the stratified fluctuation coefficients across different time periods and environmental conditions, providing a reliable quantitative basis for subsequent monitoring strategy formulation and prediction model configuration.

[0050] Furthermore, based on multiple stratified fluctuation coefficients, multiple monitoring frequencies are set for water quality sensors configured in multiple depth ranges to obtain the first water quality parameters of the first water quality sensor in the first depth range, including:

[0051] S21. Obtain the preset monitoring frequency for water quality monitoring by the water quality sensor, and perform correction calculations on the preset monitoring frequency based on multiple stratified fluctuation coefficients to obtain multiple monitoring frequencies.

[0052] S22. Configure the monitoring frequency of water quality sensors located in multiple depth ranges according to multiple monitoring frequencies.

[0053] S23. The first water quality parameter is monitored and obtained by the first water quality sensor in the first depth range, wherein the first depth range is the depth range with the smallest depth.

[0054] In one feasible implementation, firstly, a preset monitoring frequency for water quality monitoring by the water quality sensor is obtained. This preset monitoring frequency is the default baseline monitoring interval for the water quality sensor, such as monitoring once every 1 hour or once every 2 hours. Then, based on multiple obtained stratified fluctuation coefficients, the preset monitoring frequency is corrected. The correction algorithm uses a direct proportional relationship: the stratified fluctuation coefficient for each depth interval is multiplied by the preset monitoring frequency. The deeper the stratified fluctuation coefficient, the higher the corrected monitoring frequency, i.e., the shorter the monitoring interval; the deeper the stratified fluctuation coefficient, the lower the corrected monitoring frequency, i.e., the longer the monitoring interval. Through the above correction calculation, multiple monitoring frequencies corresponding one-to-one with each depth interval are obtained, achieving matching between the monitored resources and the water body fluctuation characteristics.

[0055] Subsequently, based on the obtained monitoring frequencies, the parameters of the water quality sensors deployed in each depth range were reconfigured. Specifically, the sensor control system sent frequency setting commands to the water quality sensors in each depth range, writing the corresponding monitoring frequency parameters into the control program of the corresponding water quality sensor, thus completing the dynamic adjustment and optimization of the monitoring frequency. Afterwards, water quality parameters were collected by the first water quality sensor located in the first depth range. The first depth range is the range with the smallest depth value among all depth ranges, corresponding to the surface water area of ​​the reservoir. The first water quality sensor continuously monitored according to its corresponding monitoring frequency, acquiring the first water quality parameter, which is a water pollution index characterizing the pollutant concentration level in that depth range. The first water quality parameter serves as the initial input data for layer-by-layer prediction, providing a basis for subsequent deep water quality prediction.

[0056] Through the above steps, the monitoring frequency and water body characteristics were organically combined, and the efficiency of sensor resource allocation was optimized while ensuring monitoring accuracy.

[0057] Furthermore, based on the first stratification fluctuation coefficient of the first depth range, a first prediction resource is configured. Based on the first water quality parameters and the first stratification fluctuation coefficient, a predicted second water quality parameter for the second depth range is obtained. This parameter is then verified against the second water quality parameter monitored by the second water quality sensor to obtain a first prediction error coefficient, including:

[0058] S31. Obtain a water quality diffusion prediction network group, wherein the water quality diffusion prediction network group includes T water quality diffusion prediction networks, where T is a positive integer;

[0059] S32. Based on the first layer fluctuation coefficient of the first depth interval, combined with T, the number of the first network is calculated and determined, and the water quality diffusion prediction network of the first network number is randomly selected as the first prediction resource.

[0060] S33. Input the first water quality parameter and the first stratified fluctuation coefficient into the water quality diffusion prediction network of the first network number within the first predicted resource, calculate the mean of the output results, and obtain the predicted second water quality parameter for the two depth intervals.

[0061] S34. Obtain the second water quality parameters within the second depth range monitored by the second water quality sensor;

[0062] S35. Calculate the error range between the second water quality parameter and the predicted second water quality parameter to obtain the first prediction error coefficient.

[0063] In a preferred embodiment, a water quality diffusion prediction network group is pre-constructed, comprising T independently trained water quality diffusion prediction networks, where T is a positive integer, typically ranging from 10 to 100. Each water quality diffusion prediction network is built based on a machine learning algorithm, employing the same network architecture but trained using different subsets of training data. This ensures that each water quality diffusion prediction network has certain differences and complementarities in its prediction characteristics, providing a diverse computational foundation for integrated prediction.

[0064] Then, a dynamic prediction resource allocation mechanism is established to achieve an optimal balance between prediction accuracy and computational resources. Specifically, the first-layer fluctuation coefficient of the first depth interval is multiplied by the total number of networks T in the water quality diffusion prediction network group, and the product result is rounded down to determine the number of the first network. This calculation method ensures a positive correlation between prediction resources and water body fluctuation characteristics: the larger the first-layer fluctuation coefficient, the more water quality diffusion prediction networks are selected, and the higher the reliability of the integrated prediction; when the first-layer fluctuation coefficient is small, fewer water quality diffusion prediction networks are selected to improve computational efficiency. Subsequently, water quality diffusion prediction networks corresponding to the number of the first network are randomly selected from the water quality diffusion prediction network group to constitute the first prediction resource.

[0065] Subsequently, an integrated prediction calculation process is executed. The obtained first water quality parameter and first stratified fluctuation coefficient are simultaneously used as input data and input into each of the first number of water quality diffusion prediction networks in the first prediction resource for parallel computation. Each water quality diffusion prediction network independently outputs a prediction result for the water quality status in the second depth interval based on its training parameters and input data. Then, the mean of the output results of all water quality diffusion prediction networks in the first prediction resource is calculated to obtain the predicted second water quality parameter for the second depth interval. This integrated prediction method effectively reduces the random error of single-network predictions and improves the stability and reliability of the prediction results.

[0066] Simultaneously, using a second water quality sensor configured within the second depth range, water quality parameters are collected at their corresponding monitoring frequencies to obtain the second water quality parameters for the second depth range, i.e., the pollutant concentration level in that second depth range, serving as benchmark data for verifying the prediction results. Then, the obtained predicted second water quality parameters are compared and analyzed item by item with the actual second water quality parameters. The relative error between the predicted and actual second water quality parameters is calculated, i.e.: Relative Error = |Predicted Pollutant Concentration Value - Measured Pollutant Concentration Value| / Measured Pollutant Concentration Value. This relative error is the first prediction error coefficient. This first prediction error coefficient reflects the accuracy of the current prediction strategy in predicting pollutant concentrations in the second depth range, providing a unified quantitative basis for optimizing prediction parameters in subsequent depth ranges.

[0067] The aforementioned prediction-verification cycle mechanism ensures the scientific validity and reliability of water quality predictions, laying the foundation for progressively deeper predictions.

[0068] Furthermore, obtain a network of water quality diffusion prediction systems, including:

[0069] S311. Based on historical data of reservoir water quality monitoring, collect a set of sample water quality parameters and a set of sample stratified fluctuation coefficients, and collect water quality parameters for the next depth interval of different sample water quality parameters and sample stratified fluctuation coefficients to obtain a set of sample predicted water quality parameters.

[0070] S312. The set of sample water quality parameters, the set of sample stratified fluctuation coefficients, and the set of sample predicted water quality parameters are randomly divided T times to obtain T sets of water quality diffusion training data.

[0071] S313. Based on machine learning, construct T water quality diffusion prediction networks with the same architecture, and use the T sets of water quality diffusion training data for iterative supervised training. After training and verification convergence, obtain a group of water quality diffusion prediction networks.

[0072] In a preferred embodiment, firstly, training samples for machine learning are constructed based on historical monitoring data. Specifically, based on historical data accumulated from long-term water quality monitoring of the reservoir, data is organized and features are extracted according to time series and depth levels. Water quality parameter data for each depth interval, i.e., pollutant concentration data for each depth interval, are collected from the historical data to form a sample water quality parameter set. Simultaneously, based on historical temperature monitoring data, historical stratified fluctuation coefficients corresponding to each depth interval are obtained to form a sample stratified fluctuation coefficient set. Further, for each set of sample water quality parameters and corresponding sample stratified fluctuation coefficients, actual water quality parameter data for the next depth interval are collected as the prediction target value to form a sample predicted water quality parameter set. Through the above data collection process, a complete training sample containing input features (water quality parameters and stratified fluctuation coefficients) and output targets (water quality parameters for the next depth interval) is established.

[0073] To construct a diverse ensemble of prediction networks, the training samples were randomly partitioned. The set of sample water quality parameters, the set of sample stratified fluctuation coefficients, and the set of sample predicted water quality parameters were used as the overall data source. T independent partitions were performed using random sampling, with each partition yielding a subset of training data containing both input features and output targets. Through these T random partitions, T independent and differentiated water quality diffusion training datasets were obtained. Each dataset differs in sample distribution and feature coverage, providing a data foundation for subsequently constructing multiple complementary water quality diffusion prediction networks.

[0074] Subsequently, a water quality diffusion prediction network ensemble was constructed based on machine learning techniques. Using deep learning or other machine learning algorithms, T identical water quality diffusion prediction networks were built. Each network employs a multi-layer neural network structure: the input layer receives water quality parameters and layered fluctuation coefficients, the hidden layer performs feature extraction and nonlinear mapping, and the output layer predicts the water quality parameters for the next depth interval. T sets of water quality diffusion training data were input into the T water quality diffusion prediction networks for independent iterative supervised training. During training, the network weights and bias parameters were continuously adjusted using the backpropagation algorithm to minimize the prediction error. The training process was completed when the training loss function of each network converged and the validation set performance reached a preset threshold, resulting in a water quality diffusion prediction network ensemble containing T independent water quality diffusion prediction networks.

[0075] Due to the differences in training data, the various water quality diffusion prediction networks in the water quality diffusion prediction network group have complementary and diverse prediction characteristics, which provides reliable technical support for subsequent integrated prediction.

[0076] Furthermore, based on the first prediction error coefficient and the second stratified fluctuation coefficient of the second depth interval, the second prediction resource is configured to predict and obtain the third water quality parameter of the third depth interval, until the target water quality parameter of the target depth and the final prediction error coefficient are predicted, the water quality impact parameters are calculated, and the water quality environment monitoring results of the reservoir are obtained, including:

[0077] S41. Calculate the second compensation fluctuation coefficient based on the first prediction error coefficient and the second layered fluctuation coefficient of the second depth range.

[0078] S42. Based on the second compensation fluctuation coefficient and combined with T, the number of the second network is determined, and the water quality diffusion prediction in the third depth range is carried out to obtain the third water quality parameters in the third depth range.

[0079] S43. Continue water quality prediction until the target water quality parameters and the final prediction error coefficient at the target depth are obtained, where the target depth is the depth of the next depth interval among multiple depth intervals.

[0080] S44. Calculate the water quality impact parameters based on the target water quality parameters and the final prediction error coefficient, and obtain the water quality environment monitoring results of the reservoir.

[0081] In a preferred embodiment, the obtained first prediction error coefficient is first summed with the second layer fluctuation coefficient of the second depth range, i.e., second compensated fluctuation coefficient = first prediction error coefficient + second layer fluctuation coefficient. To ensure that the compensated fluctuation coefficient is within a reasonable range, the maximum value of the second compensated fluctuation coefficient is set to 1. When the summation result exceeds 1, the second compensated fluctuation coefficient is limited to 1. This compensation mechanism effectively integrates the accuracy information of the previous layer prediction and the water wave characteristics of the current layer, improving the adaptability of the prediction strategy through error feedback.

[0082] Subsequently, optimized prediction resources are configured based on the second compensation fluctuation coefficient. The second compensation fluctuation coefficient is multiplied by the total number of networks T in the water quality diffusion prediction network group, and the product is rounded down to determine the number of second networks. Water quality diffusion prediction networks corresponding to the number of second networks are randomly selected from the water quality diffusion prediction network group to form the second prediction resource. The second water quality parameters and the second compensation fluctuation coefficient for the second depth interval are used as input data and input to each water quality diffusion prediction network in the second prediction resource for parallel computation. The mean of all output results is calculated to obtain the predicted third water quality parameters for the third depth interval. Specifically, the predicted pollutant concentration values ​​output by all water quality diffusion prediction networks are averaged to obtain the predicted third water quality parameters for the third depth interval.

[0083] Following the layer-by-layer prediction model in steps S41 to S42, water quality prediction continues in deeper depth intervals. For each subsequent depth interval, a compensation fluctuation coefficient is calculated based on the prediction error coefficient of the previous layer and the stratified fluctuation coefficient of the current layer. Prediction resources are dynamically allocated, and integrated prediction calculations are performed. This cyclical prediction process continues until the water quality prediction for the depth interval containing the target depth is completed, obtaining the target water quality parameters and the final prediction error coefficient at the target depth. The target depth is the preset reservoir monitoring depth.

[0084] Subsequently, water quality impact parameters are assessed based on the target water quality parameters and the final prediction error coefficient. First, the ratio of the target water quality parameter to the rated water quality parameter corresponding to the target depth is calculated to obtain the basic water quality impact parameters. The higher the pollutant concentration, the larger the ratio, indicating a more severe water quality impact. Then, compensation calculations are performed on the basic water quality impact parameters based on the final prediction error coefficient: Water quality impact parameter = Basic water quality impact parameter × (1 + Final prediction error coefficient), yielding the comprehensive water quality impact parameters. These comprehensive water quality impact parameters reflect the degree of deviation of the deep water quality status of the reservoir from the rated water quality and the impact of prediction uncertainty, forming the water quality environmental monitoring results of the reservoir.

[0085] Through the aforementioned error compensation and progressive prediction mechanism, the limitations of traditional monitoring methods in deep water quality detection are effectively addressed, achieving high-precision full-depth water quality prediction and assessment, and providing technical support for reservoir environmental management.

[0086] Furthermore, water quality impact parameters are calculated based on the target water quality parameters and the final prediction error coefficient, including:

[0087] S441. Calculate the ratio of the target water quality parameter to the rated water quality parameter at the target depth to obtain the basic water quality influence parameter;

[0088] S442. Based on the final prediction error coefficient, calculate the basic water quality impact parameters to obtain the water quality impact parameters.

[0089] In a preferred embodiment, firstly, the rated water quality parameter corresponding to the target depth is obtained as the evaluation benchmark. The rated water quality parameter is the pollutant concentration limit for that target depth specified in the water quality standard, representing the standard water quality status of the depth range where the target depth is located. The ratio of the obtained target water quality parameter to the rated water quality parameter is calculated, i.e.: Basic water quality impact parameter = Target water quality parameter / Rated water quality parameter. The basic water quality impact parameter reflects the degree of deviation of the actual pollutant concentration from the standard limit: when the ratio is equal to 1, it indicates that the pollutant concentration meets the standard requirements; when the ratio is greater than 1, it indicates that the pollutant concentration exceeds the standard limit, and the larger the value, the more serious the water quality impact; when the ratio is less than 1, it indicates that the pollutant concentration is lower than the standard limit, and the water quality status is good. Through the above ratio calculation, the quantified basic water quality impact parameter is obtained.

[0090] Then, a prediction uncertainty compensation mechanism is introduced to improve the reliability of the water quality impact assessment. Considering that the target water quality parameters are obtained through layer-by-layer prediction, which introduces a certain prediction error, it is necessary to compensate and correct the basic water quality impact parameters based on the final prediction error coefficient. The compensation calculation formula is: Water quality impact parameter = Basic water quality impact parameter × (1 + Final prediction error coefficient). The design principle of this compensation mechanism is that the larger the final prediction error coefficient, the higher the prediction uncertainty. Therefore, a corresponding safety margin needs to be added to the basic impact parameters to ensure the conservatism and reliability of the water quality impact assessment. Through the above compensation calculation, the final water quality impact parameters, comprehensively considering the degree of pollution and prediction uncertainty, are obtained.

[0091] This water quality impact parameter comprehensively reflects the deviation of pollutant concentration levels at the target depth of the reservoir from standard limits, as well as the accumulated uncertainties during the prediction process, providing a quantitative basis for reservoir environmental management decisions. A higher water quality impact parameter indicates a higher risk to the water environment, necessitating corresponding pollution control and treatment measures.

[0092] The above technical solution effectively solves the technical problem in existing technologies where water quality sensors cannot cover the entire depth, leading to inaccurate water quality monitoring at target depths and high costs. Compared to traditional full-depth dense sensor deployment schemes, this technical solution, through temperature stratification fluctuation analysis and layer-by-layer prediction error compensation, requires only a small number of sensors in each depth range to achieve accurate monitoring and assessment of water quality at target depths. This reduces sensor deployment costs while ensuring monitoring accuracy and assessment reliability, providing an economical and efficient monitoring solution for reservoir environmental monitoring.

[0093] Example 2, as Figure 2 As shown, based on the same inventive concept as the reservoir environmental monitoring method provided in Embodiment 1, this embodiment of the invention also provides a reservoir environmental monitoring system, including:

[0094] Temperature stratification monitoring module 11 is used to test and acquire vertical temperature time series data in multiple depth intervals within the reservoir, obtain multiple interval temperature sequences, calculate multiple interval temperature change rates, and obtain multiple stratification fluctuation coefficients.

[0095] The water quality parameter acquisition module 12 is used to set multiple monitoring frequencies of water quality sensors configured in multiple depth ranges based on multiple stratified fluctuation coefficients, and to acquire the first water quality parameter of the first water quality sensor in the first depth range.

[0096] Prediction verification module 13 is used to configure first prediction resources based on the first stratified fluctuation coefficient of the first depth range, predict the second water quality parameters of the second depth range according to the first water quality parameters and the first stratified fluctuation coefficient, verify them with the second water quality parameters monitored by the second water quality sensor, and obtain the first prediction error coefficient.

[0097] The water quality assessment module 14 is used to configure the second prediction resources based on the first prediction error coefficient and the second stratified fluctuation coefficient of the second depth range, predict and obtain the third water quality parameters of the third depth range, until the target water quality parameters and the final prediction error coefficient of the target depth are predicted and obtained, calculate the water quality impact parameters, and obtain the water quality environment monitoring results of the reservoir.

[0098] Furthermore, the execution steps of the temperature stratification monitoring module 11 include:

[0099] The test acquires temperature data at multiple times within a preset time range across multiple depth intervals in the reservoir, obtaining vertical temperature time-series data.

[0100] The temperature data from multiple depth ranges are sorted according to time sequence to obtain multiple range temperature sequences.

[0101] Based on multiple temperature interval sequences, multiple interval temperature change rates are calculated, and multiple stratified fluctuation coefficients are obtained.

[0102] Furthermore, the execution steps of the temperature stratification monitoring module 11 also include:

[0103] Based on multiple temperature interval sequences, the average change rate of temperature data in multiple depth intervals per unit time is calculated to obtain the temperature change rate of multiple intervals.

[0104] The ratios of temperature change rates in multiple intervals to the historical maximum temperature change rate are calculated to obtain multiple stratified fluctuation coefficients.

[0105] Furthermore, the execution steps of the water quality parameter acquisition module 12 include:

[0106] The preset monitoring frequency for water quality monitoring by the water quality sensor is obtained. Based on multiple stratified fluctuation coefficients, the preset monitoring frequency is corrected and calculated to obtain multiple monitoring frequencies.

[0107] The monitoring frequency of water quality sensors configured in multiple depth ranges is set according to multiple monitoring frequencies.

[0108] The first water quality parameter is monitored and acquired by the first water quality sensor within the first depth range, wherein the first depth range is the depth range with the smallest depth.

[0109] Furthermore, the execution steps of the prediction verification module 13 include:

[0110] Obtain a water quality diffusion prediction network group, wherein the water quality diffusion prediction network group includes T water quality diffusion prediction networks, where T is a positive integer;

[0111] Based on the first layer fluctuation coefficient of the first depth range, the number of the first network is determined by combining T, and the water quality diffusion prediction network of the first network number is randomly selected as the first prediction resource.

[0112] The first water quality parameter and the first stratified fluctuation coefficient are input into the first number of water quality diffusion prediction networks within the first predicted resource, and the mean of the output results is calculated to obtain the predicted second water quality parameter for the two depth intervals.

[0113] Acquire the second water quality parameters within the second depth range monitored by the second water quality sensor;

[0114] Calculate the error range between the second water quality parameter and the predicted second water quality parameter to obtain the first prediction error coefficient.

[0115] Furthermore, the execution steps of the prediction verification module 13 also include:

[0116] Based on historical data of reservoir water quality monitoring, a set of sample water quality parameters and a set of sample stratified fluctuation coefficients are collected. Water quality parameters for the next depth interval of different sample water quality parameters and sample stratified fluctuation coefficients are also collected to obtain a set of sample predicted water quality parameters.

[0117] The sample water quality parameter set, the sample stratified fluctuation coefficient set, and the sample predicted water quality parameter set are randomly divided T times to obtain T sets of water quality diffusion training data.

[0118] Based on machine learning, T water quality diffusion prediction networks with the same architecture are constructed. Each network is iteratively supervised and trained using the T sets of water quality diffusion training data. After training and verification convergence, a group of water quality diffusion prediction networks is obtained.

[0119] Furthermore, the execution steps of water quality assessment module 14 include:

[0120] The second compensation fluctuation coefficient is calculated based on the first prediction error coefficient and the second layered fluctuation coefficient of the second depth range.

[0121] The number of the second network is determined by combining the second compensation fluctuation coefficient with T, and the water quality diffusion prediction in the third depth range is carried out to obtain the third water quality parameters in the third depth range.

[0122] Continue water quality prediction until the target water quality parameters and the final prediction error coefficient at the target depth are obtained, where the target depth is the depth of the next depth interval among multiple depth intervals;

[0123] Based on the target water quality parameters and the final prediction error coefficient, the water quality impact parameters are calculated, and the water quality environment monitoring results of the reservoir are obtained.

[0124] Furthermore, the execution steps of water quality assessment module 14 also include:

[0125] Calculate the ratio of the target water quality parameter to the rated water quality parameter at the target depth to obtain the basic water quality impact parameters;

[0126] Based on the final prediction error coefficient, the basic water quality impact parameters are calculated to obtain the water quality impact parameters.

[0127] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0128] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0129] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0130] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0131] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0132] Although preferred embodiments of the invention have been described, those skilled in the art, once they have learned the basic inventive concept, can make other changes and modifications to these embodiments.

[0133] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of this invention and its equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for monitoring the environment of a reservoir, characterized in that, The method includes: The test obtained vertical temperature time series data at multiple depth intervals within the reservoir, resulting in multiple interval temperature sequences. The test also calculated the temperature change rate for multiple intervals and obtained multiple stratified fluctuation coefficients. Based on multiple stratified fluctuation coefficients, multiple monitoring frequencies of water quality sensors configured in multiple depth ranges are set to obtain the first water quality parameters of the first water quality sensor in the first depth range. Based on the first layered fluctuation coefficient of the first depth range, configure the first prediction resource, predict the second water quality parameter of the second depth range according to the first water quality parameter and the first layered fluctuation coefficient, verify it with the second water quality parameter monitored by the second water quality sensor, and obtain the first prediction error coefficient. Based on the first prediction error coefficient and the second stratified fluctuation coefficient of the second depth range, the second prediction resource is configured to predict and obtain the third water quality parameter of the third depth range until the target water quality parameter and the final prediction error coefficient of the target depth are predicted and obtained. The water quality impact parameters are calculated and the water quality environment monitoring results of the reservoir are obtained.

2. The reservoir environmental monitoring method according to claim 1, characterized in that, The test acquired vertical temperature time-series data at multiple depth intervals within the reservoir, obtaining multiple interval temperature sequences, and calculated multiple interval temperature change rates, resulting in multiple stratified fluctuation coefficients, including: The test acquires temperature data at multiple times within a preset time range across multiple depth intervals in the reservoir, obtaining vertical temperature time-series data. The temperature data from multiple depth ranges are sorted according to time sequence to obtain multiple range temperature sequences. Based on multiple temperature interval sequences, multiple interval temperature change rates are calculated, and multiple stratified fluctuation coefficients are obtained.

3. The reservoir environmental monitoring method according to claim 2, characterized in that, Based on multiple temperature interval series, multiple interval temperature change rates are calculated, resulting in multiple stratified fluctuation coefficients, including: Based on multiple temperature interval sequences, the average change rate of temperature data in multiple depth intervals per unit time is calculated to obtain the temperature change rate of multiple intervals. The ratios of temperature change rates in multiple intervals to the historical maximum temperature change rate are calculated to obtain multiple stratified fluctuation coefficients.

4. The reservoir environmental monitoring method according to claim 1, characterized in that, Based on multiple stratified fluctuation coefficients, multiple monitoring frequencies are set for water quality sensors configured in multiple depth ranges to obtain the first water quality parameters of the first water quality sensor in the first depth range, including: The preset monitoring frequency for water quality monitoring by the water quality sensor is obtained. Based on multiple stratified fluctuation coefficients, the preset monitoring frequency is corrected and calculated to obtain multiple monitoring frequencies. The monitoring frequency of water quality sensors configured in multiple depth ranges is set according to multiple monitoring frequencies. The first water quality parameter is monitored and acquired by the first water quality sensor within the first depth range, wherein the first depth range is the depth range with the smallest depth.

5. The reservoir environmental monitoring method according to claim 1, characterized in that, Based on the first stratified fluctuation coefficient of the first depth range, a first prediction resource is configured. Based on the first water quality parameters and the first stratified fluctuation coefficient, a predicted second water quality parameter for the second depth range is obtained. This parameter is then verified against the second water quality parameter monitored by the second water quality sensor to obtain a first prediction error coefficient, including: Obtain a water quality diffusion prediction network group, wherein the water quality diffusion prediction network group includes T water quality diffusion prediction networks, where T is a positive integer; Based on the first layer fluctuation coefficient of the first depth range, the number of the first network is determined by combining T, and the water quality diffusion prediction network of the first network number is randomly selected as the first prediction resource. The first water quality parameter and the first stratified fluctuation coefficient are input into the first number of water quality diffusion prediction networks within the first predicted resource, and the mean of the output results is calculated to obtain the predicted second water quality parameter for the two depth intervals. Acquire the second water quality parameters within the second depth range monitored by the second water quality sensor; Calculate the error range between the second water quality parameter and the predicted second water quality parameter to obtain the first prediction error coefficient.

6. The reservoir environmental monitoring method according to claim 5, characterized in that, Obtain a network of water quality diffusion prediction systems, including: Based on historical data of reservoir water quality monitoring, a set of sample water quality parameters and a set of sample stratified fluctuation coefficients are collected. Water quality parameters for the next depth interval of different sample water quality parameters and sample stratified fluctuation coefficients are also collected to obtain a set of sample predicted water quality parameters. The sample water quality parameter set, the sample stratified fluctuation coefficient set, and the sample predicted water quality parameter set are randomly divided T times to obtain T sets of water quality diffusion training data. Based on machine learning, T water quality diffusion prediction networks with the same architecture are constructed. Each network is iteratively supervised and trained using the T sets of water quality diffusion training data. After training and verification convergence, a group of water quality diffusion prediction networks is obtained.

7. The reservoir environmental monitoring method according to claim 5, characterized in that, Based on the first prediction error coefficient and the second stratified fluctuation coefficient of the second depth interval, the second prediction resource is configured to predict and obtain the third water quality parameter of the third depth interval, until the target water quality parameter of the target depth and the final prediction error coefficient are predicted. Water quality impact parameters are then calculated to obtain the water quality environment monitoring results of the reservoir, including: The second compensation fluctuation coefficient is calculated based on the first prediction error coefficient and the second layered fluctuation coefficient of the second depth range. The number of the second network is determined by combining the second compensation fluctuation coefficient with T, and the water quality diffusion prediction in the third depth range is carried out to obtain the third water quality parameters in the third depth range. Continue water quality prediction until the target water quality parameters and the final prediction error coefficient at the target depth are obtained, where the target depth is the depth of the next depth interval among multiple depth intervals; Based on the target water quality parameters and the final prediction error coefficient, the water quality impact parameters are calculated, and the water quality environment monitoring results of the reservoir are obtained.

8. The reservoir environmental monitoring method according to claim 7, characterized in that, Water quality impact parameters are calculated based on the target water quality parameters and the final prediction error coefficient, including: Calculate the ratio of the target water quality parameter to the rated water quality parameter at the target depth to obtain the basic water quality impact parameters; Based on the final prediction error coefficient, the basic water quality impact parameters are calculated to obtain the water quality impact parameters.

9. A reservoir environmental monitoring system, characterized in that, For implementing the reservoir environmental monitoring method as described in any one of claims 1 to 8, the system comprises: The temperature stratification monitoring module is used to test and acquire vertical temperature time series data at multiple depth intervals within the reservoir, obtain multiple interval temperature sequences, calculate the temperature change rate of multiple intervals, and obtain multiple stratification fluctuation coefficients. The water quality parameter acquisition module is used to set multiple monitoring frequencies of water quality sensors configured in multiple depth ranges based on multiple stratified fluctuation coefficients, and to acquire the first water quality parameter of the first water quality sensor in the first depth range. The prediction verification module is used to configure the first prediction resource based on the first stratified fluctuation coefficient of the first depth range, predict the second water quality parameter of the second depth range according to the first water quality parameter and the first stratified fluctuation coefficient, verify it with the second water quality parameter monitored by the second water quality sensor, and obtain the first prediction error coefficient. The water quality assessment module is used to configure the second prediction resources based on the first prediction error coefficient and the second stratified fluctuation coefficient of the second depth range, predict and obtain the third water quality parameters of the third depth range, until the target water quality parameters and the final prediction error coefficient of the target depth are predicted and obtained, calculate the water quality impact parameters, and obtain the water quality environment monitoring results of the reservoir.

Images