Lake and reservoir algal bloom prediction and monitoring time window scheduling method, device and storage medium

Abstract

The application discloses a lake and reservoir algae bloom prediction and monitoring time window scheduling method, equipment and storage medium, comprising: based on observation data, using a data assimilation algorithm to perform online assimilation update on the state and parameters of the lake and reservoir digital twin; using the assimilated digital twin to predict the algae bloom in the future period, and performing physical constraint correction and uncertainty quantification processing on the prediction result to output prediction information; generating a detection time window candidate set according to the prediction information; based on the detection time window candidate set and a preset operation constraint, optimization solving is performed to obtain a detection time window scheduling strategy, which is issued to an unmanned detection platform, and the measured data collected after the execution of the base detection task is fed back to the observation data to perform data registration and model assimilation on the digital twin. Through the construction of a closed-loop system of perception-assimilation-prediction-scheduling-feedback, the application realizes dynamic optimization configuration of monitoring resources according to the algae bloom risk, and improves the early warning timeliness and resource utilization efficiency.

Description

Technical Field

[0001] This application relates to the field of lake and reservoir water environment management and cyanobacterial bloom control technology, and in particular to a method, equipment and storage medium for predicting and monitoring time windows of algal blooms in lakes and reservoirs. Background Technology

[0002] In the field of lake and reservoir water environment management and cyanobacterial bloom control, relevant technologies mostly rely on fixed-station / buoy timed monitoring, unmanned vessel inspections, statistical or numerical model predictions, and path planning and scheduling. However, these technical solutions have obvious limitations and are fragmented, making it difficult to meet the urgent need for efficient early warning and accurate monitoring of algal blooms.

[0003] Specifically, the shortcomings of this technical system are as follows: fixed-period monitoring cannot capture the rapid spatiotemporal evolution of algal blooms, leading to detection delays; while single-modal statistical models or offline numerical models, lacking physical constraints and real-time data assimilation, experience a sharp amplification of prediction errors when external conditions change abruptly, and error accumulation cannot be avoided. Furthermore, traditional methods often focus on spatial path planning, seriously neglecting the crucial factor of the "monitoring time window," resulting in high-risk periods not being prioritized for coverage and low utilization of monitoring resources; at the same time, model prediction results are disconnected from monitoring task execution, forming information silos where "prediction is prediction and scheduling is scheduling," leading to a general lack of real-time compensation for sensor data drift, quantification of the uncertainty of prediction results, and degradation and fault tolerance strategies under abnormal conditions, making it difficult to guarantee the data validity, scientific decision-making, and service continuity of the entire system.

[0004] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main purpose of this application is to provide a method, device and storage medium for scheduling time windows of lake and reservoir algal bloom prediction and monitoring, which aims to solve the technical problems of resource mismatch and response delay caused by the disconnect between static prediction and rigid monitoring scheduling of lake and reservoir algal blooms in the prior art.

[0006] To achieve the above objectives, this application proposes a method for scheduling time windows for predicting and monitoring algal blooms in lakes and reservoirs. The method includes:

[0007] Acquire multi-source detection data from multiple sensors in a lake or reservoir, and register the data errors of the multi-source detection data to obtain registered observation data;

[0008] The observation data is subjected to model data assimilation processing, and the model state and model parameters of the digital twin are updated online based on the data assimilation results. The digital twin is a virtual dynamic mapping model constructed based on the physical entities of the lake and reservoir.

[0009] Predict future algal bloom information based on the assimilated and updated digital twin, and generate algal bloom prediction results based on the algal bloom information. The algal bloom prediction results include algal bloom outbreak probability and risk distribution.

[0010] Based on the algal bloom probability and the risk distribution, a candidate set of detection time windows is generated for different lake and reservoir detection areas;

[0011] Based on the candidate set of detection time windows and the preset running constraints, an optimization solution is performed to obtain the detection time window scheduling strategy.

[0012] The detection time window scheduling strategy is sent to the unmanned detection platform to execute the detection task, and the measured data corresponding to the executed detection task is obtained. The measured data is then fed back to the observation data to assimilate and update the digital twin.

[0013] In one embodiment, the step of registering the data errors of the multi-source detection data to obtain registered observation data includes:

[0014] The multi-source detection data is subjected to time synchronization and spatial alignment operations to generate a spatiotemporally consistent fusion dataset;

[0015] The mean of the residuals is calculated based on the residuals between the fused dataset and the predicted values ​​of the digital twin;

[0016] When the mean residual exceeds a preset residual threshold, it is determined that there is drift data in the fused dataset, and the drift data is an abnormal value caused by sensor error or environmental interference.

[0017] Numerical compensation is performed on the drift data, and the fused dataset is updated based on the numerical compensation results;

[0018] The updated fused dataset is then subjected to outlier identification and filtering to obtain the observation data.

[0019] In one embodiment, the step of predicting algal bloom information for future periods based on the assimilated and updated digital twin, and generating algal bloom prediction results based on the algal bloom information, includes:

[0020] Based on the assimilated and updated model state and model parameters, the digital twin is driven to perform simulation calculations of algal bloom information for future time periods, and a preliminary algal bloom prediction sequence is obtained based on the simulation calculation results.

[0021] Physical constraints are introduced to correct the preliminary prediction sequence of algal blooms. The physical constraints include at least one of the laws of conservation of mass and conservation of energy.

[0022] Uncertainty quantification is performed on the corrected preliminary algal bloom prediction sequence to generate confidence information, which includes confidence intervals or probability distributions;

[0023] The corrected preliminary algal bloom prediction sequence is combined with the confidence information to generate the algal bloom prediction result.

[0024] In one embodiment, the step of driving the digital twin to perform future algal bloom information simulation calculations based on the assimilated and updated model state and model parameters, and obtaining a preliminary algal bloom prediction sequence based on the simulation calculation results, includes:

[0025] Using the model state and model parameters as initial conditions, and based on preset meteorological and hydrological forcing data for future time periods, the coupled hydrodynamic sub-model, biochemical sub-model, and photothermal sub-model in the digital twin are solved simultaneously.

[0026] The coupled calculations of the hydrodynamic sub-model, the biochemical sub-model, and the photothermal sub-model are performed cyclically according to a preset time step, and the spatiotemporal change sequence within the future time period is output.

[0027] The spatiotemporal variation sequence is used as the preliminary prediction sequence for algal blooms.

[0028] In one embodiment, the step of cyclically performing the coupled calculations of the hydrodynamic sub-model, the biochemical sub-model, and the photothermal sub-model at a preset time step, and outputting the spatiotemporal change sequence within a future time period, includes:

[0029] Within each time step, the hydrodynamic sub-model and the photothermal sub-model are respectively driven to calculate the flow field distribution and the vertical distribution of water temperature under illumination;

[0030] The flow field distribution and the vertical distribution of light and water temperature are used as inputs to the biochemical sub-model to calculate the spatiotemporal changes of algal biomass and nutrient concentration.

[0031] The calculation result of the current time step of the biochemical sub-model is used as the initial condition for the next time step. The simulation is repeated until the preset future time period is completed, and the continuous spatiotemporal change sequence is output.

[0032] In one embodiment, the step of generating a candidate set of detection time windows for different lake / reservoir detection areas based on the algal bloom probability and the risk distribution includes:

[0033] Based on the probability of algal blooms and the risk distribution, the lake and reservoir detection area is divided into multiple spatial units;

[0034] Based on the monitoring priority and preset monitoring frequency requirements of each space unit, the number and duration of time windows for each space unit in the future scheduling cycle are determined, and the monitoring priority is determined based on the probability of algal blooms.

[0035] Based on the geographical location of the space unit, the travel speed of the unmanned monitoring platform, and the preset operational constraints, the feasible start time range for each time window is calculated.

[0036] The spatial unit, time window duration, and feasible start time range are combined to generate the detection time window candidate set containing multiple candidate monitoring tasks.

[0037] In one embodiment, the step of optimizing the detection time window scheduling strategy based on the candidate set of detection time windows and preset running constraints includes:

[0038] Obtain the platform parameters of the unmanned detection platform, wherein the platform parameters include at least one of the following: endurance time, travel speed, and sensor type;

[0039] The platform parameters are converted into resource constraints and time conflict constraints of a preset optimization model;

[0040] Candidate time windows that satisfy the resource constraints and time conflict constraints are selected from the candidate time windows.

[0041] The candidate time windows are comprehensively evaluated based on the comprehensive evaluation function, and a combination of candidate time windows is obtained based on the evaluation results;

[0042] The detection time window scheduling strategy is generated based on the candidate time window combination.

[0043] In one embodiment, the step of feeding back the measured data to the observed data for assimilation and updating of the digital twin includes:

[0044] The measured data are marked as special data with feedback identifiers;

[0045] The specific data is input into the multi-source registration and quality control processing flow, and assimilated with the observation data to generate a feedback observation vector;

[0046] The model state and model parameters of the digital twin are updated online using the feedback observation vector.

[0047] In addition, to achieve the above objectives, this application also proposes a time window scheduling device for predicting and monitoring algal blooms in lakes and reservoirs. The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. The computer program is configured to implement the steps of the time window scheduling method for predicting and monitoring algal blooms in lakes and reservoirs as described above.

[0048] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of the lake and reservoir algal bloom prediction and monitoring time window scheduling method described above.

[0049] One or more technical solutions proposed in this application have at least the following technical effects:

[0050] The technical solution of this application involves acquiring multi-source detection data from multiple sensors in a lake or reservoir, registering the data errors of the multi-source detection data to obtain registered observation data, performing model data assimilation processing on the observation data, and updating the model state and parameters of a digital twin online based on the data assimilation results. The digital twin is a virtual dynamic mapping model constructed based on the physical entities of the lake or reservoir. Based on the assimilated and updated digital twin, the application predicts algal bloom information for future periods and generates algal bloom prediction results, including the probability and risk distribution of algal bloom outbreaks. Based on the algal bloom outbreak probability and the risk distribution, a candidate set of detection time windows is generated for different lake or reservoir detection areas. Based on the candidate set of detection time windows and preset operating constraints, an optimization solution is performed to obtain a detection time window scheduling strategy. The detection time window scheduling strategy is then distributed to an unmanned detection platform to execute detection tasks, and the measured data corresponding to the executed detection tasks is acquired. The measured data is then fed back to the observation data for assimilation and updating of the digital twin.

[0051] This application constructs a closed-loop system of perception-assimilation-prediction-scheduling-feedback, which enables dynamic optimization of monitoring resources according to the risk of algal blooms, thereby improving the timeliness of early warning and the efficiency of resource utilization. Attached Figure Description

[0052] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0053] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 This is a flowchart illustrating the first embodiment of the lake and reservoir algal bloom prediction and monitoring time window scheduling method of this application;

[0055] Figure 2 This is a detailed step diagram based on step S10 in the first embodiment;

[0056] Figure 3 This is a detailed step diagram based on step S30 in the first embodiment;

[0057] Figure 4 This is a detailed step diagram based on step S40 in the first embodiment;

[0058] Figure 5 This is a detailed step diagram based on step S50 in the first embodiment;

[0059] Figure 6 This is a detailed step diagram based on step S60 in the first embodiment;

[0060] Figure 7 A flowchart illustrating time-window-based scheduling;

[0061] Figure 8 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the lake and reservoir algal bloom prediction and monitoring time window scheduling method in the embodiments of this application.

[0062] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0063] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0064] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0065] The main solution of this application embodiment is as follows: First, acquire multi-source detection data from multiple sensors in a lake / reservoir. Second, register the data errors of the multi-source detection data to obtain registered observation data. Third, perform model data assimilation processing on the observation data, and update the model state and parameters of the digital twin online based on the assimilation results. The digital twin is a virtual dynamic mapping model constructed based on the physical entities of the lake / reservoir. Fourth, predict algal bloom information for future periods based on the assimilated and updated digital twin, and generate algal bloom prediction results based on the algal bloom information. The algal bloom prediction results include the algal bloom outbreak probability and risk distribution. Fifth, generate candidate sets of detection time windows for different lake / reservoir detection areas based on the algal bloom outbreak probability and the risk distribution. Sixth, optimize and solve the detection time window candidate sets and preset operating constraints to obtain a detection time window scheduling strategy. Seventh, distribute the detection time window scheduling strategy to an unmanned detection platform to execute detection tasks, acquire the measured data corresponding to the executed detection tasks, and feed the measured data back to the observation data for assimilation and updating of the digital twin.

[0066] In existing technologies, the disconnect between algal bloom prediction models and monitoring and scheduling systems leads to prediction lags because static models cannot adapt to the dynamic evolution of algal blooms. Fixed monitoring cycles make it difficult to capture sudden hotspots. Delays in human decision-making caused by the decoupling of prediction results from resource allocation, as well as insufficient system reliability due to the lack of uncertainty quantification and closed-loop correction mechanisms, ultimately result in delayed early warning responses and serious misallocation of monitoring resources.

[0067] This application provides a solution that achieves three core effects by constructing a closed-loop technical system encompassing "data perception, model assimilation, risk prediction, dynamic scheduling, and feedback feedback": First, by utilizing online assimilation and physical constraint correction using digital twins, the accuracy and timeliness of algal bloom prediction are significantly improved; second, by generating and optimizing detection time windows based on prediction information, adaptive focusing of monitoring resources on high-risk spatiotemporal units is achieved, greatly improving resource utilization efficiency; finally, through uncertainty quantification and data feedback mechanisms, the robustness and adaptability of the system are enhanced, forming a complete technical closed loop from accurate perception to intelligent execution.

[0068] Based on this, embodiments of this application provide a method for scheduling time windows for predicting and monitoring algal blooms in lakes and reservoirs, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the lake / reservoir algal bloom prediction and monitoring time window scheduling method of this application. In this embodiment, the lake / reservoir algal bloom prediction and monitoring time window scheduling method includes steps S10~S60:

[0069] Step S10: Obtain multi-source detection data detected by multiple sensors in the lake / reservoir, and register the data errors of the multi-source detection data to obtain registered observation data;

[0070] In this embodiment, multi-source detection data from multiple sensors in the lake / reservoir are acquired, and the data errors of the multi-source detection data are registered to obtain registered observation data. The multi-source detection data includes real-time or periodic data from water quality sensors (such as YSI EXO2 measuring chlorophyll a and dissolved oxygen), hydro-meteorological sensors (such as RM Young 05103 anemometer measuring wind speed and Onset HOBO photonics sensor measuring photosynthetically active radiation), and offline nutrient detection equipment (such as Skalar SAN++ continuous flow analyzer measuring total nitrogen and total phosphorus). The data processing logic begins in the data acquisition phase, where the sensors operate according to a preset sampling cycle: once per minute in the core monitoring area, once every five minutes in the regular monitoring area, and once every ten minutes in the background area, ensuring that the data covers the spatiotemporal variability of the lake / reservoir. The multi-source detection data is transmitted via 4G / 5G or satellite links, with transmission latency controlled within 10 seconds (4G / 5G) or within 30 seconds (satellite) to ensure data timeliness.

[0071] Subsequently, the multi-source detection data enters the data error registration stage, which includes time synchronization, spatial alignment, drift compensation, and outlier removal. In time synchronization, the National Time Service Center's NTP server is used for time synchronization, calibrated every 10 seconds to ensure that the timestamp error of all sensors does not exceed 1 second, forming a unified time reference. Spatial alignment uses GNSS positioning data to map sensor coordinates to a standardized grid system of the lake and reservoir (e.g., a 500m×500m grid in the core area and a 1000m×1000m grid in the regular area), achieving spatial consistency of multi-source data. Drift compensation is based on the residual sliding window method, using nearly 72 hours of effective data (quality flag qflag is 0) to construct a sliding window with a window size of 30 data points, calculating the residual between sensor observations and model predictions, estimating the drift amount in real time (e.g., the linear drift rate of the chlorophyll a sensor is 0.02μg / L / h), and performing numerical compensation to update the data. Outlier removal uses the 3σ criterion, marking observations exceeding the range of "mean ± 3 times standard deviation" as outliers (qflag is 3) and excluding them from subsequent processing. Finally, the registered observation data is output as a spatiotemporally consistent fusion dataset for subsequent model assimilation and prediction.

[0072] The data error registration process relies on edge computing units (such as NVIDIA Jetson AGX Xavier) to ensure low-latency processing. Quality control of the multi-source detection data also includes data format standardization, such as unifying chlorophyll a concentration units to μg / L with one decimal place, dissolved oxygen concentration units to mg / L, and recording a quality flag qflag to indicate data status. Through this series of processes, the registered observation data possesses high consistency and reliability, providing a foundation for online updates of the digital twin.

[0073] Step S20: The observation data is subjected to model data assimilation processing, and the model state and model parameters of the digital twin are updated online based on the data assimilation results. The digital twin is a virtual dynamic mapping model constructed based on the physical entities of the lake and reservoir.

[0074] In this embodiment, a data assimilation algorithm is used to calculate the registered observation data. Based on the calculation results, the state and parameters of the digital twin are updated online. This process aims to reduce the discrepancy between the model and reality through a data-driven approach. The digital twin is defined as a dynamic mapping model of the lake / reservoir physical entity in virtual space. Its core data structure adopts a layered architecture: the bottom layer is a numerical model group coupling hydrodynamic, biochemical, and photothermal multiphysics fields; the middle layer is a real-time synchronized spatiotemporal state database; and the top layer is a model-data interaction interface. Specifically, the digital twin includes a three-dimensional gridded field of state variables (velocity vector, chlorophyll a concentration, nutrient concentration, water temperature profile) and a parameter set (maximum phytoplankton growth rate 0.5-1.2d). - ¹, Diffusion coefficient 10-50 m² / s, Attenuation coefficient 0.2-1.0 m - ¹), all data elements are organized through a unified spatiotemporal index, supporting millisecond-level status query and update operations.

[0075] The data assimilation algorithm employs either Ensemble Kalman Filtering (EnKF) or Four-Dimensional Variational Calculation (4D-Var) methods, using the registered observation data as input to adjust the state variables and model parameters of the digital twin in real time. The data processing logic, based on Bayesian estimation principles, fuses observation data with model predictions to generate optimal state and parameter estimates. Furthermore, the assimilation period Δt_a is set according to the priority of the monitoring area: 30 minutes for the core monitoring area and 60 minutes for the regular and background monitoring areas, balancing computational efficiency and model accuracy. The number of sets N_e is preferably between 20 and 30; for example, 20 sets are used for core area assimilation, and 30 sets are used for the entire lake / reservoir assimilation, ensuring that the EnKF algorithm maintains high assimilation accuracy while keeping computation time below 10 seconds. The observation error covariance matrix R is adaptively estimated using a sliding window method, calculating the variance using the residuals (i.e., the difference between observed values ​​and model predictions) of the past 10 assimilation periods, and dynamically adjusting the R value based on residual changes. For example, when the residuals increase, the R value can be increased to increase the observation weights and enhance the assimilation robustness.

[0076] In each assimilation cycle, the data assimilation algorithm first initializes the ensemble state, generating multiple ensemble members based on the current model state and parameters. Then, the registered observation data (such as chlorophyll a concentration, water temperature, and wind speed) is compared with the predicted values ​​of each ensemble member. The Kalman gain is calculated through the EnKF update step, and the state and parameters are adjusted. The state update of the digital twin employs an incremental write mechanism, modifying only the changed data blocks to ensure efficient memory management. For example, for the chlorophyll a concentration field, assimilation can reduce the root mean square error of the prediction to below 7 μg / L. The updated state and parameters are fed back to each computing module through the digital twin's data bus interface for the next round of simulation prediction. The online assimilation update is executed on a cloud server (such as Alibaba Cloud ECS) or edge computing unit, ensuring efficient processing of large-scale data through a distributed computing architecture. Through this process, the digital twin can continuously adapt to the dynamic changes in the lake / reservoir environment, and its built-in version management mechanism can record the state evolution history, supporting retrospective analysis of model performance.

[0077] Step S30: Based on the assimilated and updated digital twin, predict algal bloom information for future periods, and generate algal bloom prediction results based on the algal bloom information. The algal bloom prediction results include algal bloom probability and risk distribution.

[0078] In this embodiment, the observed data undergoes model data assimilation processing, and the model state and parameters of the digital twin are updated online based on the assimilation results. The digital twin is a virtual dynamic mapping model constructed based on the physical entities of the lake and reservoir. The core data structure of the digital twin adopts a layered architecture: the bottom layer is a numerical model group coupling hydrodynamic, biochemical, and photothermal multiphysics fields; the middle layer is a real-time synchronized spatiotemporal state database; and the top layer is a model-data interaction interface. Specifically, the digital twin includes a three-dimensional gridded state variable field (velocity vector, chlorophyll a concentration, nutrient concentration, water temperature profile) and a parameter set (maximum phytoplankton growth rate 0.5-1.2d). - ¹, Diffusion coefficient 10-50 m² / s, Attenuation coefficient 0.2-1.0 m - ¹), all data elements are organized through a unified spatiotemporal index, supporting millisecond-level status query and update operations.

[0079] The model data assimilation process employs either Ensemble Kalman Filtering (EnKF) or Four-Dimensional Variational Calculation (4D-Var) methods, using the registered observation data as input to adjust the state variables and model parameters of the digital twin in real time. The data processing logic is based on Bayesian estimation principles, fusing observation data with model predictions to generate optimal state and parameter estimates. Furthermore, the assimilation period Δt_a is set according to the priority of the monitoring area: 30 minutes for the core monitoring area and 60 minutes for the regular and background monitoring areas, balancing computational efficiency and model accuracy. The number of sets N_e is preferably between 20 and 30; for example, 20 sets are used for core area assimilation, and 30 sets are used for the entire lake / reservoir assimilation, ensuring that the EnKF algorithm maintains high assimilation accuracy while keeping computation time below 10 seconds. The observation error covariance matrix R is adaptively estimated using a sliding window method, calculating the variance using the residuals (i.e., the difference between observed values ​​and model predictions) of the past 10 assimilation periods, and dynamically adjusting the R value based on residual changes. For example, when the residuals increase, the R value can be increased to increase the observation weights and enhance the assimilation robustness.

[0080] In each assimilation cycle, the model data assimilation process first initializes the ensemble state, generating multiple ensemble members based on the current model state and parameters. Then, the registered observation data (such as chlorophyll a concentration, water temperature, and wind speed) is compared with the predicted values ​​of each ensemble member. The Kalman gain is calculated using the EnKF update step, and the state and parameters are adjusted. The state update of the digital twin employs an incremental write mechanism, modifying only the changed data blocks to ensure efficient memory management. For example, for the chlorophyll a concentration field, assimilation can reduce the root mean square error of the prediction to below 7 μg / L. The updated state and parameters based on the data assimilation results are fed back to each computing module through the digital twin's data bus interface for the next round of simulation prediction. The online assimilation update is executed on a cloud server (such as Alibaba Cloud ECS) or edge computing unit, ensuring efficient processing of large-scale data through a distributed computing architecture. Through this process, the digital twin can continuously adapt to the dynamic changes in the lake / reservoir environment, and its built-in version management mechanism can record the state evolution history, supporting retrospective analysis of model performance.

[0081] Step S40: Based on the algal bloom probability and the risk distribution, generate a candidate set of detection time windows for different lake and reservoir detection areas;

[0082] In this embodiment, based on the probability of algal bloom and the risk distribution, a candidate set of detection time windows is generated for different lake and reservoir detection areas. Specifically, this data processing logic divides the lake and reservoir into multiple spatial units based on the probability of algal bloom and the risk distribution, and calculates the number, duration and feasible start time of time windows for each unit to form a candidate set of detection time windows.

[0083] Specifically, the lake / reservoir monitoring area is divided into grids: the core monitoring area uses a 500m×500m grid, and the regular monitoring area uses a 1000m×1000m grid to ensure the precision of spatial coverage. For each spatial unit, based on the monitoring priority determined by the algal bloom probability and risk distribution (based on the bloom probability and confidence level) and the preset monitoring frequency requirements, the number and duration of time windows within a future scheduling cycle (e.g., 6-72 hours) are determined. For example, high-risk units (bloom probability ≥70%) are set with multiple short time windows (e.g., 10 minutes) to capture rapid changes; low-risk units have longer time windows (e.g., 30 minutes). The feasible start time range is calculated based on the unit's geographical location, the unmanned monitoring platform's travel speed (e.g., maximum speed 2.5m / s), and preset operational constraints (e.g., no-navigation periods) to ensure that the time windows are feasible in both time and space.

[0084] In the specific implementation process, candidate monitoring task items are generated. Each candidate monitoring task item includes a spatial unit identifier, a time window start time, an end time, a duration, and a priority. Then, all candidate monitoring task items are combined into a candidate set of monitoring time windows, which serves as input for optimization. For example, for the nearshore area of ​​Taihu Lake, the candidate set may contain hundreds of time window options, covering different risks and confidence levels. Through this process, the candidate set of monitoring time windows achieves preliminary structuring of monitoring tasks, providing a foundation for subsequent optimized scheduling and ensuring dynamic matching of resource allocation with algal bloom risk.

[0085] Step S50: Based on the candidate set of detection time windows and the preset running constraints, perform optimization to obtain the detection time window scheduling strategy;

[0086] In this embodiment, the optimization process based on the candidate set of detection time windows and preset operating constraints yields the detection time window scheduling strategy, focusing on selecting and combining optimal time windows under constraints. Specifically, using the candidate set of detection time windows as input, a mathematical programming method is used to solve the problem and output a scheduling strategy to maximize monitoring efficiency and minimize resource consumption.

[0087] Based on this optimization solution, including constraint set assembly, cost function definition, and mixed-integer linear programming (MILP) solution, the preset operational constraints are derived from unmanned monitoring platform parameters, such as power constraints (total energy consumption ≤ 1600Wh), flight time constraints (total flight time ≤ 8 hours), no-navigation constraints (e.g., no-navigation in aquaculture areas from 6:00 to 18:00 daily), and service level requirements. These constraints are transformed into mathematical forms, such as linear inequalities, to limit the feasible solution space. The cost function J is defined as a weighted sum, including risk coverage weight w1 = 0.4, energy consumption weight w2 = 0.2, delay weight w3 = 0.15, and prediction instability penalty weight w4 = 0.15. The risk coverage is calculated based on the predicted outbreak probability and confidence level, and the energy consumption is estimated based on the platform's flight power consumption model.

[0088] Furthermore, the MILP optimizer is executed on the edge computing unit, with a solution time limit of no more than 5 seconds. The optimizer first filters candidate time windows that satisfy all constraints from the candidate set, then calculates the cost function value of each candidate window, and finally selects the combination of time windows that minimizes the total cost as the scheduling strategy. For example, the output includes a set of time windows W, where each element contains the monitoring unit u_i, start time t_start, end time t_end, priority prio, confidence conf, and energy budget. If no feasible solution is found, a degradation strategy is triggered, such as adjusting the time window duration or prioritizing the coverage of high-risk units. The detection time window scheduling strategy ensures the optimal allocation of monitoring tasks in terms of timing and resources, supporting efficient execution of unmanned platforms.

[0089] Step S60: The detection time window scheduling strategy is sent to the unmanned detection platform to execute the detection task, and the measured data corresponding to the executed detection task is obtained. The measured data is fed back to the observation data for assimilation and updating of the digital twin.

[0090] In this embodiment, the detection time window scheduling strategy is issued to the unmanned detection platform to execute the detection task, and the measured data corresponding to the executed detection task is obtained. The measured data is then fed back to the observation data for assimilation and updating of the digital twin. The processing begins with the issuance of the scheduling strategy. The unmanned detection platform (such as the BlueBoat unmanned vessel) executes the monitoring task according to the time window set W, collects the measured data, and then injects this data into the upstream processing flow through a feedback mechanism.

[0091] Based on the above processing, including task assignment, data transmission, feedback marking, and assimilation updates, the detection time window scheduling strategy is transmitted to the unmanned platform via a wireless communication link (such as 4G / 5G). The platform autonomously navigates and samples according to the start and end times, priorities, and energy consumption budget of the time window. After execution, the measured data (such as chlorophyll a concentration and water temperature) is transmitted back to the edge computing unit and marked as special data with feedback identifiers (quality flag qflag is set to 4) to distinguish it from regular observation data. The special data first undergoes timeliness verification and format standardization to ensure consistency with existing data, and then is fed back to the observation data to participate in time synchronization, spatial alignment, and drift compensation processing.

[0092] During the assimilation and update phase, the specialized data is input to the model data assimilation process, used to update the model state and parameters of the digital twin online. For example, in EnKF assimilation, feedback data increases the number of observed samples, improving assimilation accuracy. Simultaneously, the assimilation and update mechanism periodically (e.g., every 24 hours) updates the residual network parameters of the prediction module based on the feedback data, with a learning rate set to 0.001, adapting to data changes through a sliding window method. The process also includes fault-tolerant handling, such as enabling local caching for offline feedback when the link is interrupted. Through this closed-loop process, the feedback of the measured data enables continuous optimization of the system, enhancing the accuracy and robustness of the digital twin.

[0093] In summary, by constructing a closed-loop technology system encompassing "data perception, model assimilation, risk prediction, dynamic scheduling, and feedback updates," three core effects were achieved: First, the online assimilation and physical constraint correction using digital twins significantly improved the accuracy and timeliness of algal bloom prediction; second, by generating and optimizing detection time windows based on prediction information, adaptive focusing of monitoring resources on high-risk spatiotemporal units was achieved, greatly improving resource utilization efficiency; and finally, through uncertainty quantification and data feedback mechanisms, the robustness and adaptability of the system were enhanced, forming a complete technical closed loop from precise perception to intelligent execution.

[0094] Furthermore, you can also view Figure 2 , Figure 2 This is a detailed step diagram based on step S10 in the first embodiment. Figure 2 The steps for registering the data errors of the multi-source detection data to obtain registered observation data include: steps S11-15:

[0095] Step S11: Perform time synchronization and spatial alignment operations on the multi-source detection data to generate a spatiotemporally consistent fusion dataset;

[0096] Step S12: Calculate the mean of the residuals based on the residuals between the fused dataset and the predicted values ​​of the digital twin;

[0097] Step S13: When the mean residual exceeds a preset residual threshold, it is determined that there is drift data in the fused dataset. The drift data is an abnormal value caused by sensor error or environmental interference.

[0098] Step S14: Perform numerical compensation on the drift data, and update the fused dataset based on the numerical compensation result;

[0099] Step S15: Perform outlier identification and filtering on the updated fused dataset to obtain the observation data.

[0100] In this embodiment, time synchronization and spatial alignment operations are performed on multi-source detection data to form a spatiotemporally consistent fused dataset based on the operation results. Specifically, the time synchronization operation uses NTP service provided by a professional time service center to unify the timestamps of each sensor to the same time reference through periodic time service signals, ensuring that the time error of data from different sources does not exceed 1 second. The spatial alignment operation is based on GNSS positioning data, mapping the spatial coordinates of each sensor to a standardized lake / reservoir grid coordinate system, where the core monitoring area uses a 500m × 500m grid and the regular monitoring area uses a 1000m × 1000m grid. This mapping process involves coordinate transformation and spatial interpolation algorithms to ensure that all observation data form the fused dataset within a unified spatiotemporal framework. The fused dataset retains the numerical characteristics of the original observations and possesses spatiotemporal consistency, providing standardized input for subsequent quality control.

[0101] Based on the residuals between the fused dataset and the predicted values ​​of the digital twin, the mean residual is calculated. During implementation, a sliding time window method is used to identify drift phenomena. Specifically, a 72-hour time window is used to calculate the residual sequence between the observed values ​​and the model's predicted values ​​within the window. When the mean residual exceeds a preset residual threshold, drift data is determined to exist in the fused dataset. This drift data is anomaly values ​​caused by sensor errors or environmental interference. For example, the preset residual threshold for a chlorophyll a sensor is 0.5 μg / L; when the mean residual continuously exceeds this threshold, drift data is determined to exist.

[0102] As shown above, numerical compensation is performed on the drift data, and the fused dataset is updated based on the numerical compensation results. For determined drift data, a linear compensation model is used for numerical correction. This model calculates the compensation amount based on historical drift trends, for example, performing reverse compensation on the chlorophyll a sensor at a rate of 0.02 μg / L / hour. The updated fused dataset not only corrects systematic errors but also retains compensation records for subsequent quality assessment.

[0103] The updated fused dataset undergoes outlier identification and screening to obtain the observation data. The outlier identification and screening process employs the 3σ criterion based on statistical distribution. First, the mean and standard deviation of each monitoring parameter under the same spatiotemporal conditions are calculated. Then, data points deviating from the mean by more than three times the standard deviation are marked as outliers. Identified outliers are assigned a specific quality flag (qflag=3) and excluded from the dataset. The data after screening forms the observation data, which possesses complete quality metadata, including timestamps, spatial coordinates, numerical records, and quality flags. This observation data, as a standardized output, is directly input into the subsequent data assimilation process, providing high-quality observation input for updating the digital twin. The entire data error registration process operates in a closed loop on the edge computing unit, with status information transmitted between steps via quality flags to ensure the traceability of the data processing trajectory. Through this hierarchical processing mechanism, the system can maintain data quality consistency in complex lake and reservoir monitoring environments.

[0104] Furthermore, you can also view Figure 3 , Figure 3 This is a detailed step diagram based on step S30 in the first embodiment. Figure 3 The step of predicting algal bloom information for future periods based on the assimilated and updated digital twin, and generating algal bloom prediction results based on the algal bloom information, includes steps S31-34:

[0105] Step S31: Based on the assimilated and updated model state and model parameters, drive the digital twin to perform future algal bloom information simulation calculations, and obtain a preliminary algal bloom prediction sequence based on the simulation calculation results;

[0106] Step S32: Introduce physical constraints to correct the preliminary prediction sequence of algal blooms. The physical constraints include at least one of the laws of conservation of mass and conservation of energy.

[0107] Step S33: Quantize the uncertainty of the corrected preliminary algal bloom prediction sequence to generate confidence information, which includes confidence intervals or probability distributions.

[0108] Step S34: The corrected preliminary algal bloom prediction sequence is combined with the confidence information to generate the algal bloom prediction result.

[0109] In this embodiment, the digital twin is driven to perform simulation calculations of algal bloom information for future time periods based on the assimilated and updated model state and model parameters, and a preliminary algal bloom prediction sequence is obtained based on the simulation results. Specifically, the model state includes the flow velocity distribution of the hydrodynamic field, the chlorophyll a concentration profile of the biochemical field, and the vertical temperature gradient of the photothermal field; the model parameters cover key ecological parameters such as the maximum growth rate of phytoplankton and the nutrient half-saturation constant. Based on the above initial ecological parameter conditions, the digital twin enters the forward simulation stage, and the calculation is advanced according to a preset time step by coupling and solving the hydrodynamic sub-model, the biochemical sub-model, and the photothermal sub-model. In specific implementation, a 15-minute simulation step is used as the basic simulation step size, and each sub-model is solved sequentially in each time step: the hydrodynamic sub-model calculates the convection and diffusion effects of the flow field, the photothermal model simulates radiation transfer and thermodynamic processes, and the biochemical model describes the algal growth and nutrient cycle dynamics. The simulation continues until the preset future time period ends (6-72 hours), and outputs the preliminary prediction sequence of the key algal bloom parameters. This sequence records the evolution trajectory of parameters such as chlorophyll a concentration in the form of a spatiotemporal matrix.

[0110] Physical constraints are introduced to correct the preliminary prediction sequence of algal blooms. These physical constraints include at least one of the laws of conservation of mass and energy. In the process of improving the reliability and interpretability of the prediction results through physical constraint correction and uncertainty quantification, the physical constraint correction employs a physical information machine learning framework to construct a lightweight residual network to correct biases in the preliminary prediction sequence. This lightweight residual network uses the prediction residuals from the previous time step as input features and outputs the correction amount for the current time step. Physical constraints such as the non-negativity of chlorophyll a concentration and the conservation of nutrient mass are embedded in the network structure to ensure that the corrected prediction values ​​conform to the inherent laws of the ecosystem. In the uncertainty quantification stage, based on 20 ensemble members of an ensemble Kalman filter, the confidence information is calculated by statistically analyzing the output distribution of each member during the prediction period. Specifically, this includes the 5% and 95th percentile values ​​of chlorophyll a concentration, forming the confidence interval of the prediction value. Simultaneously, the prediction confidence index for each spatial unit is calculated to provide a quantitative basis for risk assessment.

[0111] Uncertainty quantification is performed on the corrected preliminary algal bloom prediction sequence to generate confidence information, which includes confidence intervals or probability distributions. In the uncertainty quantification stage, based on the 20 ensemble members of an ensemble Kalman filter, the confidence information is calculated by statistically analyzing the output distribution of each member during the prediction period. Specifically, this includes the 5% and 95th percentile values ​​of chlorophyll a concentration, forming confidence intervals for the predicted values. Simultaneously, the prediction confidence index for each spatial unit is calculated to provide a quantitative basis for risk assessment.

[0112] The corrected preliminary algal bloom prediction sequence is combined with the confidence information to generate the algal bloom prediction result. In this process, the chlorophyll a concentration field, corrected for physical constraints, is spatiotemporally matched with the corresponding confidence interval to generate a structured prediction dataset. The output includes a spatiotemporal distribution map of key algal bloom parameters for the future period, an outbreak probability assessment for each monitoring unit, and a description of prediction uncertainty. The algal bloom prediction result is transmitted to the downstream processing module in a standardized data format. The outbreak probability is calculated based on a comparison between the corrected chlorophyll a concentration and historical thresholds, while the confidence information is directly derived from the uncertainty quantification result. The final output prediction information provides a comprehensive risk assessment for time window scheduling decisions, ensuring the accuracy and adaptability of monitoring resource allocation.

[0113] Based on the above Figure 3 The content of step S31 is further refined, namely, the step of driving the digital twin to perform future algal bloom information simulation calculation based on the assimilated and updated model state and model parameters, and obtaining the preliminary algal bloom prediction sequence based on the simulation calculation results, includes steps S31-1 to S31-31-3:

[0114] Step S31-1: Using the model state and model parameters as initial conditions, based on the preset meteorological and hydrological forcing data for future time periods, simultaneously solve the coupled hydrodynamic sub-model, biochemical sub-model and photothermal sub-model in the digital twin;

[0115] Step S32-2: Execute the coupled calculations of the hydrodynamic sub-model, the biochemical sub-model, and the photothermal sub-model in a cyclical manner according to a preset time step, and output the spatiotemporal change sequence within the future time period;

[0116] Step S33-3: Use the spatiotemporal change sequence as the preliminary prediction sequence for algal blooms.

[0117] In this embodiment, the digital twin is driven to simulate data for future periods using the assimilated and updated model state and model parameters, thereby obtaining a preliminary prediction sequence of algal blooms, which provides basic data support for subsequent correction and uncertainty quantification.

[0118] Using the model state and model parameters as initial conditions, and based on preset future meteorological and hydrological forcing data, the coupled hydrodynamic sub-model, biochemical sub-model, and photothermal sub-model in the digital twin are solved simultaneously. Specifically, using the model state and model parameters as initial conditions, a complete state description of the digital twin is obtained from the latest assimilation cycle. The model state includes the three-dimensional velocity distribution of the hydrodynamic field, the nutrient concentration profile of the biochemical field, and the vertical temperature gradient of the photothermal field; the model parameters cover key ecological parameters such as the maximum phytoplankton growth rate and the nutrient half-saturation constant. Based on these key ecological parameters as initial conditions, and combined with preset future meteorological and hydrological forcing data, the coupled hydrodynamic sub-model, biochemical sub-model, and photothermal sub-model in the digital twin are solved simultaneously. The meteorological and hydrological forcing data includes the wind speed field, photosynthetically active radiation intensity, and inflow and outflow boundary conditions for the next 72 hours. These data are obtained through numerical weather prediction models and spatially interpolated. During the synchronous solution process, the operator splitting method is used to handle the coupling relationship between the sub-models. First, the Navier-Stokes equations of the hydrodynamic sub-model are solved to obtain the flow field distribution. Then, the radiative transfer equations of the photothermal sub-model are solved to obtain the vertical distribution of water temperature. Finally, the calculation results are used as source terms to input into the biochemical sub-model to complete the single-step coupled calculation.

[0119] Furthermore, the coupled calculations of the hydrodynamic sub-model, the biochemical sub-model, and the photothermal sub-model are executed cyclically according to a preset time step, and the spatiotemporal variation sequence within the future time period is output. Specifically, the preset time step is set to 15 minutes according to the simulation accuracy requirements, and the solution sequence of the three sub-models is executed sequentially within each calculation step. In the specific implementation process, the hydrodynamic sub-model is first driven to calculate the velocity field and mixing layer depth at the current moment, and the governing equations are discretized using the finite volume method, with a horizontal diffusion coefficient of 30 m² / s and a vertical diffusion coefficient of 10. -4 m² / s. Subsequently, the photothermal model calculates the attenuation process of photosynthetically active radiation in the water body based on the radiative transfer equation, with the attenuation coefficient ranging from 0.2 to 1.0 m² / s based on real-time turbidity data. - ¹Dynamically adjusted within a certain range. Finally, the hydrodynamic and photothermal calculation results are used as inputs to the biochemical sub-model to solve the differential equations of the nutrient-phytoplankton-dissolved oxygen three-compartment model, where the maximum phytoplankton growth rate is set to 0.8 days. - ¹, the nutrient uptake half-saturation constants TN and TP are 0.03 mg / L. This iterative calculation process continuously iterates, using the calculation results of the previous time step as the initial conditions for the next time step, until the simulation of the preset future time period is completed, and the spatiotemporal change sequence of the future time period is output.

[0120] Furthermore, the spatiotemporal variation sequence is used as the preliminary prediction sequence for algal blooms. Specifically, the preliminary prediction sequence records the spatiotemporal evolution of chlorophyll a concentration, algal biomass, and nutrient concentration over the next 6-72 hours in the form of a multidimensional array. The spatial dimension corresponds to the lake / reservoir monitoring grid (500m×500m in the core area and 1000m×1000m in the regular area), and the temporal dimension is at 15-minute intervals. The preliminary prediction sequence also includes complete metadata information, including simulation start time, spatial grid parameters, and model version identifier, ensuring data traceability. The output data format adopts the NetCDF standard, supporting efficient storage and fast retrieval, serving as a complete input dataset for subsequent processing.

[0121] Specifically, the step of cyclically executing the coupled calculations of the hydrodynamic sub-model, the biochemical sub-model, and the photothermal sub-model at a preset time step, and outputting the spatiotemporal change sequence within the future time period includes:

[0122] Within each time step, the hydrodynamic sub-model and the photothermal sub-model are respectively driven to calculate the flow field distribution and the vertical distribution of water temperature under illumination;

[0123] The flow field distribution and the vertical distribution of light and water temperature are used as inputs to the biochemical sub-model to calculate the spatiotemporal changes of algal biomass and nutrient concentration.

[0124] The calculation result of the current time step of the biochemical sub-model is used as the initial condition for the next time step. The simulation is repeated until the preset future time period is completed, and the continuous spatiotemporal change sequence is output.

[0125] During the process of performing coupled calculations of the hydrodynamic sub-model, biochemical sub-model, and photothermal sub-model based on a preset time step, a complete model coupling mechanism needs to be established within the framework of the digital twin to ensure the coordination and unity of each physical and biochemical process in the spatiotemporal dimensions.

[0126] At the beginning of each time step, the system drives the hydrodynamic sub-model and the photothermal sub-model in parallel for calculation. The hydrodynamic sub-model is constructed based on the shallow water equations and spatially discretized using the finite volume method. Within each time step, the momentum conservation equation and the continuity equation are solved to calculate the flow field distribution, including the horizontal velocity field and the vertical mixing coefficient. Based on the model parameters, the horizontal diffusion coefficient is set to 30 m² / s, and the wind stress coefficient is dynamically adjusted according to wind speed conditions: when the wind speed at a height of 10 meters is less than 10 m / s, it is taken as 1.3 × 10⁻⁶ m² / s. - ³, when the speed is higher than 10 m / s, take 1.5 × 10. -³. Simultaneously, the photothermal sub-model, based on radiative transfer theory, calculates the attenuation process of photosynthetically active radiation in water and solves the heat conduction equation to obtain the vertical distribution of the illuminated water temperature. This model employs a two-flow approximation algorithm to handle radiative transfer, with the attenuation coefficient ranging from 0.2 to 1.0 m based on real-time turbidity monitoring data. - ¹The solar radiation flux absorbed by the surface is dynamically adjusted within the range and determined based on the radiation intensity in the meteorological forcing data.

[0127] After completing the hydrodynamic and photothermal calculations, the system enters the model coupling stage. The velocity field and turbulence diffusion coefficient calculated by the hydrodynamic sub-model, and the light intensity profile and vertical water temperature distribution calculated by the photothermal sub-model, are used as input parameters for the biochemical sub-model. The biochemical sub-model is constructed based on a three-compartment structure of nutrients, phytoplankton, and dissolved oxygen. At each time step, a set of ordinary differential equations is solved to calculate the spatiotemporal changes in algal biomass and nutrient concentration. Specifically, the phytoplankton growth term is described using the Monod formula, with a maximum growth rate set at 0.8 days. - ¹(Under the condition of water temperature 25℃), the half-saturated light intensity is 100 μE·m - ²·s - ¹; In the nutrient uptake term, the total nitrogen half-saturation constant is 0.3 mg / L, and the total phosphorus half-saturation constant is 0.03 mg / L. The model simultaneously considers processes such as algal respiration, death, and nutrient regeneration to ensure the conservation of ecosystem mass.

[0128] During the iterative advancement phase, the calculation results of the current time step of the biochemical model are used as the initial conditions for the next time step, establishing a cyclical iterative mechanism. At the end of each time step, the updated algal biomass and nutrient concentration fields are used as the initial fields for the next calculation step, while maintaining the continuity of hydrodynamic and photothermal conditions. This iterative process continues with a fixed step size of 15 minutes until a complete simulation of the preset future time period (6-72 hours) is completed. The final output spatiotemporal variation sequence of the key parameters of the algal bloom includes complete spatiotemporal evolution data of parameters such as chlorophyll a concentration, algal biomass, and nutrient concentration. The spatial resolution corresponds to the lake and reservoir monitoring grid system, and the temporal resolution is consistent with the calculation step size, forming a structured multidimensional dataset, providing a complete numerical basis for subsequent prediction correction and uncertainty analysis.

[0129] Furthermore, you can also view Figure 4 , Figure 4 This is a detailed step diagram based on step S40 in the first embodiment. Figure 4 The step of generating a candidate set of detection time windows for different lake and reservoir detection areas based on the algal bloom probability and the risk distribution includes steps S41-44:

[0130] Step S41: Based on the algal bloom probability and the risk distribution, the lake / reservoir detection area is divided into multiple spatial units;

[0131] Step S42: Based on the monitoring priority and preset monitoring frequency requirements of each space unit, determine the number and duration of time windows for each space unit in the future scheduling cycle. The monitoring priority is determined based on the probability of algal bloom.

[0132] Step S43: Based on the geographical location of the space unit, the travel speed of the unmanned monitoring platform, and the preset operating constraints, calculate the feasible start time range for each time window;

[0133] Step S44: Combine the spatial unit, time window duration, and feasible start time range to generate the detection time window candidate set containing multiple candidate monitoring task items.

[0134] In this embodiment, a candidate set of detection time windows is generated based on the prediction information. Based on the algal bloom prediction output, a complete set of candidate monitoring schemes is constructed through spatial discretization, temporal planning, and resource assessment.

[0135] Specifically, spatial units are divided based on the predicted algal bloom probability and risk distribution. This spatial unit division process reads the risk field data output by the prediction module, including the bloom probability value and corresponding confidence index for each grid unit. Based on the spatial distribution of the algal bloom probability, a threshold segmentation method is used to divide the lake / reservoir area into different risk levels: areas with a bloom probability higher than 70% are classified as high-risk units, 30%-70% as medium-risk units, and below 30% as low-risk units. Simultaneously, the division results are optimized and adjusted using the confidence data, marking units with a confidence level below 80%. The spatial unit division uses a fixed grid size standard, maintaining a 500m × 500m resolution in the core monitoring area and a 1000m × 1000m grid in the regular monitoring area. Each spatial unit records complete attribute information, including unit number, center point coordinates, risk level, and confidence score, forming a structured spatial management unit system.

[0136] When completing the monitoring priority assessment and time window parameters, the monitoring priority is calculated based on the risk level and confidence level of the spatial unit using a weighted scoring method, where the risk level has a weight of 0.6 and the confidence level has a weight of 0.4. Based on the priority scoring results and the preset monitoring frequency requirements, the time window configuration for each spatial unit in the future scheduling cycle is determined. The specific configuration rules based on the time window configuration are as follows: units with a priority score higher than 80 points are configured with 4-6 time windows, each lasting 10 minutes; units with a score of 60-80 points are configured with 2-3 time windows, each lasting 20 minutes; and units with a score lower than 60 points are configured with 1-2 time windows, each lasting 30 minutes. The preset monitoring frequency requirements are set according to the lake and reservoir management specifications, requiring at least one monitoring session per hour in the core area and at least one monitoring session every three hours in the regular area, ensuring that monitoring coverage meets the early warning requirements.

[0137] Based on the geographical location of the spatial units, the unmanned monitoring platform's travel speed, and the preset operational constraints, the feasible start time range for each time window is calculated. A travel time model for the unmanned monitoring platform is established based on the geographical location information of the spatial units. First, the travel distance from the platform's current position to the center of each spatial unit is calculated. Combined with the unmanned monitoring platform's travel speed of 2.5 m / s, the one-way travel time is estimated. Then, the preset operational constraints are comprehensively considered, including the platform's endurance of 8 hours, a power limit of 1600Wh, and prohibited navigation periods (e.g., navigation is prohibited in aquaculture areas from 6:00 to 18:00). Based on these constraints, a time conflict detection algorithm is used to eliminate infeasible periods, and the feasible start time range for each time window is calculated. The calculation process also considers the time continuity requirements of the monitoring task, ensuring a reasonable time buffer between adjacent time windows.

[0138] The spatial units, time window durations, and feasible start time ranges are combined to generate a candidate set of detection time windows containing multiple candidate monitoring tasks. The spatial unit information, time window duration parameters, and feasible start time ranges output above are combined and matched to generate standardized candidate monitoring tasks. Each candidate task includes complete descriptive fields: spatial unit number, time window start time, end time, duration, priority weight, and expected energy consumption. The candidate set of detection time windows is organized in a tree structure, arranging all candidate tasks in chronological order, and a spatial index is established to support fast retrieval. The candidate set output format adopts a standardized JSON structure, including a metadata header and a task list, ensuring that the generated set of candidate monitoring tasks forms a complete candidate set of detection time windows, providing sufficient decision space for subsequent optimization solutions.

[0139] Furthermore, you can also view Figure 5 , Figure 5This is a detailed step diagram based on step S50 in the first embodiment. Figure 5 The steps for optimizing and solving the detection time window scheduling strategy based on the candidate set of detection time windows and preset running constraints include steps S51-55:

[0140] Step S51: Obtain the platform parameters of the unmanned detection platform, wherein the platform parameters include at least one of the following: endurance time, travel speed, and sensor type;

[0141] Step S52: Convert the platform parameters into resource constraints and time conflict constraints of the preset optimization model;

[0142] Step S53: Select candidate time windows that satisfy the resource constraints and time conflict constraints from the candidate time window set;

[0143] Step S54: Perform a comprehensive evaluation on the candidate time windows according to the comprehensive evaluation function, and obtain a combination of candidate time windows based on the evaluation results;

[0144] Step S55: Generate the detection time window scheduling strategy based on the candidate time window combination.

[0145] In this embodiment, platform parameters of the unmanned detection platform are obtained, including at least one of the following: endurance, speed, and sensor type. Specifically, the implementation process based on the shown time window can be viewed... Figure 7 , Figure 7 A flowchart illustrating time-window-based scheduling.

[0146] The platform parameters are transformed into resource constraints and time conflict constraints in a preset optimization model. The core parameters, through the transformation process of the preset optimization model, form two types of constraints: resource constraints include total power consumption not exceeding 1600Wh and total navigation time not exceeding 8 hours; time conflict constraints include avoiding prohibited navigation periods (navigation is prohibited in aquaculture areas from 6:00 to 18:00, and in waterways from 22:00 to 5:00 the next day) and task interval requirements (adjacent monitoring tasks must retain at least 10 minutes of navigation buffer time). Based on the above transformation process, a linear programming method is used to quantify the platform parameters into mathematical model inputs. The power constraint is transformed into ensuring that the total energy consumption of each candidate time window does not exceed the limit, and the navigation time constraint is transformed into ensuring that the total execution time and navigation time of each time window do not exceed the limit, forming a complete constraint system.

[0147] Candidate time windows that satisfy the resource constraints and time conflict constraints are selected from the candidate set of detection time windows. When filtering the candidate set based on the mathematical programming strategy, a mixed-integer linear programming framework is used to establish binary selection variables for the candidate time windows, and then resource constraints and time conflict constraints are loaded. A constraint propagation algorithm is used to quickly eliminate candidates that do not meet the basic conditions, including those with excessive power, overlapping times, or violations of no-fly zones. Specifically, a depth-first search strategy is used to traverse the candidate set, and each candidate time window is checked for constraint satisfaction: verifying whether its start time is within the platform's reachable time range, whether its duration meets monitoring requirements, and whether its energy consumption budget is within the allowable range of remaining power. Through multiple rounds of filtering iterations, candidate time windows that satisfy all constraints are output. These candidate time windows retain the priority and attribute information of the original candidate set, providing input for subsequent optimization and evaluation.

[0148] The candidate time windows are comprehensively evaluated using a comprehensive evaluation function, and a combination of candidate time windows is obtained based on the evaluation results. In the implementation of the optimal solution through the calculation of the comprehensive evaluation function, the comprehensive evaluation function is constructed using a weighted summation method. This construction includes four core indicators: a risk coverage weight of 0.4, calculated based on the probability of algal blooms in the corresponding spatial units of the candidate time windows; an energy consumption weight of 0.2, estimated based on the platform's power consumption model; a latency weight of 0.15, considering the urgency of task execution; and a prediction instability penalty weight of 0.15, adjusted in conjunction with the prediction confidence. During the calculation process, the evaluation function value is calculated for each candidate time window combination, and a branch and bound algorithm is used to search for the optimal solution. The candidate time window combination is obtained based on the calculation results. The detection time window scheduling strategy is generated based on the candidate time window combination. The detection time window scheduling strategy is output in a structured data format, containing complete execution parameters for each time window: monitoring unit number, start time, duration, priority order, and energy consumption allocation. It also includes a constraint satisfaction verification report to ensure the executability and optimal resource utilization of the scheduling scheme.

[0149] Furthermore, you can also view Figure 6 , Figure 6 This is a detailed step diagram based on step S60 in the first embodiment. Figure 6 The steps include steps S61-63: distributing the detection time window scheduling strategy to the unmanned detection platform to execute the detection task, acquiring the measured data collected after the detection task is executed, and feeding the measured data back into the observation data for data registration and model assimilation of the digital twin.

[0150] Step S61: Mark the measured data as special data with feedback identifier;

[0151] Step S62: Input the special data into the multi-source registration and quality control processing flow, assimilate it with the observation data, and generate a feedback observation vector;

[0152] Step S63: Update the model state and model parameters of the digital twin online using the feedback observation vector.

[0153] In this embodiment, the measured data is marked as specialized data with feedback identifiers. Specifically, after the unmanned monitoring platform completes its monitoring task, the raw monitoring data transmitted back via the 4G / 5G wireless link is first assigned a specific data identifier, marking it as specialized data with feedback identifiers. Subsequently, the timeliness verification is performed, checking the deviation between the data timestamp and the current system time, and discarding data records with a time delay exceeding 30 minutes to ensure the real-time validity of the specialized data. In the format standardization process, the system converts heterogeneous data collected by different sensors into a unified standardized format: chlorophyll a concentration is standardized to μg / L with one decimal place, dissolved oxygen concentration is converted to mg / L, spatial coordinates are converted to standard grid encoding, and timestamps are standardized to UTC time format. The standardization process also includes a data integrity check, marking data records missing key fields and triggering a retransmission mechanism.

[0154] The specialized data is input into the multi-source registration and quality control processing flow, assimilated with the observation data, and a feedback observation vector is generated. The validated specialized data is then injected into the multi-source data registration pipeline and assimilated together with real-time monitoring data from conventional monitoring sources such as fixed stations and buoys. In this process, the specialized data first participates in the time synchronization stage, using NTP time synchronization service for timestamp calibration to ensure time consistency with other data sources. In the spatial alignment stage, the GNSS coordinates carried by the specialized data are mapped to the lake / reservoir standard grid system and fused with the real-time monitoring data within a unified spatial framework. Subsequently, the specialized data participates in drift compensation calculations, and its observations are incorporated into residual sliding window analysis for sensor drift detection and compensation model updates. Finally, all data sources undergo outlier removal processing, identifying and excluding abnormal data points based on the 3σ criterion, generating the feedback observation vector.

[0155] Subsequently, the model state and parameters of the digital twin are updated online using the feedback observation vector. When inputting the feedback observation vector into the data assimilation algorithm, an ensemble Kalman filter is employed, using the feedback observation vector as additional observation input to enhance the observation network density of the assimilation system. In the assimilation calculation, the high spatiotemporal resolution observation values ​​provided by the feedback observation vector are compared with the model prediction values, and the analytical values ​​of the state variables are adjusted using the Kalman gain matrix. The specific update process includes: adjusting the phytoplankton biomass field of the biochemical sub-model based on the re-irrigated chlorophyll a concentration data, optimizing the parameter settings of the photothermal sub-model using the feedback water temperature data, and achieving online updates of the model state and parameters of the digital twin through multi-parameter collaborative assimilation. The assimilation cycle is adaptively adjusted according to the characteristics of the specific data, shortening the assimilation interval to 20 minutes during periods of dense specific data to ensure the system responds quickly to environmental changes.

[0156] To continuously optimize the algal bloom prediction model, parameter calibration is required based on the updated digital twin output. This calibration focuses on adjusting the model parameters using real observational information provided by specialized data, targeting the physical constraint machine learning component within the prediction module. Specifically, the calibration process involves calculating the residual sequence between the predicted values ​​and the observed values ​​from the specialized data, then updating the weight parameters of the PIML residual network using a backpropagation algorithm with a learning rate set to 0.001. Simultaneously, based on long-term comparative analysis of the specialized data and prediction results, the threshold parameters in calculating the algal bloom probability are dynamically adjusted to optimize the prediction's sensitivity and specificity. The calibrated prediction model not only possesses better adaptability but also more accurately reflects the actual patterns of algal blooms in lakes and reservoirs, providing a more reliable decision-making basis for the next round of monitoring and scheduling.

[0157] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the time window scheduling method for predicting and monitoring algal blooms in lakes and reservoirs in this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0158] This application provides a time window scheduling device for predicting and monitoring algal blooms in lakes and reservoirs. The time window scheduling device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the time window scheduling method for predicting and monitoring algal blooms in lakes and reservoirs in the first embodiment described above.

[0159] The following is for reference. Figure 8The diagram illustrates a structural schematic of a time window scheduling device for predicting and monitoring algal blooms in lakes and reservoirs, suitable for implementing embodiments of this application. The time window scheduling device for predicting and monitoring algal blooms in lakes and reservoirs in embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 8 The lake and reservoir algal bloom prediction and monitoring time window scheduling device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0160] like Figure 8 As shown, the lake / reservoir algal bloom prediction and monitoring time window scheduling device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the lake / reservoir algal bloom prediction and monitoring time window scheduling device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the lake / reservoir algal bloom prediction and monitoring time window scheduling equipment to exchange data wirelessly or via wired communication with other devices. Although the figure shows a lake / reservoir algal bloom prediction and monitoring time window scheduling equipment with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.

[0161] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0162] The lake and reservoir algal bloom prediction and monitoring time window scheduling device provided in this application adopts the lake and reservoir algal bloom prediction and monitoring time window scheduling method in the above embodiments, which can solve the technical problems of resource misallocation and response delay caused by the disconnect between static prediction and rigid monitoring scheduling of lake and reservoir algal blooms in the prior art. Compared with the prior art, the beneficial effects of the lake and reservoir algal bloom prediction and monitoring time window scheduling device provided in this application are the same as the beneficial effects of the lake and reservoir algal bloom prediction and monitoring time window scheduling method provided in the above embodiments, and other technical features in the lake and reservoir algal bloom prediction and monitoring time window scheduling device are the same as the features disclosed in the previous embodiment method, and will not be repeated here.

[0163] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0164] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0165] This application provides a storage medium, which is a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, which are used to execute the lake and reservoir algal bloom prediction and monitoring time window scheduling method in the above embodiments.

[0166] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0167] The aforementioned computer-readable storage medium may be included in the lake and reservoir algal bloom prediction and monitoring time window scheduling equipment; or it may exist independently and not be assembled into the lake and reservoir algal bloom prediction and monitoring time window scheduling equipment.

[0168] The aforementioned computer-readable storage medium carries one or more programs. When the aforementioned one or more programs are executed by the lake and reservoir algal bloom prediction and monitoring time window scheduling device, the lake and reservoir algal bloom prediction and monitoring time window scheduling device implements the technical content of the lake and reservoir algal bloom prediction and monitoring time window scheduling method embodiment shown above.

[0169] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0170] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0171] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0172] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described lake and reservoir algal bloom prediction and monitoring time window scheduling method. This solves the technical problems of resource mismatch and response delays caused by the disconnect between static prediction and rigid monitoring scheduling of lake and reservoir algal blooms in the prior art. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the lake and reservoir algal bloom prediction and monitoring time window scheduling method provided in the above embodiments, and will not be repeated here.

Claims

1. A method for scheduling time windows for predicting and monitoring algal blooms in lakes and reservoirs, characterized in that, The method for scheduling time windows for predicting and monitoring algal blooms in lakes and reservoirs includes the following steps: Acquire multi-source detection data from multiple sensors in a lake or reservoir, and register the data errors of the multi-source detection data to obtain registered observation data; The observation data is subjected to model data assimilation processing, and the model state and model parameters of the digital twin are updated online based on the data assimilation results. The digital twin is a virtual dynamic mapping model constructed based on the physical entities of the lake and reservoir. Predict future algal bloom information based on the assimilated and updated digital twin, and generate algal bloom prediction results based on the algal bloom information. The algal bloom prediction results include algal bloom outbreak probability and risk distribution. Based on the algal bloom probability and the risk distribution, a candidate set of detection time windows is generated for different lake and reservoir detection areas; Based on the candidate set of detection time windows and the preset running constraints, an optimization solution is performed to obtain the detection time window scheduling strategy. The detection time window scheduling strategy is sent to the unmanned detection platform to execute the detection task, and the actual measurement data corresponding to the executed detection task is obtained. The actual measurement data is fed back to the observation data to assimilate and update the digital twin. The step of predicting algal bloom information for future periods based on the assimilated and updated digital twin, and generating algal bloom prediction results based on the algal bloom information, includes: The process of driving the digital twin to simulate and calculate algal bloom information for future periods based on the assimilated and updated model state and model parameters, and obtaining a preliminary algal bloom prediction sequence based on the simulation results, includes: using the model state and model parameters as initial conditions, and simultaneously solving the coupled hydrodynamic sub-model, biochemical sub-model, and photothermal sub-model in the digital twin based on preset future meteorological and hydrological forcing data; cyclically executing the coupled calculation of the hydrodynamic sub-model, the biochemical sub-model, and the photothermal sub-model at preset time steps, and outputting the spatiotemporal change sequence within the future periods; The spatiotemporal variation sequence is used as the preliminary prediction sequence for algal blooms. Physical constraints are introduced to correct the preliminary prediction sequence of algal blooms. The physical constraints include at least one of the laws of conservation of mass and conservation of energy. Uncertainty quantification is performed on the corrected preliminary algal bloom prediction sequence to generate confidence information, which includes confidence intervals or probability distributions; The corrected preliminary algal bloom prediction sequence is combined with the confidence information to generate the algal bloom prediction result.

2. The method for scheduling time windows for predicting and monitoring algal blooms in lakes and reservoirs as described in claim 1, characterized in that, The steps for registering the multi-source detection data to obtain registered observation data include: The multi-source detection data is subjected to time synchronization and spatial alignment operations to generate a spatiotemporally consistent fusion dataset; The mean of the residuals is calculated based on the residuals between the fused dataset and the predicted values ​​of the digital twin; When the mean residual exceeds a preset residual threshold, it is determined that there is drift data in the fused dataset, and the drift data is an abnormal value caused by sensor error or environmental interference. Numerical compensation is performed on the drift data, and the fused dataset is updated based on the numerical compensation results; The updated fused dataset is then subjected to outlier identification and filtering to obtain the observation data.

3. The method for scheduling time windows for predicting and monitoring algal blooms in lakes and reservoirs as described in claim 1, characterized in that, The step of cyclically performing the coupled calculations of the hydrodynamic sub-model, the biochemical sub-model, and the photothermal sub-model at a preset time step, and outputting the spatiotemporal change sequence within a future time period, includes: Within each time step, the hydrodynamic sub-model and the photothermal sub-model are respectively driven to calculate the flow field distribution and the vertical distribution of water temperature under illumination; The flow field distribution and the vertical distribution of light and water temperature are used as inputs to the biochemical sub-model to calculate the spatiotemporal changes of algal biomass and nutrient concentration. The calculation result of the current time step of the biochemical sub-model is used as the initial condition for the next time step. The simulation is repeated until the preset future time period is completed, and the continuous spatiotemporal change sequence is output.

4. The method for scheduling time windows for predicting and monitoring algal blooms in lakes and reservoirs as described in claim 1, characterized in that, The steps for generating candidate sets of detection time windows for different lake / reservoir detection areas based on the algal bloom probability and the risk distribution include: Based on the probability of algal blooms and the risk distribution, the lake and reservoir detection area is divided into multiple spatial units; Based on the monitoring priority and preset monitoring frequency requirements of each space unit, the number and duration of time windows for each space unit in the future scheduling cycle are determined, and the monitoring priority is determined based on the probability of algal blooms. Based on the geographical location of the space unit, the travel speed of the unmanned monitoring platform, and the preset operational constraints, the feasible start time range for each time window is calculated. The spatial unit, time window duration, and feasible start time range are combined to generate the detection time window candidate set containing multiple candidate monitoring tasks.

5. The method for scheduling time windows for predicting and monitoring algal blooms in lakes and reservoirs as described in claim 1, characterized in that, The steps for optimizing the detection time window scheduling strategy based on the candidate set of detection time windows and preset running constraints include: Obtain the platform parameters of the unmanned detection platform, wherein the platform parameters include at least one of the following: endurance time, travel speed, and sensor type; The platform parameters are converted into resource constraints and time conflict constraints of a preset optimization model; Candidate time windows that satisfy the resource constraints and time conflict constraints are selected from the candidate time windows. The candidate time windows are comprehensively evaluated based on the comprehensive evaluation function, and a combination of candidate time windows is obtained based on the evaluation results; The detection time window scheduling strategy is generated based on the candidate time window combination.

6. The method for scheduling time windows for predicting and monitoring algal blooms in lakes and reservoirs as described in claim 1, characterized in that, The step of feeding back the measured data to the observed data for assimilation and updating of the digital twin includes: The measured data are marked as special data with feedback identifiers; The specific data is input into the multi-source registration and quality control processing flow, and assimilated with the observation data to generate a feedback observation vector; The model state and model parameters of the digital twin are updated online using the feedback observation vector.

7. A time window scheduling device for predicting and monitoring algal blooms in lakes and reservoirs, characterized in that, The lake and reservoir algal bloom prediction and monitoring time window scheduling device stores a computer program, which, when executed by a processor, implements the lake and reservoir algal bloom prediction and monitoring time window scheduling method according to any one of claims 1-6.

8. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the lake and reservoir algal bloom prediction and monitoring time window scheduling method according to any one of claims 1-6.

Images