Nuclear power water intake biomass prediction method based on deep learning

Abstract

The invention relates to the technical field of nuclear power water intake biomass prediction, and discloses a nuclear power water intake biomass prediction method based on deep learning, and the method comprises the steps: firstly collecting data such as environmental factors, biological distribution and historical prediction deviation, building a model selection association equation, and selecting a current prediction model according to the model selection association equation; and whether the model is optimized or not is judged through actual biomass data, and finally an optimized prediction model structure is obtained. In the method, a monitoring area and a target organism type are delimited, a statistical time period and key parameters are set, a swarm intelligent optimization technology is utilized to construct an associated equation and a weight data set, a model is determined by comparing success rates, and whether the structure of the model is optimized or not is judged according to an effect difference critical value. The method can improve prediction accuracy and adaptability, and provides an effective scheme for nuclear power water intake biomass prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of biological quantity prediction for nuclear power plant water intakes, and specifically to a method for predicting the biological quantity of nuclear power plant water intakes based on deep learning. Background Art

[0002] During the operation of a nuclear power plant, the prediction of the biological quantity at the nuclear power plant water intake is crucial. The cooling system of a nuclear power plant needs to draw a large amount of water from the outside to maintain the normal operation of the unit, and the biological quantity and species distribution in the water intake area will have an important impact on the operation efficiency of the cooling system. If the biological quantity in the water intake area is too large, it may block the pipeline, affect the water quality, and even cause the failure of the cooling system, thus affecting the safe and stable operation of the nuclear power plant.

[0003] Traditional methods for predicting the biological quantity at nuclear power plant water intakes have many deficiencies. Traditional methods often rely on simple statistical models or empirical formulas and are difficult to accurately capture the complex non-linear relationships between biological quantity and environmental factors. Changes in environmental factors such as water temperature, salinity, dissolved oxygen, etc. will all affect the growth, reproduction, and distribution of organisms, and there are interactions between these factors. Traditional methods are difficult to comprehensively and accurately describe these relationships; traditional methods do not make full use of historical data. In actual situations, environmental factor data, biological species distribution data, and historical prediction deviation data at different time periods all contain rich information, and traditional methods often cannot effectively integrate these data, resulting in low accuracy and reliability of prediction results.

[0004] In addition, traditional methods lack adaptability and optimization capabilities. When environmental conditions change, traditional methods are difficult to automatically adjust model parameters or structures to adapt to new situations, and a large amount of manual adjustment and optimization work is required, which is not only time-consuming and laborious but also difficult to ensure the accuracy and timeliness of the adjustment.

[0005] With the continuous development of deep learning technology, its applications in various fields are becoming more and more extensive. Deep learning has powerful non-linear fitting capabilities and self-learning capabilities, can effectively handle complex non-linear problems, and provides new ideas and methods for predicting the biological quantity at nuclear power plant water intakes. However, at present, the research on applying deep learning technology to predict the biological quantity at nuclear power plant water intakes is relatively scarce, and there is a lack of a systematic and perfect prediction method based on deep learning.

[0006] Therefore, there is an urgent need for a method for predicting the biological quantity of nuclear power plant water intakes based on deep learning to solve the deficiencies of traditional methods, improve the accuracy and reliability of prediction, and provide strong support for the safe and stable operation of nuclear power plants. Summary of the Invention

[0007] The object of the present invention is to provide a method for predicting the number of organisms at a nuclear power plant water intake based on deep learning to solve the problems raised in the above background art.

[0008] To achieve the above object, the present invention provides a method for predicting the number of organisms at a nuclear power plant water intake based on deep learning, the method comprising:

[0009] S1. Collect environmental factor data, biological species distribution data, and historical prediction deviation data in the nuclear power plant water intake area at different time periods to obtain a current environmental factor data set, a current biological distribution data set, and a current prediction deviation data set;

[0010] S2. Collect historical application success rate data of each candidate prediction model, environmental factor data of each candidate prediction model, biological distribution data at different time periods, and prediction deviation data, and construct a final model selection correlation equation;

[0011] S3. Select a current prediction model according to the final model selection correlation equation, the current environmental factor data set, the current biological distribution data set, and the current prediction deviation data set;

[0012] S4. Determine whether parameter adjustment or structure optimization of the current prediction model is required by collecting actual biological quantity data after using the current prediction model, and obtain a determination result;

[0013] S5. Perform structure optimization processing on the current prediction model according to the determination result to obtain a final prediction model structure.

[0014] Preferably, the S1 includes the following steps:

[0015] S11. Delimit the nuclear power plant water intake monitoring area to be predicted and corresponding several types of biological species to obtain a target biological species set; then set several key parameter types affecting the selection of the prediction model to obtain a prediction model influence parameter type set;

[0016] S12. Set a first current statistical time period; in cooperation with the first current statistical time period and the prediction model influence parameter type set, obtain environmental factor data, biological distribution data at different time periods, and prediction deviation data of each biological species in the target biological species set to obtain a current environmental factor data set, a current biological distribution data set, and a current prediction deviation data set.

[0017] Preferably, the S2 includes the following steps:

[0018] S21. Set the historical statistical period; in combination with the target biological species set and the prediction model influence parameter type set, collect the application success rate data of each candidate prediction model, the environmental factor data of each candidate prediction model, the biological distribution data of different periods, and the prediction deviation data within the historical statistical period to obtain the historical candidate model success rate data set, the historical environmental factor data set, the historical biological distribution data set, and the historical prediction deviation data set;

[0019] S22. Use the historical candidate model success rate data set, the historical environmental factor data set, the historical biological distribution data set, and the historical prediction deviation data set to construct the final model selection correlation equation and the final environmental factor weight data set.

[0020] Preferably, in S22, the group intelligence optimization technology is used to construct the final model selection correlation equation and the final environmental factor weight data set.

[0021] Preferably, the said S3 includes the following steps:

[0022] S31. Set the current initial prediction model; substitute each data in the current environmental factor data set, the current biological distribution data set, the current prediction deviation data set, and the final environmental factor weight data set into the final model selection correlation equation for correlation calculation to obtain the current candidate model success rate data set;

[0023] S32. When the candidate prediction model corresponding to the highest success rate data in the current candidate model success rate data set is the same as the current initial prediction model, the current prediction model is not replaced; otherwise, the candidate prediction model corresponding to the highest success rate data in the current candidate model success rate data set is used as the current prediction model, both are denoted as the current prediction model.

[0024] Preferably, the said S4 includes the following steps:

[0025] S41. Set the second current statistical period; set a number of time nodes within the second current statistical period to obtain the current time node set; set a number of verification parameters that can reflect the application effect of the prediction model to obtain the model verification parameter type set;

[0026] S42. In combination with the model verification parameter type set and the current time node set, collect the actual biological quantity data after using the current prediction model to obtain the current verification parameter matrix; then collect the average biological quantity data of a number of time nodes before using the current prediction model to obtain the historical average quantity data set; calculate the difference values between the historical average quantity data set and each row of data in the current verification parameter matrix to obtain the current effect difference data set;

[0027] S43. Set the first effect difference threshold and the second effect difference threshold; when there is a current effect difference value in the current effect difference dataset that is less than the second effect difference threshold, switch the current prediction model back to the current initial prediction model again; when there is a current effect difference value in the current effect difference dataset that is greater than or equal to the second effect difference threshold and less than the first effect difference threshold, proceed to S5; otherwise, proceed to S44;

[0028] S44. Predict the biological quantity data at future time nodes according to the current verification parameter matrix and using a feedforward neural network model to obtain a future verification parameter matrix; then calculate the difference values between each row of data in the historical average quantity dataset and the future verification parameter matrix to obtain a future effect difference dataset; when there is a future effect difference value in the future effect difference dataset that is less than the second effect difference threshold, proceed to S5; otherwise, do not process.

[0029] Preferably, S5 includes the following steps:

[0030] S51. Set the initial model structure optimization parameters; perform structure optimization processing on the current prediction model using the initial model structure optimization parameters and store it; after completing the structure optimization processing and storage of the current prediction model, collect the current biological quantity data according to the model verification parameter type set to obtain the current optimized verification parameter set; calculate the difference value between the current optimized verification parameter set and the historical average quantity dataset to obtain the current optimized difference value;

[0031] S52. When the current optimized difference value is greater than or equal to the first effect difference threshold, use the initial model structure optimization parameters as the final prediction model structure; otherwise, adjust the initial model structure optimization parameters until the current optimized difference value is greater than or equal to the first effect difference threshold.

[0032] Preferably, the adjustment of the initial model structure optimization parameters in S52 includes the following steps:

[0033] S521. Set the value range of the initial model structure optimization parameters to obtain the current parameter value range; construct a model structure optimization population; set the maximum number of iterations and the current number of iterations of the model structure optimization population, denoted as the structure optimization maximum number of iterations and the structure optimization current number of iterations respectively;

[0034] S522. Set the initial positions of each individual in the model structure optimization population according to the current parameter value range to obtain a second initial position set;

[0035] S523. Construct a fitness evaluation function for the model structure optimization population;

[0036] S524. Start the iteration. Before the iteration, set the current iteration number of structure optimization to 1. During each iteration, use the fitness evaluation function of the model structure optimization population to calculate the fitness values of each individual position in the model structure optimization population updated in the previous iteration and update the positions of each individual in the model structure optimization population updated in the previous iteration.

[0037] S525. When the current iteration number of structure optimization is equal to the maximum iteration number of structure optimization, stop the iteration to obtain the second final global optimal fitness and the second final global optimal position; otherwise, continue the iteration until the current iteration number of structure optimization is equal to the maximum iteration number of structure optimization. Take the second final global optimal fitness as the current optimized difference value after optimization. When the current optimized difference value after optimization is greater than or equal to the first effect difference critical value, perform structure optimization processing on the current prediction model using the second final global optimal position and store it to obtain the final prediction model structure; otherwise, return to S524 to continue the iteration until the current optimized difference value after optimization is greater than or equal to the first effect difference critical value.

[0038] Preferably, the model verification parameter type set includes water temperature parameter, salinity parameter, dissolved oxygen parameter, and biological density parameter.

[0039] Preferably, the determination of the target biological species set includes the following steps:

[0040] S111. Conduct a census of the biological community in the nuclear power water intake area and record the biological species with occurrence frequencies exceeding a preset threshold in different time periods.

[0041] S112. Combine the operating characteristics of the nuclear power cooling system to screen out the key biological species that may affect the water intake efficiency and summarize them to form the target biological species set.

[0042] Compared with the prior art, the beneficial effects of the present invention are:

[0043] By collecting the environmental factor data, biological species distribution data, and historical prediction deviation data in the nuclear power water intake area at different times, this method can comprehensively and fully utilize various data information related to biological quantity prediction. The collection of these data lays a solid foundation for subsequent model construction and selection, enabling the model to better understand and capture the laws of biological quantity changes.

[0044] In terms of model selection, a final model selection correlation equation is constructed and combined with multiple datasets to select the current prediction model, which changes the limitations of traditional single models. It can dynamically select the most suitable prediction model according to the actual environment and biological distribution, greatly improving the adaptability of the model to different scenarios. For example, under different seasons and different hydrological conditions, this method can automatically select the model that can most accurately reflect the current change in biological quantity, thus improving the accuracy of prediction.

[0045] By collecting the actual biological quantity data after using the current prediction model to determine whether parameter adjustment or structural optimization of the model is needed, this process realizes the self-optimization and improvement of the model. When the model shows prediction deviation in actual application, it can be detected and adjusted in time, enabling the model to continuously adapt to the changes in the environment and the dynamic development of biological quantity. This self-optimization mechanism avoids the cumbersome work of manually adjusting the model frequently in traditional methods, and at the same time improves the prediction accuracy and reliability of the model.

[0046] In the process of model structure optimization, advanced technologies and methods such as swarm intelligence optimization technology and feedforward neural network model are adopted to further improve the performance of the model. Swarm intelligence optimization technology can search for the optimal model structure parameters in the parameter space, making the model structure more reasonable; the feedforward neural network model can accurately predict the future biological quantity data, providing strong support for the optimization of the model.

[0047] In addition, various parameters and critical values set in this method, such as the first effect difference critical value, the second effect difference critical value, etc., provide clear standards and bases for the adjustment and optimization of the model. This makes the optimization process of the model more scientific and reasonable, avoids the interference of subjective factors, and ensures the effect and stability of model optimization.

[0048] The method of the present invention significantly improves the accuracy, reliability and adaptability of the biological quantity prediction at the nuclear power water intake through comprehensive data collection, dynamic model selection, self-optimizing model adjustment and advanced technology application. It can provide more accurate biological quantity prediction information for the operation of the cooling system of the nuclear power plant, ensure the safe and stable operation of the nuclear power plant, and at the same time reduce the manual intervention and maintenance costs, having important practical application value and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is the working principle diagram of the method for predicting the biological quantity at the nuclear power water intake based on deep learning according to the present invention;

[0050] Figure 2 is the flowchart of S1 data collection;

[0051] Figure 3Flowchart for selecting the current prediction model for S3;

[0052] Figure 4 Flowchart for determining the adjustment of the S4 model;

[0053] Figure 5 Flowchart for parameter adjustment of S52. Detailed implementation manners

[0054] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0055] Please refer to Figures 1 - 5 , the present invention provides a method for predicting the number of organisms at the nuclear power plant water intake based on deep learning, including the following steps:

[0056] Collect environmental factor data, biological species distribution data, and historical prediction deviation data in the nuclear power plant water intake area at different time periods to obtain the current environmental factor data set, the current biological distribution data set, and the current prediction deviation data set.

[0057] Collect historical application success rate data of each candidate prediction model, environmental factor data of each candidate prediction model, biological distribution data at different time periods, and prediction deviation data, and construct a final model selection correlation equation.

[0058] Select the current prediction model according to the final model selection correlation equation, the current environmental factor data set, the current biological distribution data set, and the current prediction deviation data set.

[0059] By collecting the actual number of organisms data after using the current prediction model, determine whether it is necessary to adjust the parameters or optimize the structure of the current prediction model to obtain a determination result.

[0060] Perform structure optimization processing on the current prediction model according to the determination result to obtain the final prediction model structure.

[0061] Embodiment 1: This embodiment details step S1. First, it is necessary to delimit the monitoring area of the nuclear power plant water intake to be predicted. The determination of this area needs to comprehensively consider factors such as the actual geographical location of the nuclear power plant water intake, the influence range of the water flow, and the main area of biological activities. For example, according to the pipeline layout of the nuclear power plant water intake, the water flow velocity distribution, and historical biological monitoring data, a specific geographical range can be reasonably defined as the object area for subsequent monitoring and data collection.

[0062] It is necessary to determine a corresponding number of biological species categories to form a set of target biological species. This process requires a prior census of the biological communities in the nuclear power plant water intake area. During the census, various biological species that appear during different time periods are recorded in detail, and the occurrence frequency of each biological species in different time periods is counted. The time periods here can be divided according to factors such as the living habits of the organisms and seasonal changes, such as monthly, quarterly, or according to different hydrological cycles, etc. Then, a preset threshold is set, and the biological species with occurrence frequencies exceeding this threshold are recorded.

[0063] Combined with the operating characteristics of the nuclear power cooling system, the recorded biological species are screened. During the operation of the nuclear power cooling system, parameters such as water temperature and water flow velocity will change to a certain extent. Some organisms may gather near the water intake due to these changes, thus affecting the water intake efficiency. Therefore, it is necessary to analyze factors such as the living habits of each biological species, its adaptability to the environment, and the possible impact on the cooling system, and screen out the key biological species that may affect the water intake efficiency. Summing up these key biological species forms the set of target biological species.

[0064] After completing the delineation of the set of target biological species, it is necessary to set the first current statistical period. The setting of the first current statistical period needs to comprehensively consider factors such as the cycle of biological quantity changes, the feasibility of data collection, and the timeliness of prediction. For example, if the quantity changes of the target biological species have an obvious seasonal cycle, then the first current statistical period can be set as a complete season or a combination of several seasons to ensure that sufficient data reflecting the law of biological quantity changes can be collected.

[0065] After setting the first current statistical period and the pre-determined set of impact parameter types of the prediction model, start obtaining the relevant data of each biological species in the set of target biological species. The set of impact parameter types of the prediction model includes key parameter types that affect the prediction of biological quantity, such as environmental parameter types like water temperature parameters, salinity parameters, dissolved oxygen parameters, etc., and other factor parameter types that may affect the biological distribution.

[0066] For each target biological species, within the first current statistical period, according to different time periods, collect its corresponding environmental factor data. The collection of environmental factor data requires the use of professional monitoring equipment. For example, water temperature can be measured by a temperature sensor, salinity can be detected by a salinometer, and dissolved oxygen content can be measured by a dissolved oxygen meter, etc. At the same time, collect the biological distribution data for different time periods. The collection of biological distribution data can adopt methods such as field sampling and image monitoring, and record information such as the distribution location and quantity of each target biological species in different time periods.

[0067] In addition, it is also necessary to collect historical prediction deviation data. Historical prediction deviation data refers to the difference data between the prediction results of the prediction model and the actual biological quantity during the previous prediction process. These data can be obtained from historical prediction records. By comparing and analyzing the historical prediction results and actual monitoring data, the prediction deviation value for each time period is calculated.

[0068] Through the above series of operations, the collected environmental factor data, biological distribution data, and prediction deviation data are sorted and stored to obtain the current environmental factor dataset, the current biological distribution dataset, and the current prediction deviation dataset respectively. These datasets will serve as important bases for subsequent model selection and optimization. During the data sorting process, it is necessary to ensure the accuracy and integrity of the data, and perform necessary cleaning and preprocessing on the collected data to remove abnormal data and noise interference, so as to improve the accuracy and reliability of the subsequent prediction model. For example, for the obviously unreasonable values in the environmental factor data, verification and correction are required; for the missing parts in the biological distribution data, they need to be supplemented by reasonable methods.

[0069] Embodiment 2: This embodiment details step S2.

[0070] Set the historical statistical period, and the determination of this period should comprehensively consider the effectiveness and comprehensiveness of the data. Usually, a time period covering multiple biological growth cycles, different seasonal climate conditions, and different operating states of the nuclear power cooling system is selected, such as a time range of the past 3 to 5 years. This is to ensure that the collected data can reflect the impacts of various possible environmental changes and system operating conditions on the biological quantity and prediction model.

[0071] After setting the historical statistical period, in conjunction with the determined target biological species set and the prediction model influence parameter type set, relevant data collection begins. The target biological species set is obtained through a census and screening of the biological community in the nuclear power water intake area in the previous step, and includes the key biological species that may affect the water intake efficiency. The prediction model influence parameter type set includes various key parameters that affect the prediction of biological quantity, such as environmental factor parameters such as water temperature, salinity, and dissolved oxygen content, as well as biological distribution data, prediction deviation data, etc.

[0072] The data collection work specifically includes: for each candidate prediction model within the historical statistical period, record the application success rate data at different time periods. The calculation of the application success rate is based on the degree of agreement between the prediction results of the prediction model and the actual biological quantity. For example, the success or failure of each model in a specific time period can be determined by comparing the error range between the predicted value and the measured value, and these results are statistically sorted. At the same time, collect the corresponding environmental factor data for each candidate prediction model within the corresponding time period. The collection method of these data is the same as that of the environmental factor data in step S1, and is measured and recorded by professional monitoring equipment at different time nodes.

[0073] It is also necessary to collect the biological distribution data and prediction deviation data at different time periods within the historical statistical period. The collection of biological distribution data also uses methods such as field sampling and image monitoring, and details the distribution locations and quantity changes of each target biological species at different time periods. The prediction deviation data is obtained from historical prediction records, and the prediction results of each candidate prediction model within the historical statistical period are compared with the actual monitoring data to calculate the prediction deviation value for each time period.

[0074] Through the above data collection work, a historical candidate model success rate dataset, a historical environmental factor dataset, a historical biological distribution dataset, and a historical prediction deviation dataset can be obtained. These datasets contain a large amount of information about the performance of candidate prediction models within the historical statistical period and the corresponding environmental and biological data, providing rich materials for selecting the correlation equation and the final environmental factor weight dataset for constructing the final model.

[0075] It is necessary to use these datasets to construct the final model selection correlation equation and the final environmental factor weight dataset. The group intelligence optimization technology is adopted in the construction process. This technology simulates the intelligent behavior of biological groups to find the optimal solution. For example, the particle swarm optimization algorithm simulates the foraging behavior of bird flocks. Each particle represents a possible solution, and through the mutual cooperation and information sharing among the particles, the position is continuously adjusted in the solution space to find the optimal model selection correlation equation parameters and environmental factor weights.

[0076] In specific implementation, first, it is necessary to determine the relevant parameters of the group intelligence optimization technology, such as the group size, the number of iterations, the learning factor, etc. The setting of these parameters will affect the optimization effect and efficiency, and need to be reasonably adjusted according to the actual data scale and problem complexity. Then, input the historical candidate model success rate dataset, the historical environmental factor dataset, the historical biological distribution dataset, and the historical prediction deviation dataset into the optimization algorithm as the basis for optimization.

[0077] The optimization algorithm will, based on this data, through continuous iterative calculations, adjust the structure and parameters of the model selection correlation equation, as well as the weight values of each environmental factor. In each iteration, the algorithm will evaluate the fitting degree of the current model selection correlation equation and the combination of environmental factor weights to the historical data, that is, calculate the prediction accuracy of the success rate of the historical candidate models under the current combination. Through continuous optimization, the model selection correlation equation can better reflect the relationship between the success rate of the candidate prediction models, environmental factors, biological distribution data, and prediction deviation data, and at the same time determine the weight of each environmental factor in model selection.

[0078] When the optimization algorithm reaches the preset number of iterations or meets other termination conditions, stop the iteration to obtain the final model selection correlation equation and the final environmental factor weight data set. The final model selection correlation equation describes the mathematical relationship between the success rate of the candidate prediction models and various types of data, while the final environmental factor weight data set clarifies the importance of each environmental factor in the model selection process.

[0079] During the entire construction process, it is necessary to perform reasonable preprocessing and normalization on the data to ensure the stability and accuracy of the optimization algorithm. For example, standardize the environmental factor data with different dimensions so that they are within the same numerical range to avoid bias in the optimization results caused by dimension differences. At the same time, it is necessary to identify and process the outliers in the data to ensure that the constructed model selection correlation equation and environmental factor weight data set can accurately reflect the actual situation.

[0080] Example 3: This example details the implementation process of step S3.

[0081] It is necessary to set the current initial prediction model, and the determination of this model can be based on historical application situations or preliminary screening. For example, assume that in the past prediction of the number of organisms at the nuclear power plant water intake, the LSTM model was used under similar environmental conditions, and its structure and parameter settings have a certain degree of fit with the current prediction requirements, then the LSTM model can be set as the current initial prediction model.

[0082] Substitute each data in the current environmental factor data set, the current biological distribution data set, the current prediction deviation data set, and the final environmental factor weight data set into the final model selection correlation equation for correlation calculation. Taking the water temperature data in the current environmental factor data set as an example, assume that the water temperature data collected at a certain time period is 25°C, the salinity data is 30‰, and the dissolved oxygen data is 5 mg / L. These data need to be formatted and normalized according to the input requirements of the correlation equation. For example, convert the water temperature data to a value within the range of [0,1] through linear transformation, such as (25 - 15) / (35 - 15) = 0.5 (assuming the historical water temperature range is 15°C to 35°C).

[0083] The density data of a certain target biological species in the current biological distribution dataset at this time period is 100 individuals / m³, and it needs to be standardized according to the preset rules. For example, divide it by the historical maximum density value of 200 individuals / m³ to obtain a standardized value of 0.5. The historical prediction deviation of this time period recorded in the current prediction deviation dataset is 5%, which also needs to be converted into the corresponding numerical form. In the final environmental factor weight dataset, assume that the weight of water temperature is 0.3, the weight of salinity is 0.2, the weight of dissolved oxygen is 0.3, and the weight of other factors is 0.2. These weight values are used to adjust the influence degree of each environmental factor in the correlation calculation.

[0084] When substituting data, it is necessary to strictly follow the operation logic of the correlation equation selected by the final model. For example, the correlation equation may be an expression of weighted summation, where the environmental factor data is multiplied by the corresponding weights and then accumulated, and then combined with the biological distribution data and prediction deviation data for comprehensive calculation. Assume the form of the correlation equation is: success rate = a×(water temperature×0.3 + salinity×0.2 + dissolved oxygen×0.3 + biological density×0.1 + prediction deviation×0.1) + b, where a and b are the equation parameters. At this time, substitute the standardized water temperature of 0.5, salinity of 0.6 (assuming 30‰ is processed to 0.6), dissolved oxygen of 0.5 (5mg / L corresponds to 0.5), biological density of 0.5 (100 / 200), and prediction deviation of 0.05 (5% is converted to 0.05) into the equation to calculate the success rate value of this candidate model under the current data.

[0085] Perform the above data substitution and calculation process for each candidate prediction model. For example, the candidate models also include the CNN model, BP neural network model, etc. Calculate their success rates under the current environmental factor dataset, biological distribution dataset, and prediction deviation dataset respectively, so as to obtain the current candidate model success rate dataset. Assume that after calculation, the success rate of the LSTM model is 0.75, the CNN model is 0.82, and the BP model is 0.68. Then the current candidate model success rate dataset contains the success rate values of these three models.

[0086] Compare the values in the current candidate model success rate dataset to find the candidate prediction model corresponding to the highest success rate data. In the above example, the 0.82 of the CNN model is the highest success rate. At this time, it is necessary to judge whether the candidate model corresponding to the highest success rate is consistent with the current initial prediction model. If the current initial prediction model is the LSTM model and the one corresponding to the highest success rate is the CNN model, then the current prediction model needs to be replaced with the CNN model; if the model corresponding to the highest success rate is the same as the initial model, such as the initial model is CNN and its success rate is the highest after calculation, then no model replacement is performed.

[0087] In actual operation, the collection and processing of data must strictly follow the preset standards. For example, the collection frequency of environmental factor data needs to match the time resolution of the prediction model. If the model makes predictions on a daily basis, the environmental factor data needs to be collected once a day. During the data processing, if it is found that the water temperature data in a certain period shows abnormal values due to sensor failures, interpolation processing needs to be carried out using the data in adjacent periods to ensure the accuracy and reliability of the data input into the correlation equation.

[0088] In addition, the parameters of the correlation equation selected by the final model may need to be updated regularly according to historical data. For example, when new historical statistical data is accumulated, the swarm intelligence optimization technology can be reused to optimize the parameters of the correlation equation to adapt to environmental changes and the evolution of model performance. In this embodiment, it is assumed that the parameters of the correlation equation have been determined through the previous optimization process and do not need to be adjusted in this step.

[0089] The entire model selection process needs to ensure the integrity of the data and the accuracy of the calculation. For example, when calculating the success rate, it is necessary to clarify the definition method of the success rate, whether it is based on the mean square error, mean absolute error between the predicted value and the measured value, or other indicators. In this embodiment, it is assumed that the success rate is calculated based on the proportion of prediction errors within the range of ±10%. That is, if a model has 80 errors within ±10% in the past 100 predictions, its success rate is 80%.

[0090] Through the above steps, the most suitable current prediction model can be selected from the candidate prediction models according to the current actual data situation, providing a basis for subsequent biological quantity prediction and model optimization. During the implementation process, attention needs to be paid to the details of data processing and the rigor of the calculation logic to avoid model selection errors caused by data errors or calculation mistakes.

[0091] Embodiment 4: This embodiment elaborates on step S4 in detail. Set the second current statistical period, which needs to be determined according to the biological quantity change characteristics and model verification requirements. For example, assume that the target organism is a migratory fish, and its quantity is significantly affected by water temperature and water flow from April to June every year. The second current statistical period can be set from April 1, 2025 to June 30, 2025, with a duration of 3 months to cover the main activity cycle of this organism.

[0092] Set several time nodes within the second current statistical period to form the current time node set. For example, with an interval of 10 days, 9 time nodes are set within 3 months, which are April 10th, April 20th... June 20th. At the same time, set the model verification parameter type set, including water temperature parameter, salinity parameter, dissolved oxygen parameter, and biological density parameter. These parameters need to be able to intuitively reflect the application effect of the prediction model. For example, the water temperature parameter is collected once an hour by an underwater temperature sensor, and the daily average value is taken as the daily data.

[0093] Combine the model validation parameter type set with the current time node set to collect actual biomass data using the current prediction model. Assuming the current prediction model is the CNN model selected in step S3, on April 10th, using a combination of underwater camera monitoring and manual sampling, the target organism density data for that day was 85 individuals / m³. The average water temperature for that day was also recorded as 22°C, salinity as 28‰, and dissolved oxygen as 6.2 mg / L, forming a row of data in the current validation parameter matrix. This process continues until data collection for all time nodes is complete.

[0094] Collect average biomass data for several time points prior to the current prediction model to form a historical average biomass dataset. For example, select data from the three preceding years (April to June 2024) corresponding to the current time point and calculate the average biomass density at each time point. Assuming the biomass density on April 10, 2022, 2023, and 2024 was 80 individuals / m³, 78 individuals / m³, and 82 individuals / m³, respectively, the average biomass density at that time point in the historical average biomass dataset would be (80 + 78 + 82) / 3 = 80 individuals / m³.

[0095] Calculate the difference between the historical average population dataset and each row of data in the current validation parameter matrix to obtain the current effect difference dataset. For example, if the current biomass density on April 10th is 85 individuals / m³ and the historical average is 80 individuals / m³, the difference is 5 individuals / m³. If the current biomass density at another node is 70 individuals / m³ and the historical average is 80 individuals / m³, the difference is -10 individuals / m³.

[0096] Set the first and second effect difference thresholds. Assume the first threshold is 15 units / m³ and the second threshold is 5 units / m³. If the difference value in the current effect difference dataset is less than 5 units / m³, such as a node with a difference of 3 units / m³, the current prediction model is ineffective and you need to switch back to the initial prediction model (such as the LSTM model). If the difference value is greater than or equal to 5 units / m³ and less than 15 units / m³, such as 8 units / m³, proceed to step S5 to optimize the model structure.

[0097] If all values in the current effect difference dataset are greater than or equal to 15 individuals / m³, for example, if the difference at a particular node is 20 individuals / m³, then further predictions of biomass at future time points are required. A feedforward neural network model is used to predict biomass data for future time points (e.g., July 10th and July 20th). Assuming the environmental factor data and biomass density data in the current validation parameter matrix are input, a future validation parameter matrix is generated. For example, the biomass density on July 10th is predicted to be 90 individuals / m³, the water temperature to be 28°C, the salinity to be 29‰, and the dissolved oxygen to be 5.8 mg / L.

[0098] Calculate the difference values between the historical average quantity dataset and each row of data in the future verification parameter matrix to obtain the future effect difference dataset. Assume that the average biological density in the same period of history (July) is 85 individuals / m³, and the future predicted value is 90 individuals / m³, and the difference value is 5 individuals / m³. If there is a value less than 5 individuals / m³ in the future effect difference dataset, such as 4 individuals / m³, proceed to step S5; if all difference values are greater than or equal to 5 individuals / m³, no processing is required.

[0099] During the data collection process, the accuracy of the equipment needs to be ensured. For example, the temperature sensor needs to be calibrated regularly, and the shooting range of the underwater camera needs to cover the main area of the water intake to avoid data deviation caused by sampling blind spots. If the dissolved oxygen data at a certain time node is missing due to equipment failure, the value of the adjacent time node needs to be used for linear interpolation to ensure the integrity of the data.

[0100] In addition, the time span of the historical average quantity dataset needs to be selected reasonably. If the selected time is too short, it may not be able to reflect the long-term change law of the biological quantity; if it is too long, the environmental change may lead to a decrease in the reference value of the data. In this embodiment, the data of the previous 3 years are selected, taking into account both the timeliness and regularity of the data.

[0101] The entire determination process needs to be strictly executed according to the preset critical value. For example, when there are both values greater than 15 individuals / m³ and less than 5 individuals / m³ in the current effect difference dataset, the situation with a difference less than 5 individuals / m³ needs to be processed first, that is, switch back to the initial model to ensure the prediction accuracy.

[0102] Through the above steps, it is possible to determine whether the current prediction model needs to be adjusted based on the actually collected biological quantity data, providing a basis for subsequent model optimization. In implementation, attention needs to be paid to the standardization of data collection, the rigor of the calculation logic, and the rationality of the critical value setting to ensure the reliability of the determination result.

[0103] Example 5: This example details the implementation process of step S5. Assume that it is determined in step S4 that the structure of the current prediction model needs to be optimized. For example, the current prediction model is a CNN model, and the effect difference value in the second current statistical period is 8 individuals / m³, which is between the second effect difference critical value of 5 individuals / m³ and the first effect difference critical value of 15 individuals / m³. At this time, proceed to step S5.

[0104] Set the initial model structure optimization parameters, which include the number of model layers, the number of neurons, the convolutional kernel size, etc. For example, initially set the number of convolutional layers of the CNN model to 3 layers, the number of neurons in each layer to 64, 128, and 256 respectively, the convolutional kernel size to 3×3, the number of pooling layers to 2 layers, and the number of neurons in the fully connected layer to 100. Use these initial model structure optimization parameters to perform structure optimization processing on the current CNN model, adjust the network structure of the model, and store the optimized model.

[0105] After completing the structure optimization processing, collect the current biological quantity data according to the model verification parameter type set. For example, after the optimized model is put into use, at the time node of May 10, 2025, collect the density data of the target organism as 90 individuals / m³, and at the same time record the water temperature as 24°C, the salinity as 29‰, and the dissolved oxygen as 6.0 mg / L to form the current optimized verification parameter set. Calculate the difference value between the biological density data in the current optimized verification parameter set and the average biological quantity at the corresponding time node in the historical average quantity dataset. Assume that the average biological density on May 10 in the historical average quantity dataset is 80 individuals / m³, then the current optimized difference value is 10 individuals / m³.

[0106] Compare the current optimized difference value with the first effect difference critical value of 15 individuals / m³. Since 10 individuals / m³ is less than 15 individuals / m³, it is necessary to adjust the initial model structure optimization parameters. First, set the value range of the initial model structure optimization parameters. For example, the value range of the number of convolutional layers is 2 to 5 layers, the value range of the number of neurons in each layer is 32 to 512, the value range of the convolutional kernel size is 3×3 to 5×5, the value range of the number of pooling layers is 1 to 3 layers, and the value range of the number of neurons in the fully connected layer is 50 to 200.

[0107] Construct a model structure optimization population, which consists of multiple individuals, and each individual represents a set of possible model structure optimization parameters. Assume that an optimization population containing 50 individuals is constructed, and the parameter combinations of each individual are different. Set the maximum number of iterations of the model structure optimization population to 100 times, and the current number of iterations to 1 time.

[0108] Set the initial position of each individual in the model structure optimization population according to the current parameter value range. For example, the first individual has 3 convolutional layers, the number of neurons is 64, 128, 256, the convolutional kernel size is 3×3, the number of pooling layers is 2 layers, and the number of neurons in the fully connected layer is 100; the second individual has 4 convolutional layers, the number of neurons is 32, 64, 128, 256, the convolutional kernel size is 4×4, the number of pooling layers is 1 layer, and the number of neurons in the fully connected layer is 150, and so on, to obtain the second initial position set.

[0109] Construct a fitness evaluation function for the model structure optimization population, which is used to evaluate the quality of each individual's position. The design of the fitness evaluation function is based on the prediction error of the model. For example, the reciprocal of the mean square error between the predicted value and the measured value is used as the fitness value. The smaller the mean square error, the larger the fitness value.

[0110] Start the iteration. Before the iteration, set the current iteration number of the structure optimization to 1. In each iteration process, use the fitness evaluation function to calculate the fitness value of each individual's position in the model structure optimization population updated in the previous iteration. For example, for the model structure corresponding to each individual, use the current environmental factor dataset and the biological distribution dataset to make predictions, calculate the mean square error between the prediction result and the actual biological quantity data, and take its reciprocal as the fitness value.

[0111] Then update the position of each individual in the model structure optimization population updated in the previous iteration. The update method adopts the rules in the swarm intelligence optimization algorithm. For example, in the particle swarm optimization algorithm, each individual adjusts its position according to its own historical best position and the global best position of the population to generate a new parameter combination.

[0112] When the current iteration number of the structure optimization is equal to the maximum iteration number of the structure optimization, which is 100 times, stop the iteration to obtain the second final global optimal fitness and the second final global optimal position. Assume that after the iteration, the parameter combination of the optimal individual is 4 convolutional layers, the number of neurons is 64, 128, 256, 512, the convolutional kernel size is 3×3, the number of pooling layers is 2 layers, the number of neurons in the fully connected layer is 150, and the corresponding fitness value is 0.85.

[0113] Take the difference value corresponding to the second final global optimal fitness as the currently optimized difference value after optimization. Assume that according to the correspondence between the fitness evaluation function and the difference value, the fitness of 0.85 corresponds to a difference value of 16 / m³, which is greater than the first effect difference critical value of 15 / m³. At this time, use the parameter combination corresponding to the second final global optimal position to perform structure optimization processing on the current prediction model and store it to obtain the final prediction model structure.

[0114] If the currently optimized difference value after optimization does not reach the first effect difference critical value, for example, the fitness of 0.7 corresponds to a difference value of 14 / m³, which is less than 15 / m³, then return to continue the iteration until the currently optimized difference value after optimization is greater than or equal to the first effect difference critical value.

[0115] During the iteration process, if the position of an individual in a certain iteration exceeds the parameter value range, it needs to be corrected. For example, the number of convolutional layers can be adjusted to the maximum or minimum value within the value range. At the same time, to avoid falling into local optima, random factors can be introduced to increase the diversity of the population.

[0116] During the data collection process, if the dissolved oxygen data at a certain time node is abnormal, interpolation processing using adjacent node data is required to ensure the accuracy of the data input into the model. The historical average quantity dataset needs to be updated regularly to reflect the latest change trend of the biological quantity.

[0117] The entire optimization process needs to strictly record the parameter combinations and fitness values of each iteration, which is convenient for analyzing the optimization effect and adjusting the optimization strategy. The final obtained final prediction model structure needs to be verified multiple times to ensure that it can maintain good prediction performance under different environmental conditions.

[0118] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device.

[0119] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for predicting the number of organisms at nuclear power water intakes based on deep learning, characterized in that: It includes the following steps: S1. Collect the environmental factor data, biological species distribution data, and historical prediction deviation data in the nuclear power water intake area at different times to obtain the current environmental factor dataset, the current biological distribution dataset, and the current prediction deviation dataset; S2. Collect the historical application success rate data of each candidate prediction model, the environmental factor data of each candidate prediction model, the biological distribution data at different times, and the prediction deviation data, and construct the final model selection correlation equation; S3. Select the current prediction model according to the final model selection correlation equation, the current environmental factor dataset, the current biological distribution dataset, and the current prediction deviation dataset; S4. Determine whether parameter adjustment or structure optimization of the current prediction model is required by collecting the actual biological quantity data after using the current prediction model, and obtain the determination result; S5. Perform structure optimization processing on the current prediction model according to the determination result to obtain the final prediction model structure.

2. The method for predicting the biological quantity of a nuclear power water intake based on deep learning according to claim 1, characterized in that, The S1 includes the following steps: S11. Delimit the nuclear power water intake monitoring area to be predicted and the corresponding several types of biological species to obtain the target biological species set; then set several key parameter types affecting the prediction model selection to obtain the prediction model influence parameter type set; S12. Set the first current statistical period; cooperate with the first current statistical period and the prediction model influence parameter type set to obtain the environmental factor data, the biological distribution data at different times, and the prediction deviation data of each biological species in the target biological species set, to obtain the current environmental factor dataset, the current biological distribution dataset, and the current prediction deviation dataset.

3. The method for predicting the number of organisms at the nuclear power water intake based on deep learning according to claim 2, characterized in that, The S2 includes the following steps: S21. Set the historical statistical period; cooperate with the target biological species set and the prediction model influence parameter type set to collect the application success rate data of each candidate prediction model, the environmental factor data of each candidate prediction model, the biological distribution data at different times, and the prediction deviation data during the historical statistical period, to obtain the historical candidate model success rate dataset, the historical environmental factor dataset, the historical biological distribution dataset, and the historical prediction deviation dataset; S22. Use the historical candidate model success rate dataset, the historical environmental factor dataset, the historical biological distribution dataset, and the historical prediction deviation dataset to construct the final model selection correlation equation and the final environmental factor weight dataset.

4. The method for predicting the number of organisms at nuclear power water intakes based on deep learning according to claim 3 is characterized in that: The swarm intelligence optimization technology is used to construct the final model selection correlation equation and the final environmental factor weight dataset in S22.

5. A method for predicting the number of organisms at a nuclear power water intake based on deep learning according to claim 4, characterized in that, The S3 includes the following steps: S31. Set the current initial prediction model; substitute each data in the current environmental factor dataset, the current biological distribution dataset, the current prediction deviation dataset, and the final environmental factor weight dataset into the final model selection correlation equation for correlation calculation to obtain the current candidate model success rate dataset; S32. When the candidate prediction model corresponding to the highest success rate data in the current candidate model success rate dataset is the same as the current initial prediction model, the current prediction model is not replaced; otherwise, the candidate prediction model corresponding to the highest success rate data in the current candidate model success rate dataset is used as the current prediction model, and both are denoted as the current prediction model.

6. The method for predicting the number of organisms at the nuclear power water intake based on deep learning according to claim 5, wherein, The above S4 includes the following steps: S41. Set the second current statistical period; set a number of time nodes within the second current statistical period to obtain the current time node set; set a number of verification parameters that can reflect the application effect of the prediction model to obtain the model verification parameter type set; S42. Combine the model verification parameter type set and the current time node set to collect the actual biological quantity data after using the current prediction model to obtain the current verification parameter matrix; then collect the average biological quantity data of several time nodes before using the current prediction model to obtain the historical average quantity dataset; calculate the difference values between the historical average quantity dataset and each row of data in the current verification parameter matrix to obtain the current effect difference dataset; S43. Set the first effect difference critical value and the second effect difference critical value; when there is a current effect difference value in the current effect difference dataset that is less than the second effect difference critical value, switch the current prediction model back to the current initial prediction model again; when there is a current effect difference value in the current effect difference dataset that is greater than or equal to the second effect difference critical value and less than the first effect difference critical value, go to S5; otherwise, go to S44; S44. According to the current verification parameter matrix and using a feedforward neural network model, predict the biological quantity data of future time nodes to obtain the future verification parameter matrix; then calculate the difference values between the historical average quantity dataset and each row of data in the future verification parameter matrix to obtain the future effect difference dataset; when there is a future effect difference value in the future effect difference dataset that is less than the second effect difference critical value, go to S5; otherwise, do not process.

7. A method for predicting the number of organisms at the nuclear power water intake based on deep learning according to claim 6, characterized in that, The above S5 includes the following steps: S51. Set the initial model structure optimization parameter; use the initial model structure optimization parameter to perform structure optimization processing on the current prediction model and store it; after completing the structure optimization processing and storage of the current prediction model, collect the current biological quantity data according to the model verification parameter type set to obtain the current optimized verification parameter set; calculate the difference value between the current optimized verification parameter set and the historical average quantity dataset to obtain the current optimized difference value; S52. When the current optimized difference value is greater than or equal to the first effect difference critical value, use the initial model structure optimization parameter as the final prediction model structure; otherwise, adjust the initial model structure optimization parameter until the current optimized difference value is greater than or equal to the first effect difference critical value.

8. A method for predicting the number of organisms at a nuclear power water intake based on deep learning according to claim 7, characterized in that, The adjustment of the initial model structure optimization parameter in S52 includes the following steps: S521. Set the value range of the initial model structure optimization parameters to obtain the current parameter value range; construct the model structure optimization population; set the maximum number of iterations and the current number of iterations of the model structure optimization population, denoted as the maximum number of structure optimization iterations and the current number of structure optimization iterations respectively; S522. Set the initial position of each individual in the model structure optimization population according to the current parameter value range to obtain the second initial position set; S523. Construct the fitness evaluation function of the model structure optimization population; S524. Start the iteration. Before the iteration, set the current number of structure optimization iterations to 1; in each round of iteration, use the fitness evaluation function of the model structure optimization population to calculate the fitness value of the position of each individual in the model structure optimization population updated in the previous round of iteration and update the position of each individual in the model structure optimization population updated in the previous round of iteration; S525. When the current number of structure optimization iterations is equal to the maximum number of structure optimization iterations, stop the iteration to obtain the second final global optimal fitness and the second final global optimal position; otherwise, continue the iteration until the current number of structure optimization iterations is equal to the maximum number of structure optimization iterations; take the second final global optimal fitness as the optimized current optimized difference value; when the optimized current optimized difference value is greater than or equal to the first effect difference critical value, use the second final global optimal position to perform structure optimization processing on the current prediction model and store it to obtain the final prediction model structure; otherwise, return to S524 to continue the iteration until the optimized current optimized difference value is greater than or equal to the first effect difference critical value.

9. The method for predicting the number of organisms at the nuclear power water intake based on deep learning according to claim 8, wherein The model verification parameter type set includes water temperature parameters, salinity parameters, dissolved oxygen parameters, and biological density parameters.

10. The method for predicting the number of organisms at nuclear power water intakes based on deep learning according to claim 2, characterized in that: The delineation of the target biological species set includes the following steps: S111. Conduct a census of the biological community in the nuclear power water intake area and record the biological species with a frequency of occurrence exceeding the preset threshold in different time periods; S112. Combine the operating characteristics of the nuclear power cooling system to screen out the key biological species that may affect the water intake efficiency and summarize them to form the target biological species set.

Images