Nuclear power station main water supply system valve blocking fault prediction method and device, computer equipment, readable storage medium and program product

Abstract

The invention relates to a nuclear power station main water supply system valve blocking fault prediction method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: collecting data of a valve of a main water supply system of a nuclear power station during operation to form a plurality of samples; inputting the water supply flow data and the steam flow data in the sample into a to-be-trained stuck valve fault prediction network to obtain valve opening prediction data; calculating network loss by adopting a preset loss function based on the prediction data and the valve opening real data; optimizing parameters of the to-be-trained stuck valve fault prediction network according to the network loss; continuing training based on the stuck valve fault prediction network after parameter adjustment, and stopping until a stop condition is reached to obtain a trained stuck valve fault prediction network; and the trained valve jamming fault prediction network is used for predicting whether the nuclear power station main water supply system valve jamming fault occurs in real time. By adopting the method, the fault of the main feed water valve of the nuclear power station can be timely and accurately found.

Description

Technical Field

[0001] This application relates to the field of nuclear power technology, and in particular to a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for predicting valve jamming failure in the main feedwater system of a nuclear power plant. Background Technology

[0002] As one of the most important and sensitive control systems in a nuclear power plant, the stability of the evaporator water level control system is crucial for the safe and stable operation of the nuclear power unit. Statistics show that abnormal evaporator water level control is the leading cause of transient reactor shutdowns or automatic reactor shutdowns. Existing evaporator water level control systems incorporate redundancy in their front-end sensors and intermediate controllers, significantly improving the reliability of the front-end and intermediate stages of the control system. However, the non-redundant design of the actuators at the back end of the control system remains a weak point.

[0003] The actuator at the back end of the evaporator water level control system is the main feedwater valve, which is in principle a pneumatic regulating valve. Existing methods cannot detect in a timely and accurate manner the main feedwater valve is stuck at a specific opening degree or the valve is continuously opened too wide or closed too narrowly, leading to malfunctions in the nuclear power plant operation. Summary of the Invention

[0004] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium, and computer program product for predicting valve jamming in the main feedwater system of a nuclear power plant, which can detect the above-mentioned technical problems in a timely and accurate manner.

[0005] In a first aspect, this application provides a method for predicting valve jamming failures in the main feedwater system of a nuclear power plant, including:

[0006] Collect data on the operation of valves in the main feedwater system of the nuclear power plant, and obtain a training set from the data. The training set includes multiple samples, each of which includes feedwater flow data, steam flow data, and actual valve opening data.

[0007] The water flow rate data and steam flow rate data in the sample are input into the valve failure prediction network to be trained to obtain valve opening prediction data; based on the prediction data and the actual valve opening data, the network loss is calculated using a preset loss function; the parameters of the valve failure prediction network to be trained are optimized according to the network loss.

[0008] The training continues based on the parameter-adjusted valve jamming fault prediction network until the stopping condition is met, resulting in a trained valve jamming fault prediction network. The trained valve jamming fault prediction network is used to predict in real time whether valves in the main feedwater system of a nuclear power plant are jammed.

[0009] In one embodiment, data on the operation of valves in the main feedwater system of the nuclear power plant is collected, and a training set is obtained from the data, including:

[0010] Collect real data on valve opening degree, feedwater flow rate, and steam flow rate of the main feedwater system at various preset acquisition times during operation of the nuclear power plant's main feedwater system; take the valve opening degree data, feedwater flow rate data, and steam flow rate data of the steam generator corresponding to each preset acquisition time as a sample; and extract a first preset proportion of samples from all samples to construct a training set.

[0011] In one embodiment, before taking the valve opening data, the main water supply system's water flow rate data, and the steam generator's steam flow rate data corresponding to each preset acquisition time as a sample, the method further includes:

[0012] For each preset data collection time, the valve opening data, the main water supply system's water flow data, and the steam generator's steam flow data are cleaned according to a preset cleaning method to obtain cleaned data; the cleaned data is then used as a sample.

[0013] In one embodiment, according to a preset cleaning method, the valve opening data, water flow rate, and steam flow rate data are cleaned to obtain cleaning data, including:

[0014] Calculate a first value and a second value for the valve opening data, water flow data, and steam flow data respectively; based on the first value and the second value, calculate a standard score for the valve opening data, water flow data, and steam flow data respectively; data in the valve opening data, water flow data, and steam flow data whose standard score is greater than a preset value are regarded as abnormal data and removed to form cleaning data.

[0015] In one embodiment, the method further includes:

[0016] The water flow rate and steam flow rate data from the validation set are input into the valve fault prediction network to be trained to obtain the current validation valve opening. The validation set is constructed by taking samples from the remaining samples after the training set is built, according to a second preset proportion. Based on the current validation valve opening and the real valve opening data in the validation set, the current validation loss is calculated using a preset loss function. If the current validation loss is less than the validation loss of the previous iteration, the preset condition has not been met, and the next iteration of training is performed. If the current validation loss is not less than the validation loss of the previous iteration, the iteration training is stopped, and the valve fault prediction network trained in the previous iteration is used as the completed valve fault prediction network.

[0017] In one embodiment, the process by which the trained valve jamming fault prediction network predicts in real time whether a valve in the main feedwater system of a nuclear power plant is jammed includes:

[0018] Real-time acquisition of feedwater flow rate and steam flow rate data for valves in the main feedwater system of a nuclear power plant over a preset time period; inputting the feedwater flow rate and steam flow rate data for the preset time period into a trained valve jamming fault prediction network to obtain the predicted valve opening; obtaining the actual valve opening; and based on the predicted valve opening and the actual valve opening, determining whether the valves in the main feedwater system of the nuclear power plant are jammed.

[0019] Secondly, this application also provides a valve jamming fault prediction device for the main feedwater system of a nuclear power plant, comprising:

[0020] The collection module is used to collect data on the valves of the main feedwater system of the nuclear power plant during operation, and to obtain a training set from the data. The training set includes multiple samples, each of which includes feedwater flow data, steam flow data, and actual valve opening data.

[0021] The training module is used to input the water flow rate data and steam flow rate data from the samples into the valve failure prediction network to be trained, and obtain valve opening prediction data; based on the prediction data and the actual valve opening data, the network loss is calculated using a preset loss function; and the parameters of the valve failure prediction network to be trained are optimized according to the network loss.

[0022] The prediction module is used to continue training the valve jamming fault prediction network based on the adjusted parameters until the stopping condition is met, thus obtaining the trained valve jamming fault prediction network. The trained valve jamming fault prediction network is used to predict in real time whether the valves in the main feedwater system of the nuclear power plant are jammed.

[0023] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0024] Collect data on the operation of valves in the main feedwater system of the nuclear power plant, and obtain a training set from the data. The training set includes multiple samples, each of which includes feedwater flow data, steam flow data, and actual valve opening data.

[0025] The water flow rate data and steam flow rate data in the sample are input into the valve failure prediction network to be trained to obtain valve opening prediction data; based on the prediction data and the actual valve opening data, the network loss is calculated using a preset loss function; the parameters of the valve failure prediction network to be trained are optimized according to the network loss.

[0026] The training continues based on the parameter-adjusted valve jamming fault prediction network until the stopping condition is met, resulting in a trained valve jamming fault prediction network. The trained valve jamming fault prediction network is used to predict in real time whether valves in the main feedwater system of a nuclear power plant are jammed.

[0027] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0028] Collect data on the operation of valves in the main feedwater system of the nuclear power plant, and obtain a training set from the data. The training set includes multiple samples, each of which includes feedwater flow data, steam flow data, and actual valve opening data.

[0029] The water flow rate data and steam flow rate data in the sample are input into the valve failure prediction network to be trained to obtain valve opening prediction data; based on the prediction data and the actual valve opening data, the network loss is calculated using a preset loss function; the parameters of the valve failure prediction network to be trained are optimized according to the network loss.

[0030] The training continues based on the parameter-adjusted valve jamming fault prediction network until the stopping condition is met, resulting in a trained valve jamming fault prediction network. The trained valve jamming fault prediction network is used to predict in real time whether valves in the main feedwater system of a nuclear power plant are jammed.

[0031] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0032] Collect data on the operation of valves in the main feedwater system of the nuclear power plant, and obtain a training set from the data. The training set includes multiple samples, each of which includes feedwater flow data, steam flow data, and actual valve opening data.

[0033] The water flow rate data and steam flow rate data in the sample are input into the valve failure prediction network to be trained to obtain valve opening prediction data; based on the prediction data and the actual valve opening data, the network loss is calculated using a preset loss function; the parameters of the valve failure prediction network to be trained are optimized according to the network loss.

[0034] The training continues based on the parameter-adjusted valve jamming fault prediction network until the stopping condition is met, resulting in a trained valve jamming fault prediction network. The trained valve jamming fault prediction network is used to predict in real time whether valves in the main feedwater system of a nuclear power plant are jammed.

[0035] The aforementioned method, apparatus, computer equipment, computer-readable storage medium, and computer program product for predicting valve jamming faults in the main feedwater system of a nuclear power plant collect data on the valves of the main feedwater system during operation. A training set is obtained from this data, comprising multiple samples, each including feedwater flow rate data, steam flow rate data, and actual valve opening data. The feedwater flow rate and steam flow rate data from the samples are input into the valve jamming fault prediction network to be trained, yielding predicted valve opening data. Based on the predicted data and the actual valve opening data, a pre-defined loss function is used to calculate the network loss. The parameters of the valve jamming fault prediction network to be trained are optimized based on the network loss. The network continues training based on the adjusted parameters until a stopping condition is met, resulting in a trained valve jamming fault prediction network. This trained network is used to predict in real time whether valve jamming faults occur in the main feedwater system of a nuclear power plant. This method can detect faults in the main feedwater valves of a nuclear power plant in a timely and accurate manner. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a flowchart illustrating a method for predicting valve jamming failures in the main feedwater system of a nuclear power plant, as shown in one embodiment.

[0038] Figure 2 Here is a detailed flowchart of training a fault prediction network for a valve in one embodiment;

[0039] Figure 3 This is a structural block diagram of a valve jamming fault prediction device for the main feedwater system of a nuclear power plant in one embodiment.

[0040] Figure 4 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0042] In one embodiment, such as Figure 1As shown, a method for predicting valve jamming faults in the main feedwater system of a nuclear power plant is provided. This embodiment illustrates the method by applying it to a terminal. It is understood that this method can also be applied to a server, and to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0043] Step 102: Collect data on the operation of valves in the main feedwater system of the nuclear power plant, and obtain a training set from the data. The training set includes multiple samples, each of which includes feedwater flow data, steam flow data, and actual valve opening data.

[0044] Optionally, when collecting data, the water flow rate and steam flow rate of the first 25 seconds are collected as input data for the model, with a cycle of 26 seconds. The actual valve opening data of the 26th second is collected as the actual data corresponding to the water flow rate and steam flow rate of the first 25 seconds.

[0045] For example, data on the operation of valves in the main feedwater system of a nuclear power plant is collected in 26-second cycles, including feedwater flow rate data and steam flow rate data for the first 25 seconds, and actual valve opening data for the 26th second; a training set is obtained from the data, which includes multiple samples, each of which includes feedwater flow rate data, steam flow rate data, and actual valve opening data for one cycle.

[0046] Step 104: Input the water flow rate data and steam flow rate data from the sample into the valve failure prediction network to be trained to obtain valve opening prediction data; calculate the network loss using a preset loss function based on the prediction data and the actual valve opening data; optimize the parameters of the valve failure prediction network to be trained based on the network loss.

[0047] Among them, the water flow rate data and steam flow rate data are training data, the prediction data is the output of the valve fault prediction network to be trained, and the actual valve opening data is the actual data used to supervise the training of the network.

[0048] Optionally, the valve fault prediction network can be a CNN-LSTM-Attention hybrid neural network, and the preset loss function can be the mean squared error loss.

[0049] Specifically, the CNN-LSTM-Attention hybrid neural network uses ReLU activation function in the convolutional layers to ensure that the output is equal to the input if the input is greater than 0, otherwise the output is 0. Max pooling is used in the pooling layers to reduce computational cost by eliminating non-maximum values. It also helps extract local dependencies between different regions, retaining the most significant information. The obtained region vectors are used as input to the LSTM network, which uses 256 units to capture temporal dependencies in the time series. The Dropout layer uses a 0.5 kill rate to prevent overfitting. The attention mechanism layer uses four attention heads to dynamically adjust the network's attention level at different time steps. The fully connected layer serves as the output layer, using linear activation function to ensure the output is sequentially arranged linearly. The Adam optimizer is used in this example. The structure of the CNN-LSTM-Attention hybrid neural network is as follows: 1. 1D convolutional layer with 64 kernels and ReLU activation function; 2. Dropout layer with a 0.5 activation rate; 3. Bidirectional LSTM layer with 2*512 hidden units; 4. Dropout layer with a 0.5 activation rate; 5. Fully connected layer with an output dimension of 512; 6. Attention mechanism layer: 4 layers with 64 kernels and ReLU activation function; 7. 1D convolutional layer with 64 kernels and ReLU activation function; 8. Dropout layer with a 0.5 activation rate; 9. Bidirectional LSTM layer with 2*512 hidden units; 10. Fully connected layer with an output dimension of 512; 11. Attention mechanism layer with 64 kernels and ReLU activation function; 12. Fully connected output layer with an output dimension of 1.

[0050] Step 106: The valve jamming fault prediction network is trained again based on the adjusted parameters until the stopping condition is met, and the trained valve jamming fault prediction network is obtained. The trained valve jamming fault prediction network is used to predict in real time whether the valves of the main feedwater system of the nuclear power plant are jammed.

[0051] Optionally, the stopping condition can be that the model's loss no longer decreases during validation, or a preset number of iterations, such as 20 training rounds.

[0052] The aforementioned method, apparatus, computer equipment, computer-readable storage medium, and computer program product for predicting valve jamming faults in the main feedwater system of a nuclear power plant collect data on the valves of the main feedwater system during operation. A training set is obtained from this data, comprising multiple samples, each including feedwater flow rate data, steam flow rate data, and actual valve opening data. The feedwater flow rate and steam flow rate data from the samples are input into the valve jamming fault prediction network to be trained, yielding predicted valve opening data. Based on the predicted data and the actual valve opening data, a pre-defined loss function is used to calculate the network loss. The parameters of the valve jamming fault prediction network to be trained are optimized based on the network loss. The network continues training based on the adjusted parameters until a stopping condition is met, resulting in a trained valve jamming fault prediction network. This trained network is used to predict in real time whether valve jamming faults occur in the main feedwater system of a nuclear power plant. This method can detect faults in the main feedwater valves of a nuclear power plant in a timely and accurate manner.

[0053] In one exemplary embodiment, data on the operation of valves in the main feedwater system of a nuclear power plant is collected, and a training set is obtained from the data, including:

[0054] Collect real data on valve opening degree, feedwater flow rate, and steam flow rate of the main feedwater system at various preset acquisition times during operation of the nuclear power plant's main feedwater system; take the valve opening degree data, feedwater flow rate data, and steam flow rate data of the steam generator corresponding to each preset acquisition time as a sample; and extract a first preset proportion of samples from all samples to construct a training set.

[0055] Optionally, a 26-second cycle can be used. Taking one cycle as an example, each preset acquisition time is the valve opening data, the water supply flow data of the main water supply system, and the steam flow data of the steam generator for each second within the 26 seconds. Then, the water supply flow data and steam flow data of the first 25 seconds are used as input data, and the actual valve opening data of the 26th second is used as the actual value.

[0056] Optionally, the first preset ratio can be 80%.

[0057] For example, with a cycle of 26 seconds, the system collects real data on valve opening degree, feedwater flow rate, and steam flow rate of the main feedwater system and steam generator during operation. The feedwater flow rate and steam flow rate data of the first 25 seconds of each cycle are used as input data, and the real data on valve opening degree at the 26th second is used as the true value to form a sample. 80% of the samples are then used as the training set.

[0058] In this embodiment, by acquiring real data on valve opening, water flow data of the main water supply system, and steam flow data of the steam generator as training data, a network for predicting whether the valve state is abnormal can be trained.

[0059] In an exemplary embodiment, before taking the valve opening data, the main water supply system's water flow rate data, and the steam generator's steam flow rate data corresponding to each preset acquisition time as a sample, the method further includes:

[0060] For each preset data collection time, the valve opening data, the main water supply system's water flow data, and the steam generator's steam flow data are cleaned according to a preset cleaning method to obtain cleaned data; the cleaned data is then used as a sample.

[0061] Optionally, the preset cleaning method can be used to clean by calculating the mean and standard deviation, and then calculating a standard score based on the mean and standard deviation.

[0062] For example, the mean and standard deviation are calculated for valve opening data, water flow rate data, and steam flow rate data, respectively. Then, standard scores are calculated based on the mean and standard deviation. Data cleaning is performed on the corresponding valve opening data, water flow rate data, and steam flow rate data according to the standard scores to obtain cleaned data. The cleaned data is used as a sample.

[0063] In this embodiment, by cleaning the data, outliers can be removed, thereby improving the effectiveness and speed of network training.

[0064] In an exemplary embodiment, according to a preset cleaning method, the valve opening data, water flow rate, and steam flow rate data are cleaned to obtain cleaning data, including:

[0065] Calculate a first value and a second value for the valve opening data, water flow data, and steam flow data respectively; based on the first value and the second value, calculate a standard score for the valve opening data, water flow data, and steam flow data respectively; data in the valve opening data, water flow data, and steam flow data whose standard score is greater than a preset value are regarded as abnormal data and removed to form cleaning data.

[0066] Optionally, the first value can be the mean, and the second value can be the standard deviation; the preset value can be 3.

[0067] For example, the mean and standard deviation are calculated for valve opening data, feedwater flow rate data, and steam flow rate data, respectively. The formula is as follows:

[0068]

[0069] in, It's the amount of data. It is the first Individual data; standard deviation The formula is as follows:

[0070]

[0071] Based on the mean and standard deviation, standard scores were calculated for valve opening data, feedwater flow data, and steam flow data, respectively. The formula is as follows:

[0072]

[0073] Data with an absolute value greater than 3 in the standard score are considered outliers and removed. Then, the data after outlier removal is standardized to form clean data. The standardization formula is:

[0074] .

[0075] in, It is standardized data.

[0076] In this embodiment, by cleaning the data, more accurate data can be obtained for training.

[0077] In one exemplary embodiment, the method further includes:

[0078] The water flow rate and steam flow rate data from the validation set are input into the valve fault prediction network to be trained to obtain the current validation valve opening. The validation set is constructed by taking samples from the remaining samples after the training set is built, according to a second preset proportion. Based on the current validation valve opening and the real valve opening data in the validation set, the current validation loss is calculated using a preset loss function. If the current validation loss is less than the validation loss of the previous iteration, the preset condition has not been met, and the next iteration of training is performed. If the current validation loss is not less than the validation loss of the previous iteration, the iteration training is stopped, and the valve fault prediction network trained in the previous iteration is used as the completed valve fault prediction network.

[0079] After constructing the training set, 20% of the total samples remain. A second preset proportion of samples is selected from all the remaining samples as the validation set.

[0080] Optionally, the second preset ratio can be 50%, that is, half of all the remaining samples are selected as the validation set, which is equivalent to 10% of the total samples.

[0081] For example, when the network loss on the validation set no longer decreases as a stopping condition, after updating the network parameters in the current iteration of training, the water flow rate data and steam flow rate data from the validation set samples are input into the valve fault prediction network to be trained after updating the parameters to obtain the current validation valve opening. The validation set is constructed by taking 50% of the samples from the remaining samples after constructing the training set. Based on the current validation valve opening and the real valve opening data in the validation set, the current validation loss is calculated using a preset loss function. If the current validation loss is less than the validation loss of the previous iteration of training, the preset condition has not been met, and the next iteration of training is performed. If the current validation loss is not less than the validation loss of the previous iteration of training, the iteration training is stopped, and the valve fault prediction network trained in the previous iteration is used as the valve fault prediction network that has been trained.

[0082] In this embodiment, the data from the validation set is used to verify whether the valve failure prediction network has been trained to its best state, which can result in a valve failure prediction network with more accurate predictions.

[0083] In an exemplary embodiment, the process by which the trained valve jamming failure prediction network predicts in real time whether a valve in the main feedwater system of a nuclear power plant is jammed includes:

[0084] Real-time acquisition of feedwater flow rate and steam flow rate data for valves in the main feedwater system of a nuclear power plant over a preset time period; inputting the feedwater flow rate and steam flow rate data for the preset time period into a trained valve jamming fault prediction network to obtain the predicted valve opening; obtaining the actual valve opening; and based on the predicted valve opening and the actual valve opening, determining whether the valves in the main feedwater system of the nuclear power plant are jammed.

[0085] In this process, after obtaining the trained valve failure prediction network, the samples excluding the training and validation sets are used as the test set. The water flow rate and steam flow rate data from the test set are input into the trained network to obtain the maximum error between the predicted valve opening and the actual value. In practical applications, the presence of a failure is determined based on the maximum error, the predicted valve opening at the time of application, and the actual valve opening data.

[0086] For example, in actual use, a cyclic test is also performed with a 26-second cycle. The water flow rate and steam flow rate data for the first 25 seconds of a cycle are collected in real time. The collected data is input into the trained valve fault prediction network, which outputs the predicted valve opening at the 26th second. The actual valve opening data at the 26th second is collected, and the error between the predicted valve opening and the actual valve opening data is calculated. If the error is greater than twice the maximum error of the test set, the valve has a fault; if it is not greater, there is no fault.

[0087] In this embodiment, by applying the trained valve failure predictions for the future, the existence of valve failures can be assessed in real time and efficiently.

[0088] In one exemplary embodiment, such as Figure 2 As shown, the training steps of the valve failure prediction network include: collecting real data on valve opening degree, feedwater flow rate, and steam generator flow rate of the main feedwater system of the nuclear power plant during operation, with a cycle of 26 seconds; using the feedwater flow rate and steam flow rate data of the first 25 seconds of each cycle as input data, and the real valve opening degree data of the 26th second as the true value; calculating the mean and standard deviation of the valve opening degree data, feedwater flow rate data, and steam flow rate data respectively. The formula is as follows:

[0089]

[0090] in, It's the amount of data. It is the first Individual data; standard deviation The formula is as follows:

[0091]

[0092] Based on the mean and standard deviation, standard scores were calculated for valve opening data, feedwater flow data, and steam flow data, respectively. The formula is as follows:

[0093]

[0094] Data with an absolute value greater than 3 in the standard score are considered outliers and removed. Then, the data after outlier removal is standardized to form clean data. The standardization formula is:

[0095] .

[0096] in, The data is standardized. The first 25 seconds of the water flow rate and steam flow rate data from each cycle of the cleaning data, along with the actual valve opening data at the 26th second, are used as samples. 80% of the samples are used as the training set. The water flow rate and steam flow rate data from these samples are input into the valve failure prediction network to be trained, resulting in valve opening prediction data. Based on the prediction data and the actual valve opening data, a preset loss function is used to calculate the network loss. The parameters of the valve failure prediction network to be trained are then optimized based on the network loss. The valve failure prediction network, after parameter adjustment, continues training until a stopping condition is met: the network's loss on the validation set no longer decreases. After updating the network parameters in the current iteration, the feedwater flow rate and steam flow rate data from the validation set are input into the updated network to obtain the current validation valve opening. The validation set is constructed by taking 50% of the remaining samples after building the training set. Based on the current validation valve opening and the actual valve opening data in the validation set, a preset loss function is used to calculate the current validation loss. If the current validation loss is less than the validation loss of the previous iteration, the preset condition has not been met, and the next iteration begins. If the current validation loss is not less than the validation loss of the previous iteration, the iteration stops, and the valve failure prediction network trained in the previous iteration is used as the completed network. In actual use, a cyclic test is also performed with a 26-second cycle. The water flow rate and steam flow rate data for the first 25 seconds of a cycle are collected in real time. The collected data is then input into the trained valve fault prediction network, which outputs the predicted valve opening at the 26th second. The actual valve opening data at the 26th second is collected, and the error between the predicted valve opening and the actual valve opening data is calculated. If the error is greater than twice the maximum error of the test set, the valve is faulty; otherwise, there is no fault.

[0097] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0098] In one exemplary embodiment, such as Figure 3As shown, a valve jamming fault prediction device for the main feedwater system of a nuclear power plant is provided, comprising: a collection module, a training module, and a prediction module, wherein:

[0099] The collection module is used to collect data on the valves of the main feedwater system of the nuclear power plant during operation, and to obtain a training set from the data. The training set includes multiple samples, each of which includes feedwater flow data, steam flow data, and actual valve opening data.

[0100] The training module is used to input the water flow rate data and steam flow rate data from the samples into the valve failure prediction network to be trained, and obtain valve opening prediction data; based on the prediction data and the actual valve opening data, the network loss is calculated using a preset loss function; and the parameters of the valve failure prediction network to be trained are optimized according to the network loss.

[0101] The prediction module is used to continue training the valve jamming fault prediction network based on the adjusted parameters until the stopping condition is met, thus obtaining the trained valve jamming fault prediction network. The trained valve jamming fault prediction network is used to predict in real time whether the valves in the main feedwater system of the nuclear power plant are jammed.

[0102] In one embodiment, the collection module is further configured to:

[0103] Collect real data on valve opening degree, feedwater flow rate, and steam flow rate of the main feedwater system at various preset acquisition times during operation of the nuclear power plant's main feedwater system; take the valve opening degree data, feedwater flow rate data, and steam flow rate data of the steam generator corresponding to each preset acquisition time as a sample; and extract a first preset proportion of samples from all samples to construct a training set.

[0104] In one embodiment, the collection module is further configured to:

[0105] For each preset data collection time, the valve opening data, the main water supply system's water flow data, and the steam generator's steam flow data are cleaned according to a preset cleaning method to obtain cleaned data; the cleaned data is then used as a sample.

[0106] In one embodiment, the collection module is further configured to:

[0107] Calculate a first value and a second value for the valve opening data, water flow data, and steam flow data respectively; based on the first value and the second value, calculate a standard score for the valve opening data, water flow data, and steam flow data respectively; data in the valve opening data, water flow data, and steam flow data whose standard score is greater than a preset value are regarded as abnormal data and removed to form cleaning data.

[0108] In one embodiment, the training module is further configured to:

[0109] The water flow rate and steam flow rate data from the validation set are input into the valve fault prediction network to be trained to obtain the current validation valve opening. The validation set is constructed by taking samples from the remaining samples after the training set is built, according to a second preset proportion. Based on the current validation valve opening and the real valve opening data in the validation set, the current validation loss is calculated using a preset loss function. If the current validation loss is less than the validation loss of the previous iteration, the preset condition has not been met, and the next iteration of training is performed. If the current validation loss is not less than the validation loss of the previous iteration, the iteration training is stopped, and the valve fault prediction network trained in the previous iteration is used as the completed valve fault prediction network.

[0110] In one embodiment, the prediction module is further configured to:

[0111] Real-time acquisition of feedwater flow rate and steam flow rate data for valves in the main feedwater system of a nuclear power plant over a preset time period; inputting the feedwater flow rate and steam flow rate data for the preset time period into a trained valve jamming fault prediction network to obtain the predicted valve opening; obtaining the actual valve opening; and based on the predicted valve opening and the actual valve opening, determining whether the valves in the main feedwater system of the nuclear power plant are jammed.

[0112] The various modules in the aforementioned valve jamming fault prediction device for the main feedwater system of a nuclear power plant can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0113] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores feedwater flow data, steam flow data, and actual valve opening data. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a method for predicting valve jamming faults in the main feedwater system of a nuclear power plant.

[0114] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0115] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0116] Collect data on the operation of valves in the main feedwater system of the nuclear power plant, and obtain a training set from the data. The training set includes multiple samples, each of which includes feedwater flow data, steam flow data, and actual valve opening data.

[0117] The water flow rate data and steam flow rate data in the sample are input into the valve failure prediction network to be trained to obtain valve opening prediction data; based on the prediction data and the actual valve opening data, the network loss is calculated using a preset loss function; the parameters of the valve failure prediction network to be trained are optimized according to the network loss.

[0118] The training continues based on the parameter-adjusted valve jamming fault prediction network until the stopping condition is met, resulting in a trained valve jamming fault prediction network. The trained valve jamming fault prediction network is used to predict in real time whether valves in the main feedwater system of a nuclear power plant are jammed.

[0119] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0120] Collect real data on valve opening degree, feedwater flow rate, and steam flow rate of the main feedwater system at various preset acquisition times during operation of the nuclear power plant's main feedwater system; take the valve opening degree data, feedwater flow rate data, and steam flow rate data of the steam generator corresponding to each preset acquisition time as a sample; and extract a first preset proportion of samples from all samples to construct a training set.

[0121] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0122] For each preset data collection time, the valve opening data, the main water supply system's water flow data, and the steam generator's steam flow data are cleaned according to a preset cleaning method to obtain cleaned data; the cleaned data is then used as a sample.

[0123] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0124] Calculate a first value and a second value for the valve opening data, water flow data, and steam flow data respectively; based on the first value and the second value, calculate a standard score for the valve opening data, water flow data, and steam flow data respectively; data in the valve opening data, water flow data, and steam flow data whose standard score is greater than a preset value are regarded as abnormal data and removed to form cleaning data.

[0125] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0126] The water flow rate and steam flow rate data from the validation set are input into the valve fault prediction network to be trained to obtain the current validation valve opening. The validation set is constructed by taking samples from the remaining samples after the training set is built, according to a second preset proportion. Based on the current validation valve opening and the real valve opening data in the validation set, the current validation loss is calculated using a preset loss function. If the current validation loss is less than the validation loss of the previous iteration, the preset condition has not been met, and the next iteration of training is performed. If the current validation loss is not less than the validation loss of the previous iteration, the iteration training is stopped, and the valve fault prediction network trained in the previous iteration is used as the completed valve fault prediction network.

[0127] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0128] Real-time acquisition of feedwater flow rate and steam flow rate data for valves in the main feedwater system of a nuclear power plant over a preset time period; inputting the feedwater flow rate and steam flow rate data for the preset time period into a trained valve jamming fault prediction network to obtain the predicted valve opening; obtaining the actual valve opening; and based on the predicted valve opening and the actual valve opening, determining whether the valves in the main feedwater system of the nuclear power plant are jammed.

[0129] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0130] Collect data on the operation of valves in the main feedwater system of the nuclear power plant, and obtain a training set from the data. The training set includes multiple samples, each of which includes feedwater flow data, steam flow data, and actual valve opening data.

[0131] The water flow rate data and steam flow rate data in the sample are input into the valve failure prediction network to be trained to obtain valve opening prediction data; based on the prediction data and the actual valve opening data, the network loss is calculated using a preset loss function; the parameters of the valve failure prediction network to be trained are optimized according to the network loss.

[0132] The training continues based on the parameter-adjusted valve jamming fault prediction network until the stopping condition is met, resulting in a trained valve jamming fault prediction network. The trained valve jamming fault prediction network is used to predict in real time whether valves in the main feedwater system of a nuclear power plant are jammed.

[0133] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0134] Collect real data on valve opening degree, feedwater flow rate, and steam flow rate of the main feedwater system at various preset acquisition times during operation of the nuclear power plant's main feedwater system; take the valve opening degree data, feedwater flow rate data, and steam flow rate data of the steam generator corresponding to each preset acquisition time as a sample; and extract a first preset proportion of samples from all samples to construct a training set.

[0135] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0136] For each preset data collection time, the valve opening data, the main water supply system's water flow data, and the steam generator's steam flow data are cleaned according to a preset cleaning method to obtain cleaned data; the cleaned data is then used as a sample.

[0137] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0138] Calculate a first value and a second value for the valve opening data, water flow data, and steam flow data respectively; based on the first value and the second value, calculate a standard score for the valve opening data, water flow data, and steam flow data respectively; data in the valve opening data, water flow data, and steam flow data whose standard score is greater than a preset value are regarded as abnormal data and removed to form cleaning data.

[0139] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0140] The water flow rate and steam flow rate data from the validation set are input into the valve fault prediction network to be trained to obtain the current validation valve opening. The validation set is constructed by taking samples from the remaining samples after the training set is built, according to a second preset proportion. Based on the current validation valve opening and the real valve opening data in the validation set, the current validation loss is calculated using a preset loss function. If the current validation loss is less than the validation loss of the previous iteration, the preset condition has not been met, and the next iteration of training is performed. If the current validation loss is not less than the validation loss of the previous iteration, the iteration training is stopped, and the valve fault prediction network trained in the previous iteration is used as the completed valve fault prediction network.

[0141] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0142] Real-time acquisition of feedwater flow rate and steam flow rate data for valves in the main feedwater system of a nuclear power plant over a preset time period; inputting the feedwater flow rate and steam flow rate data for the preset time period into a trained valve jamming fault prediction network to obtain the predicted valve opening; obtaining the actual valve opening; and based on the predicted valve opening and the actual valve opening, determining whether the valves in the main feedwater system of the nuclear power plant are jammed.

[0143] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0144] Collect data on the operation of valves in the main feedwater system of the nuclear power plant, and obtain a training set from the data. The training set includes multiple samples, each of which includes feedwater flow data, steam flow data, and actual valve opening data.

[0145] The water flow rate data and steam flow rate data in the sample are input into the valve failure prediction network to be trained to obtain valve opening prediction data; based on the prediction data and the actual valve opening data, the network loss is calculated using a preset loss function; the parameters of the valve failure prediction network to be trained are optimized according to the network loss.

[0146] The training continues based on the parameter-adjusted valve jamming fault prediction network until the stopping condition is met, resulting in a trained valve jamming fault prediction network. The trained valve jamming fault prediction network is used to predict in real time whether valves in the main feedwater system of a nuclear power plant are jammed.

[0147] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0148] Collect real data on valve opening degree, feedwater flow rate, and steam flow rate of the main feedwater system at various preset acquisition times during operation of the nuclear power plant's main feedwater system; take the valve opening degree data, feedwater flow rate data, and steam flow rate data of the steam generator corresponding to each preset acquisition time as a sample; and extract a first preset proportion of samples from all samples to construct a training set.

[0149] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0150] For each preset data collection time, the valve opening data, the main water supply system's water flow data, and the steam generator's steam flow data are cleaned according to a preset cleaning method to obtain cleaned data; the cleaned data is then used as a sample.

[0151] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0152] Calculate a first value and a second value for the valve opening data, water flow data, and steam flow data respectively; based on the first value and the second value, calculate a standard score for the valve opening data, water flow data, and steam flow data respectively; data in the valve opening data, water flow data, and steam flow data whose standard score is greater than a preset value are regarded as abnormal data and removed to form cleaning data.

[0153] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0154] The water flow rate and steam flow rate data from the validation set are input into the valve fault prediction network to be trained to obtain the current validation valve opening. The validation set is constructed by taking samples from the remaining samples after the training set is built, according to a second preset proportion. Based on the current validation valve opening and the real valve opening data in the validation set, the current validation loss is calculated using a preset loss function. If the current validation loss is less than the validation loss of the previous iteration, the preset condition has not been met, and the next iteration of training is performed. If the current validation loss is not less than the validation loss of the previous iteration, the iteration training is stopped, and the valve fault prediction network trained in the previous iteration is used as the completed valve fault prediction network.

[0155] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0156] Real-time acquisition of feedwater flow rate and steam flow rate data for valves in the main feedwater system of a nuclear power plant over a preset time period; inputting the feedwater flow rate and steam flow rate data for the preset time period into a trained valve jamming fault prediction network to obtain the predicted valve opening; obtaining the actual valve opening; and based on the predicted valve opening and the actual valve opening, determining whether the valves in the main feedwater system of the nuclear power plant are jammed.

[0157] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0158] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0159] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for predicting valve jamming faults in the main feedwater system of a nuclear power plant, characterized in that, The method includes: Data on the operation of valves in the main feedwater system of a nuclear power plant is collected, and a training set is obtained from the data. The training set includes multiple samples, each of which includes feedwater flow data, steam flow data, and actual valve opening data. The water flow rate data and steam flow rate data from the sample are input into the valve failure prediction network to be trained to obtain valve opening prediction data; based on the prediction data and the actual valve opening data, the network loss is calculated using a preset loss function; the parameters of the valve failure prediction network to be trained are optimized according to the network loss. The training continues based on the parameter-adjusted valve jamming fault prediction network until a stopping condition is met, resulting in a trained valve jamming fault prediction network. The trained valve jamming fault prediction network is used to predict in real time whether valves in the main feedwater system of a nuclear power plant are jammed.

2. The method according to claim 1, characterized in that, The process of collecting operational data from the valves of the nuclear power plant's main feedwater system and obtaining a training set from that data includes: Collect real data on valve opening degree, feedwater flow rate, and steam flow rate of the main feedwater system and steam generator at various preset acquisition times during operation of the main feedwater system of the nuclear power plant. The valve opening data, the main water supply system's water flow data, and the steam generator's steam flow data corresponding to each preset data collection time are used as a sample. Select a first preset proportion of samples from all samples to construct a training set.

3. The method according to claim 2, characterized in that, Before taking the valve opening data, main water supply flow data, and steam generator steam flow data corresponding to each preset acquisition time as a sample, the method further includes: For each preset data collection time, the valve opening data, the water supply flow data of the main water supply system, and the steam flow data of the steam generator are collected, and the data are cleaned according to the preset cleaning method to obtain cleaned data. The cleaned data is used as a sample.

4. The method according to claim 3, characterized in that, The process involves cleaning the valve opening data, water flow rate, and steam flow rate data according to a preset cleaning method to obtain cleaning data, including: Calculate a first value and a second value for the valve opening data, the water supply flow data, and the steam flow data, respectively; Based on the first value and the second value, standard scores are calculated for the valve opening data, the water supply flow data and the steam flow data, respectively; Data in the valve opening data, water flow rate data, and steam flow rate data that have a standard score greater than a preset value are identified as abnormal data and removed to generate cleaning data.

5. The method according to claim 3, characterized in that, The method further includes: The water flow rate data and steam flow rate data in the validation set samples are input into the valve failure prediction network to be trained to obtain the current validation valve opening; the validation set is constructed by obtaining a second preset proportion of samples from the remaining samples after the training set is constructed. Based on the current verification valve opening and the actual valve opening data in the verification set, the current verification loss is calculated using a preset loss function. If the current verification loss is less than the verification loss of the previous iteration of training, then the preset condition has not been met, and the next iteration of training will proceed. If the current verification loss is not less than the verification loss of the previous iteration, then stop the iteration training and use the valve fault prediction network trained in the previous iteration as the completed valve fault prediction network.

6. The method according to claim 1, characterized in that, The process by which the trained valve jamming fault prediction network predicts in real time whether a valve in the main feedwater system of a nuclear power plant is jammed includes: Real-time acquisition of feedwater flow rate and steam flow rate data of valves in the main feedwater system of nuclear power plants within a preset time period; The water flow rate data and steam flow rate data for the preset time period are input into the trained valve failure prediction network to obtain the predicted valve opening. The actual valve opening is obtained, and based on the predicted valve opening and the actual valve opening, the result of whether the valve in the main feedwater system of the nuclear power plant is stuck is obtained.

7. A valve jamming fault prediction device for the main feedwater system of a nuclear power plant, characterized in that, The device includes: The collection module is used to collect data on the valves of the main feedwater system of the nuclear power plant during operation, and to obtain a training set from the data. The training set includes multiple samples, each of which includes feedwater flow data, steam flow data, and actual valve opening data. The training module is used to input the water flow rate data and steam flow rate data from the sample into the valve failure prediction network to be trained, and obtain valve opening prediction data; based on the prediction data and the actual valve opening data, the network loss is calculated using a preset loss function; and the parameters of the valve failure prediction network to be trained are optimized according to the network loss. The prediction module is used to continue training the valve jamming fault prediction network based on the adjusted parameters until a stopping condition is met, thus obtaining a trained valve jamming fault prediction network. The trained valve jamming fault prediction network is used to predict in real time whether the valves in the main feedwater system of the nuclear power plant are jammed.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images