Data operation and maintenance management method and system for smart power plant

Abstract

The invention provides a data operation and maintenance management method and system for an intelligent power plant. The method comprises the following steps: acquiring monitoring data transmitted from a plurality of monitoring points of a generator set at a current time point; constructing a monitoring matrix based on the monitoring data transmitted by each monitoring point; inputting the real-time monitoring matrix into a pre-trained first-stage processing model, wherein the first-stage processing model outputs a first prediction monitoring matrix for predicting a target time point; the first prediction monitoring matrix serves as input of second-stage processing, the second-stage processing comprises a plurality of sub-processing stages, and in each sub-processing stage, a monitoring matrix corresponding to a time slice before a target time point is calculated through a first processing model, and inputting the monitoring matrix into a second processing model to update a prediction monitoring matrix of a target time point, and taking the prediction monitoring matrix output by the last sub-processing stage as a second prediction monitoring matrix output by the second processing stage.

Description

Technical Field

[0001] The present invention relates to the technical field of smart power plants, and in particular to a data operation and maintenance management method and system for smart power plants. Background Art

[0002] The importance of smart power plant supervision through data is self-evident. It is a key link in the intelligent transformation of the modern energy field and plays a vital role in improving energy utilization efficiency, optimizing power supply structure, and ensuring safe and efficient operation of power plants.

[0003] First, the data supervision of smart power plants can realize real-time monitoring and analysis of the operating status of power plants. Through advanced sensors, instruments and other equipment, smart power plants can collect and store various data in real time during the operation of power plants, such as equipment status, energy consumption, environmental parameters, etc. These data provide a rich information basis for the operation and management of power plants. By analyzing these data, power plant managers can timely understand the operating status of power plants, discover potential safety hazards, and take corresponding measures to prevent and deal with them to avoid accidents.

[0004] In addition, data supervision can also promote intelligent decision-making in power plants. Based on the results of data analysis, smart power plants can provide scientific and reasonable decision-making suggestions for the operation and management of power plants. For example, in terms of power generation dispatching, smart power plants can predict changes in power demand in the future based on real-time and historical data, optimize power generation plans and dispatching strategies, and improve power generation efficiency and economic benefits. In terms of equipment maintenance, smart power plants can monitor and analyze equipment operation data to formulate maintenance plans in advance and reduce the impact of equipment failures on power plant operations.

[0005] However, the maintenance solutions of existing technologies usually require prediction of the future operating parameters of the power plant, but they often require a long period of historical data as a reference, and the processing efficiency is low.

[0006] In view of this, the present invention is proposed. Summary of the invention

[0007] The purpose of the present invention is to provide a data operation and maintenance management method and system for a smart power plant. This scheme first performs coarse-grained prediction through a first-stage processing model, and then performs granular processing through multiple sub-processing stages of the second-stage processing. On the one hand, it does not require historical data of a long-term window, and on the other hand, it also ensures the accuracy of the prediction.

[0008] The present invention provides a data operation and maintenance management method for a smart power plant, the method comprising the following steps:

[0009] Acquire monitoring data transmitted from multiple monitoring points of the generator set at the current time point;

[0010] A real-time monitoring vector is constructed based on the monitoring data transmitted from each monitoring point, and a monitoring matrix is ​​constructed based on the real-time monitoring vectors of all monitoring points;

[0011] Inputting the real-time monitoring matrix into a pre-trained first-stage processing model, the first-stage processing model outputs a first prediction monitoring matrix for predicting a target time point;

[0012] The first prediction monitoring matrix is ​​used as the input of the second stage processing, and the second stage processing includes multiple sub-processing stages. In each sub-processing stage, the monitoring matrix corresponding to the time slice before the target time point is calculated by the first processing model, and the monitoring matrix is ​​input into the second processing model to update the prediction monitoring matrix of the target time point, and the prediction monitoring matrix output by the last sub-processing stage is used as the second prediction monitoring matrix output by the second processing stage.

[0013] The above scheme is adopted. This scheme only needs the monitoring data of the current time point as input data, and constructs a monitoring matrix with the monitoring data of the current time point. In the prediction process, the first stage processing model is used to perform coarse-grained prediction to obtain the first prediction monitoring matrix of preliminary prediction, and then the granularity is refined through multiple sub-processing stages of the second stage processing. In multiple sub-processing stages, the monitoring matrix of the time points before the target time point is continuously predicted, and the monitoring matrix of the target time point is generated again based on the monitoring matrix. The prediction results can be continuously refined to ensure the accuracy of the prediction.

[0014] In some embodiments of the present invention, in the processing step of inputting the real-time monitoring matrix into the pre-trained first-stage processing model, the first-stage processing model outputs a first prediction monitoring matrix for predicting the target time point:

[0015] Determine the number of loop processing times based on the number of time slices between the current time point and the target time point;

[0016] In each cycle of processing, the monitoring matrix at the next time point is predicted through the first-stage processing model.

[0017] In some embodiments of the present invention, the real-time monitoring matrix is ​​input into a pre-trained first-stage processing model, and the first-stage processing model outputs a loop processing of a first prediction monitoring matrix for predicting a target time point. Each loop processing corresponds to a monitoring matrix of a time slice, and each loop processing takes the monitoring matrix output by the previous loop processing as input, and takes the monitoring matrix output by the last loop processing as the first prediction monitoring matrix.

[0018] By adopting the above scheme, the first stage processing of this scheme is looped through the first stage processing model, and the monitoring matrix of the next time slice is predicted in each processing. The granularity of the first stage processing can be improved through multiple loop processing to ensure the processing precision.

[0019] In some embodiments of the present invention, the second stage of processing is composed of multiple sub-processing stages, each of which is completed by the first processing model and the second processing model:

[0020] During the processing of the first processing model, the real-time monitoring matrix and the predicted monitoring matrix of the currently updated target time point are combined and input into the first processing model, and the first processing model outputs a transition monitoring matrix for predicting the time slice between the current time point and the target time point;

[0021] The transition monitoring matrix is ​​input into the second processing model, and the second processing model outputs an updated prediction monitoring matrix at the target time point.

[0022] In some embodiments of the present invention, the number of sub-processing stages in the second stage processing corresponds to the number of cycles of the first processing stage, and the sub-processing stages correspond to the loop processing order of the first processing stage based on the order in the second stage processing. The first processing model outputs a transition monitoring matrix of a time slice corresponding to the loop processing order of the first processing stage, and the second processing model outputs an updated prediction monitoring matrix of the target time point based on the transition monitoring matrix of the time slice.

[0023] By adopting the above scheme, the sub-processing stage of the second stage processing of this scheme corresponds to the number of loop processing times of the first stage processing. Since each loop processing corresponds to a time slice, this scheme makes the sub-processing stage correspond one-to-one with the loop processing, so that the first processing model of the sub-processing stage outputs the transition monitoring matrix of the time slice corresponding to the loop processing, which can represent the smooth generation at different time steps, thereby capturing the time dynamics and propagation characteristics, and generating the prediction monitoring matrix of the target time point one by one, which effectively solves the problem that the prior art relies on a long observation window.

[0024] In some embodiments of the present invention, in the steps of constructing a real-time monitoring vector based on the monitoring data transmitted from each monitoring point and constructing a monitoring matrix based on the real-time monitoring vectors of all monitoring points, the monitoring data transmitted from each monitoring point is encoded to obtain a real-time monitoring vector of the corresponding monitoring point, and the real-time monitoring vectors are arranged in sequence based on the numbers of the respective monitoring points to obtain the monitoring matrix.

[0025] In some embodiments of the present invention, the method further comprises health determination, and the health determination step comprises:

[0026] Inputting the second prediction monitoring matrix into a pre-trained health determination model, outputting a health value based on the health determination model, and determining the overall health of the generator set based on the health value;

[0027] The overall health is compared with a health threshold to determine whether the generator set is overall healthy.

[0028] By adopting the above scheme, this scheme determines the overall health of the generator set through the second predictive monitoring matrix obtained by continuous refinement processing. It can determine the overall health of the generator set at a future target time point through the monitoring data at one time point, thereby ensuring the accuracy of health judgment.

[0029] In some embodiments of the present invention, the step of determining the health status further includes:

[0030] If the generator set is in a non-healthy state as a whole, the second prediction monitoring matrix is ​​split corresponding to each monitoring point to obtain a prediction monitoring vector corresponding to each monitoring point;

[0031] Based on the predicted monitoring vector, it is determined whether the corresponding monitoring point is in a waiting-for-inspection state.

[0032] In some embodiments of the present invention, in the step of splitting the second prediction monitoring matrix corresponding to each monitoring point to obtain a prediction monitoring vector corresponding to each monitoring point; and determining whether the corresponding monitoring point is in a waiting inspection state based on the prediction monitoring vector:

[0033] Obtain the power generation data and operating frequency data of the generator set at multiple historical time points;

[0034] Based on the power generation data and operating frequency data of multiple historical time points, a power prediction vector and a frequency prediction vector are respectively constructed, and based on the power prediction vector and the frequency prediction vector, a power prediction value and a frequency prediction value at a target time point are predicted;

[0035] The power prediction value and the frequency prediction value are constructed as a supplementary data group, and are spliced ​​with the prediction monitoring vector split out of the second prediction monitoring matrix. The spliced ​​vector is input into the judgment model corresponding to each monitoring point, and the judgment model is used to determine whether the monitoring point is in a state to be inspected.

[0036] Using the above scheme, when it is determined that the current generator set is in a non-overall healthy state, this scheme will split the prediction monitoring vectors corresponding to each monitoring point in the second prediction monitoring matrix. However, the separate environmental monitoring data often lacks comparison, and it is difficult to ensure the accuracy of the judgment. Then, this scheme further predicts the power generation and operating frequency of the generator set through historical power generation data and operating frequency data, and splices the predicted power generation and operating frequency into each prediction monitoring vector to provide a comparison of the load of the generator set. Then, the neural network model is used to determine whether the monitoring point is in a state to be inspected, thereby improving the accuracy of the judgment.

[0037] Another aspect of the present invention also relates to a data operation and maintenance management system for a smart power plant, the system comprising a computer device, the computer device comprising a processor and a memory, the memory storing computer instructions, the processor being used to execute the computer instructions stored in the memory, and when the computer instructions are executed by the processor, the system implements the steps implemented by the method.

[0038] In summary, the present invention has the following beneficial effects:

[0039] 1. This solution only needs the monitoring data of the current time point as input data, and constructs a monitoring matrix with the monitoring data of the current time point. In the prediction process, the first-stage processing model is used to perform coarse-grained prediction to obtain the first prediction monitoring matrix of preliminary prediction, and then the second-stage processing multiple sub-processing stages are used to perform granularity refinement. In multiple sub-processing stages, the monitoring matrix of the time point before the target time point is continuously predicted, and the monitoring matrix of the target time point is generated again based on the monitoring matrix, which can continuously refine the prediction results and ensure the accuracy of the prediction;

[0040] 2. The first stage processing of this scheme is cyclically processed through the first stage processing model. In each processing, the monitoring matrix of the next time slice is predicted. The granularity of the first stage processing can be improved through multiple cyclic processing to ensure the processing precision.

[0041] 3. The sub-processing stage of the second stage of the scheme corresponds to the number of loop processing of the first stage. Since each loop processing corresponds to a time slice, the scheme makes the sub-processing stage correspond to the loop processing one by one, so that the first processing model of the sub-processing stage outputs the transition monitoring matrix of the time slice corresponding to the loop processing, which can represent the smooth generation at different time steps, thereby capturing the time dynamics and propagation characteristics, and generating the prediction monitoring matrix of the target time point one by one, which effectively solves the problem of the prior art relying on a long observation window;

[0042] 4. When it is determined that the current generator set is in a non-overall healthy state, this scheme will split the prediction monitoring vectors corresponding to each monitoring point in the second prediction monitoring matrix. However, the separate environmental monitoring data often lacks comparison, and it is difficult to ensure the accuracy of the judgment. Then, this scheme further predicts the power generation and operating frequency of the generator set through historical power generation data and operating frequency data, and splices the predicted power generation and operating frequency into each prediction monitoring vector to provide a comparison of the load of the generator set, and then determines whether the monitoring point is in a state to be inspected through a neural network model, thereby improving the accuracy of the judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] Figure 1 It is a schematic diagram of a first implementation mode of the data operation and maintenance management method for a smart power plant according to the present invention;

[0045] Figure 2 It is a schematic diagram of a second implementation mode of the data operation and maintenance management method for a smart power plant according to the present invention;

[0046] Figure 3 It is a schematic diagram of a third implementation mode of the data operation and maintenance management method for a smart power plant according to the present invention;

[0047] Figure 4 It is a schematic diagram of the fourth implementation mode of the data operation and maintenance management method for a smart power plant according to the present invention. DETAILED DESCRIPTION

[0048] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of systems and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0049] The terms used in the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The singular forms "a", "the" and "the" used in the present invention and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0050] like Figure 1 As shown, the present invention provides a data operation and maintenance management method for a smart power plant, the method comprising the steps of:

[0051] Step s100, acquiring monitoring data transmitted from multiple monitoring points of the generator set at the current time point;

[0052] In a specific implementation process, the monitoring data includes temperature data, pressure data and sound data of the monitoring point.

[0053] Step S200, constructing a real-time monitoring vector based on the monitoring data transmitted from each monitoring point, and constructing a monitoring matrix based on the real-time monitoring vectors of all monitoring points;

[0054] In the specific implementation process, in the steps of constructing a real-time monitoring vector based on the monitoring data transmitted from each monitoring point and constructing a monitoring matrix based on the real-time monitoring vectors of all monitoring points, if the monitoring data includes temperature data, pressure data and sound data of the monitoring point, the temperature data, pressure data and sound data of each monitoring point are first constructed into a triplet, and the temperature data, pressure data and sound data in the triplet are respectively encoded into vectors of preset lengths, and the three vectors are connected as the real-time monitoring vector of one monitoring point.

[0055] Step S300, inputting the real-time monitoring matrix into the pre-trained first-stage processing model, the first-stage processing model outputting a first prediction monitoring matrix for predicting the target time point;

[0056] In the specific implementation process, the network structure of the first-stage processing model can be a sequentially connected convolution layer, attention layer, embedding layer, attention layer, embedding layer, attention layer and convolution layer; the structure of the second processing model processed in each sub-stage of the second stage processing is the same as the network structure of the first-stage processing model; both models are trained using a preset training data set. Specifically, the training process is completed by calculating the loss value of the loss function and performing back propagation.

[0057] In some embodiments of the present invention, the loss function may adopt a mean square error loss function:

[0058] Specifically, the mean square error loss function is calculated using the following formula:

[0059]

[0060] Among them, MSE represents the value of mean square error, N is the number of samples, and y i represents the true value of the i-th sample, Represents the predicted value of the i-th sample.

[0061] Step S400, using the first prediction monitoring matrix as the input of the second stage processing, the second stage processing includes multiple sub-processing stages, in each sub-processing stage, the monitoring matrix corresponding to the time slice before the target time point is calculated by the first processing model, and the monitoring matrix is ​​input into the second processing model to update the prediction monitoring matrix of the target time point, and the prediction monitoring matrix output by the last sub-processing stage is used as the second prediction monitoring matrix output by the second processing stage.

[0062] In the specific implementation process, the first processing model can adopt a convolutional neural network model, which is also completed using a preset training data set. The second processing model can also be implemented using a convolutional neural network model.

[0063] In some embodiments of the present invention, in the step of calculating the monitoring matrix corresponding to the time slice before the target time point through the first processing model, the time slice before the target time point can be any time slice between the target time point and the current time point.

[0064] In a specific implementation process, the time slice is a time period that is evenly divided into the time period between the target time point and the current time point.

[0065] The above scheme is adopted. This scheme only needs the monitoring data of the current time point as input data, and constructs a monitoring matrix with the monitoring data of the current time point. In the prediction process, the first stage processing model is used to perform coarse-grained prediction to obtain the first prediction monitoring matrix of preliminary prediction, and then the granularity is refined through multiple sub-processing stages of the second stage processing. In multiple sub-processing stages, the monitoring matrix of the time points before the target time point is continuously predicted, and the monitoring matrix of the target time point is generated again based on the monitoring matrix. The prediction results can be continuously refined to ensure the accuracy of the prediction.

[0066] like Figure 2 As shown, in some embodiments of the present invention, in the processing step of inputting the real-time monitoring matrix into the pre-trained first-stage processing model, the first-stage processing model outputs a first prediction monitoring matrix for predicting the target time point:

[0067] Step S310, determining the number of loop processing times based on the number of time slices between the current time point and the target time point;

[0068] Step S320: In each cycle of processing, the monitoring matrix at the next time point is predicted by the first stage processing model.

[0069] In the specific implementation process, the time slice can be fine-grained or coarse-grained. The finer the granularity of the slicing, the higher the processing accuracy, but the more times the loop processing is performed, the slower the processing speed. The specific granularity of the slicing can be set according to actual conditions.

[0070] In some embodiments of the present invention, the real-time monitoring matrix is ​​input into a pre-trained first-stage processing model, and the first-stage processing model outputs a loop processing of a first prediction monitoring matrix for predicting a target time point. Each loop processing corresponds to a monitoring matrix of a time slice, and each loop processing takes the monitoring matrix output by the previous loop processing as input, and takes the monitoring matrix output by the last loop processing as the first prediction monitoring matrix.

[0071] In the specific implementation process, since the length of each time slice is equal, the time length of each backward prediction is the same. Therefore, the first stage processing of this scheme can be looped through a first stage processing model, without the need to add additional models, thereby reducing the training burden of the model.

[0072] By adopting the above scheme, the first stage processing of this scheme is looped through the first stage processing model, and the monitoring matrix of the next time slice is predicted in each processing. The granularity of the first stage processing can be improved through multiple loop processing to ensure the processing precision.

[0073] In some embodiments of the present invention, the second stage of processing is composed of multiple sub-processing stages, each of which is completed by the first processing model and the second processing model:

[0074] During the processing of the first processing model, the real-time monitoring matrix and the predicted monitoring matrix of the currently updated target time point are combined and input into the first processing model, and the first processing model outputs a transition monitoring matrix for predicting the time slice between the current time point and the target time point;

[0075] The transition monitoring matrix is ​​input into the second processing model, and the second processing model outputs an updated prediction monitoring matrix at the target time point.

[0076] In the specific implementation process, in the step of combining the real-time monitoring matrix and the predicted monitoring matrix of the currently updated target time point and inputting them into the first processing model, the real-time monitoring matrix and the predicted monitoring matrix of the currently updated target time point are spliced.

[0077] In the specific implementation process, this scheme first generates a monitoring matrix of an intermediate time slice in each sub-processing stage, and then predicts the predicted monitoring matrix of the target time point through the monitoring matrix of the time slice. By continuously performing intermediate segmentation and continuously realizing refined processing, the prediction accuracy is guaranteed without the need for real data of a long time window.

[0078] In some embodiments of the present invention, the number of sub-processing stages in the second stage processing corresponds to the number of cycles of the first processing stage, and the sub-processing stages correspond to the loop processing order of the first processing stage based on the order in the second stage processing. The first processing model outputs a transition monitoring matrix of a time slice corresponding to the loop processing order of the first processing stage, and the second processing model outputs an updated prediction monitoring matrix of the target time point based on the transition monitoring matrix of the time slice.

[0079] In the specific implementation process, this solution makes the number of sub-processing stages in the second stage correspond to the number of cycles in the first stage, so as to achieve granularity alignment of the two processing stages. Through the alignment of granularity, the prediction of each time slice can be refined twice, further ensuring the processing accuracy.

[0080] By adopting the above scheme, the sub-processing stage of the second stage processing of this scheme corresponds to the number of loop processing times of the first stage processing. Since each loop processing corresponds to a time slice, this scheme makes the sub-processing stage correspond one-to-one with the loop processing, so that the first processing model of the sub-processing stage outputs the transition monitoring matrix of the time slice corresponding to the loop processing, which can represent the smooth generation at different time steps, thereby capturing the time dynamics and propagation characteristics, and generating the prediction monitoring matrix of the target time point one by one, which effectively solves the problem that the prior art relies on a long observation window.

[0081] In some embodiments of the present invention, in the steps of constructing a real-time monitoring vector based on the monitoring data transmitted from each monitoring point and constructing a monitoring matrix based on the real-time monitoring vectors of all monitoring points, the monitoring data transmitted from each monitoring point is encoded to obtain a real-time monitoring vector of the corresponding monitoring point, and the real-time monitoring vectors are arranged in sequence based on the numbers of the respective monitoring points to obtain the monitoring matrix.

[0082] like Figure 3 As shown, in some embodiments of the present invention, the steps of the method further include health determination, and the steps of health determination include:

[0083] Step S510, inputting the second prediction monitoring matrix into a pre-trained health determination model, outputting a health value based on the health determination model, and determining the overall health of the generator set based on the health value;

[0084] Step S520, comparing the overall health with a health threshold to determine whether the generator set is healthy as a whole.

[0085] In the specific implementation process, the health determination model can be a convolutional neural network model or a multi-layer perceptron model, etc.

[0086] In a specific implementation process, if the overall health is lower than the health threshold, it is determined that the generator set is in a non-overall healthy state.

[0087] By adopting the above scheme, this scheme determines the overall health of the generator set through the second predictive monitoring matrix obtained by continuous refinement processing. It can determine the overall health of the generator set at a future target time point through the monitoring data at one time point, thereby ensuring the accuracy of health judgment.

[0088] like Figure 4 As shown, in some embodiments of the present invention, the step of determining the health level further includes:

[0089] Step S530: if the generator set is in a non-healthy state as a whole, the second prediction monitoring matrix is ​​split corresponding to each monitoring point to obtain a prediction monitoring vector corresponding to each monitoring point;

[0090] Step S540: determine whether the corresponding monitoring point is in a waiting-for-inspection state based on the predicted monitoring vector.

[0091] In a specific implementation process, each row of the second prediction monitoring matrix corresponds to a monitoring point, and the data of the row corresponding to each monitoring point is split to obtain the prediction monitoring vector corresponding to each monitoring point.

[0092] In some embodiments of the present invention, in the step of splitting the second prediction monitoring matrix corresponding to each monitoring point to obtain a prediction monitoring vector corresponding to each monitoring point; and determining whether the corresponding monitoring point is in a waiting inspection state based on the prediction monitoring vector:

[0093] Obtain the power generation data and operating frequency data of the generator set at multiple historical time points;

[0094] Based on the power generation data and operating frequency data of multiple historical time points, a power prediction vector and a frequency prediction vector are respectively constructed, and based on the power prediction vector and the frequency prediction vector, a power prediction value and a frequency prediction value at a target time point are predicted;

[0095] The power prediction value and the frequency prediction value are constructed as a supplementary data group, and are spliced ​​with the prediction monitoring vector split out of the second prediction monitoring matrix. The spliced ​​vector is input into the judgment model corresponding to each monitoring point, and the judgment model is used to determine whether the monitoring point is in a state to be inspected.

[0096] In the specific implementation process, in the step of respectively constructing a power prediction vector and a frequency prediction vector based on the power generation data and the operating frequency data at multiple historical time points, and predicting the power prediction value and the frequency prediction value at the target time point based on the power prediction vector and the frequency prediction vector, the power generation data and the operating frequency data at each time point are sequentially arranged based on the timestamps in the power generation data and the operating frequency data to obtain the power prediction vector and the frequency prediction vector; and the prediction is performed through the pre-trained LSTM model;

[0097] LSTM (Long Short-Term Memory) is a special recurrent neural network (RNN).

[0098] Using the above scheme, when it is determined that the current generator set is in a non-overall healthy state, this scheme will split the prediction monitoring vectors corresponding to each monitoring point in the second prediction monitoring matrix. However, the separate environmental monitoring data often lacks comparison, and it is difficult to ensure the accuracy of the judgment. Then, this scheme further predicts the power generation and operating frequency of the generator set through historical power generation data and operating frequency data, and splices the predicted power generation and operating frequency into each prediction monitoring vector to provide a comparison of the load of the generator set. Then, the neural network model is used to determine whether the monitoring point is in a state to be inspected, thereby improving the accuracy of the judgment.

[0099] Another aspect of the present invention also relates to a data operation and maintenance management system for a smart power plant, the system comprising a computer device, the computer device comprising a processor and a memory, the memory storing computer instructions, the processor being used to execute the computer instructions stored in the memory, and when the computer instructions are executed by the processor, the system implements the steps implemented by the method.

[0100] The embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the aforementioned data operation and maintenance management method for a smart power plant is implemented. The computer-readable storage medium can be a tangible storage medium, such as a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a floppy disk, a hard disk, a removable storage disk, a CD-ROM, or any other form of storage medium known in the technical field.

[0101] It should be understood by those skilled in the art that the exemplary components, systems and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software or a combination of the two. Whether it is performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present invention are programs or code segments used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link via a data signal carried in a carrier.

[0102] It should be clear that the present invention is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present invention.

[0103] In the present invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with features of other embodiments or replace features of other embodiments.

[0104] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the embodiments of the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A data operation and maintenance management method for a smart power plant, characterized in that: The steps of the method include: Acquire monitoring data transmitted from multiple monitoring points of the generator set at the current time point; A real-time monitoring vector is constructed based on the monitoring data transmitted from each monitoring point, and a monitoring matrix is ​​constructed based on the real-time monitoring vectors of all monitoring points; Inputting the real-time monitoring matrix into a pre-trained first-stage processing model, the first-stage processing model outputs a first prediction monitoring matrix for predicting a target time point; The first prediction monitoring matrix is ​​used as the input of the second stage processing, and the second stage processing includes multiple sub-processing stages. In each sub-processing stage, the monitoring matrix corresponding to the time slice before the target time point is calculated by the first processing model, and the monitoring matrix is ​​input into the second processing model to update the prediction monitoring matrix of the target time point, and the prediction monitoring matrix output by the last sub-processing stage is used as the second prediction monitoring matrix output by the second processing stage.

2. The data operation and maintenance management method for a smart power plant according to claim 1, characterized in that: In the processing step of inputting the real-time monitoring matrix into the pre-trained first-stage processing model, the first-stage processing model outputs a first prediction monitoring matrix for predicting the target time point: Determine the number of loop processing times based on the number of time slices between the current time point and the target time point; In each cycle of processing, the monitoring matrix at the next time point is predicted through the first-stage processing model.

3. The data operation and maintenance management method for a smart power plant according to claim 1 is characterized in that: When the real-time monitoring matrix is ​​input into the pre-trained first-stage processing model, the first-stage processing model outputs a loop processing of the first prediction monitoring matrix for predicting the target time point. Each loop processing corresponds to a monitoring matrix of a time slice, and each loop processing takes the monitoring matrix output by the previous loop processing as input, and takes the monitoring matrix output by the last loop processing as the first prediction monitoring matrix.

4. The data operation and maintenance management method for a smart power plant according to claim 1, characterized in that: The second stage of processing is composed of multiple sub-processing stages, each of which is completed by the first processing model and the second processing model: During the processing of the first processing model, the real-time monitoring matrix and the predicted monitoring matrix of the currently updated target time point are combined and input into the first processing model, and the first processing model outputs a transition monitoring matrix for predicting the time slice between the current time point and the target time point; The transition monitoring matrix is ​​input into the second processing model, and the second processing model outputs an updated prediction monitoring matrix at the target time point.

5. The data operation and maintenance management method for a smart power plant according to claim 3 is characterized in that: The number of sub-processing stages in the second stage processing corresponds to the number of loops in the first processing stage. The sub-processing stages correspond to the loop processing order of the first processing stage based on the order in the second stage processing. The first processing model outputs a transition monitoring matrix of a time slice corresponding to the loop processing order of the first processing stage, and the second processing model outputs an updated prediction monitoring matrix of the target time point based on the transition monitoring matrix of the time slice.

6. The data operation and maintenance management method for a smart power plant according to claim 1, characterized in that: In the steps of constructing a real-time monitoring vector based on the monitoring data transmitted from each monitoring point and constructing a monitoring matrix based on the real-time monitoring vectors of all monitoring points, the monitoring data transmitted from each monitoring point is encoded to obtain a real-time monitoring vector of the corresponding monitoring point, and the real-time monitoring vectors are arranged in sequence based on the numbers of the respective monitoring points to obtain the monitoring matrix.

7. The data operation and maintenance management method for a smart power plant according to any one of claims 1 to 6, characterized in that: The method further includes health determination, and the health determination step includes: Inputting the second prediction monitoring matrix into a pre-trained health determination model, outputting a health value based on the health determination model, and determining the overall health of the generator set based on the health value; The overall health is compared with a health threshold to determine whether the generator set is overall healthy.

8. The data operation and maintenance management method for a smart power plant according to claim 7, characterized in that: The step of determining the health level also includes: If the generator set is in a non-healthy state as a whole, the second prediction monitoring matrix is ​​split corresponding to each monitoring point to obtain a prediction monitoring vector corresponding to each monitoring point; Based on the predicted monitoring vector, it is determined whether the corresponding monitoring point is in a waiting-for-inspection state.

9. The data operation and maintenance management method for a smart power plant according to claim 8, characterized in that: In the step of splitting the second prediction monitoring matrix corresponding to each monitoring point to obtain a prediction monitoring vector corresponding to each monitoring point; and determining whether the corresponding monitoring point is in a waiting inspection state based on the prediction monitoring vector: Obtain the power generation data and operating frequency data of the generator set at multiple historical time points; Based on the power generation data and operating frequency data of multiple historical time points, a power prediction vector and a frequency prediction vector are respectively constructed, and based on the power prediction vector and the frequency prediction vector, a power prediction value and a frequency prediction value at a target time point are predicted; The power prediction value and the frequency prediction value are constructed as a supplementary data group, and are spliced ​​with the prediction monitoring vector split out of the second prediction monitoring matrix. The spliced ​​vector is input into the judgment model corresponding to each monitoring point, and the judgment model is used to determine whether the monitoring point is in a state to be inspected.

10. A data operation and maintenance management system for a smart power plant, characterized by: The system includes a computer device, which includes a processor and a memory, wherein the memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the system implements the steps implemented by the method described in any one of claims 1 to 9.

Images