An intelligent prediction method for working conditions of a hydraulic turbine by fusing multi-source monitoring information

Abstract

The present disclosure relates to a water turbine working condition intelligent prediction method fusing multi-source monitoring information, comprising: obtaining a plurality of target fusion information of a to-be-tested unit group under a current working condition; symbolizing the plurality of target fusion information to obtain an initial feature sequence; inputting the initial feature sequence into a trained prediction model to obtain a first target feature sequence; and determining a first target predicted working condition corresponding to the first target feature sequence. According to the present disclosure, the multi-source monitoring information obtained by a plurality of sensors of the to-be-tested unit group under the current working condition is fused into a group of target fusion information, so that redundant information in the multi-source monitoring information can be eliminated; the plurality of target fusion information is symbolized to extract features in the target fusion information, the extracted features can increase the noise resistance of the information, and the extracted features are used to input the trained prediction model, so that the first target predicted working condition obtained by prediction is more reliable and the accuracy of the prediction result is enhanced.

Description

Technical Field

[0001] This disclosure relates to the field of motor control technology, specifically to a method, device, medium, and equipment for intelligent prediction of turbine operating conditions by integrating multi-source monitoring information. Background Technology

[0002] As the core mechanical equipment of a hydropower station, the safe operation of the turbine is of great significance to the regional power grid, hydraulic structures, and natural environment. Any abnormality or malfunction in the turbine's operation could lead to huge economic losses and major safety accidents for the power station, with unimaginable consequences. Therefore, every effort must be made to avoid the occurrence of any malfunction.

[0003] Commonly used monitoring signals for hydroelectric turbines include vibration signals, swing signals, pressure pulsation signals, and sound signals. In actual production, it is necessary to combine multiple signals and make timely judgments on the status of the hydroelectric unit to ensure comprehensive information. However, in actual operation, the operating conditions of hydroelectric turbines are complex, with significant differences between different units. Different sensor locations will also yield completely different monitoring results, and data is affected by environmental noise, leading to a large discrepancy between predicted and actual operating conditions. Summary of the Invention

[0004] The purpose of this disclosure is to provide an intelligent prediction method for turbine operating conditions that integrates multi-source monitoring information, in order to solve the problem that the predicted operating conditions differ significantly from the actual operating conditions due to interference from environmental noise in the data acquired by the sensors.

[0005] To achieve the above objectives, a first aspect of this disclosure proposes an intelligent prediction method for turbine operating conditions that integrates multi-source monitoring information, the method comprising: Under the current operating conditions of the unit under test, multiple target fusion information of the unit under test is acquired. The target fusion information is obtained after processing multi-source monitoring information. The fused information from the multiple targets is symbolized to obtain an initial feature sequence; The initial feature sequence is input into the trained prediction model to obtain a first target feature sequence, wherein the first target feature sequence is a sample feature sequence that matches the initial feature sequence among multiple sample feature sequences of the prediction model; Determine the first target prediction condition corresponding to the first target feature sequence.

[0006] Optionally, the training method for the prediction model includes: A labeled training sample set is obtained, wherein the training sample set includes sample feature sequences corresponding to multiple sample fusion information of the unit under test under different operating conditions, and the sample fusion information is obtained by processing the multi-source monitoring information of the samples obtained under any operating condition of the unit under test. Based on the labeled training sample set, the preset neural network is iteratively trained to obtain the trained prediction model.

[0007] Optionally, before the step of acquiring multiple target fusion information of the unit under test under its current operating condition, the method further includes: Within a first preset time period when the unit under test is in its current operating condition, the multi-source monitoring information of the unit under test is acquired once every second preset time period. The information in each group of multi-source monitoring information is time-aligned to obtain the corresponding synchronous multi-source information. The synchronous multi-source information is fused to obtain the target fused information.

[0008] Optionally, the step of fusing the synchronized multi-source information to obtain the target fused information includes: Based on the number of information items corresponding to the synchronized multi-source information and the length of each information item, sample information is obtained; The sample information is standardized to obtain standard information; Sensitive information is extracted from the standard information to obtain target fusion information.

[0009] Optionally, after symbolizing the fused information of the multiple targets to obtain an initial feature sequence, the method further includes: Within the second preset time period following the first preset time period, update the fusion information; The updated fused information is symbolized and the initial feature sequence is updated to obtain the updated feature sequence; The updated feature sequence is input into the trained prediction model to obtain the second target feature sequence; Determine the second target prediction condition corresponding to the second target feature sequence.

[0010] Optionally, the step of symbolizing the fused information of the multiple targets to obtain an initial feature sequence includes: The target fusion information is converted into a binary symbol sequence; Convert the binary symbol sequence into a decimal symbol sequence; Extract the symbol entropy features of the decimal symbol sequence to obtain the initial feature sequence.

[0011] Optionally, the step of extracting the symbol entropy features of the decimal symbol sequence to obtain the initial feature sequence includes: The symbol entropy feature of the decimal symbol sequence is calculated using the following formula:

[0012] Where SE is the symbol entropy feature, M is the number of symbol types in the decimal symbol sequence, and P... i (n) represents the probability of the i-th symbol appearing in the decimal symbol sequence.

[0013] To achieve the above objectives, in a second aspect of this disclosure, a smart predictive device for turbine operating conditions that integrates multi-source monitoring information is provided, the device comprising: The first acquisition module is used to acquire multiple target fusion information of the unit under test within a first preset time period under the current operating condition. The target fusion information is obtained after processing based on multi-source monitoring information. The first acquisition module is used to symbolize the fused information of the multiple targets to obtain an initial feature sequence; The second obtaining module is used to input the initial feature sequence into the trained prediction model to obtain a first target feature sequence, wherein the first target feature sequence is a sample feature sequence that matches the initial feature sequence among multiple sample feature sequences of the prediction model; The first determining module is used to determine the first target prediction condition corresponding to the first target feature sequence.

[0014] To achieve the above objectives, in a third aspect of this disclosure, a non-transitory computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the method described in the first aspect of this disclosure.

[0015] To achieve the above objectives, in a fourth aspect of this disclosure, an electronic device is provided, comprising: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method described in the first aspect of this disclosure.

[0016] This embodiment of the disclosure symbolizes the target fusion information corresponding to each set of multi-source monitoring information of the unit under test to obtain initial sequence features, and obtains a first target feature sequence matching the initial sequence features through a trained prediction model, thereby determining the first target predicted operating condition corresponding to the first target feature sequence. By fusing the multi-source monitoring information obtained by multiple sensors of the unit under test under the current operating condition into a set of target fusion information, redundant information in the multi-source monitoring information can be eliminated; symbolizing multiple target fusion information to extract features from the target fusion information can increase the noise resistance of the information. When the extracted features are used as input to the trained prediction model, the predicted first target predicted operating condition is more reliable, enhancing the accuracy of the prediction results.

[0017] Other features and advantages of this disclosure will be described in detail in the following detailed description section. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0019] Figure 1 This is a schematic diagram of the production equipment structure of the hardware operating environment involved in the embodiments of this application.

[0020] Figure 2 This is a flowchart illustrating an intelligent prediction method for turbine operating conditions that integrates multi-source monitoring information, according to an exemplary embodiment.

[0021] Figure 3 This is a diagram illustrating the matching process between the initial feature sequence and the sample feature sequence in an intelligent prediction method for turbine operating conditions that integrates multi-source monitoring information, according to an exemplary embodiment.

[0022] Figure 4 This is a block diagram illustrating an intelligent prediction device for turbine operating conditions that integrates multi-source monitoring information, according to an exemplary embodiment. Detailed Implementation

[0023] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0024] In related technologies, if a hydro turbine malfunctions or experiences abnormal operation, it could lead to huge economic losses and major safety accidents for the power station, with unimaginable consequences. Therefore, it is imperative to avoid all kinds of malfunctions as much as possible. Thus, research and application of hydro turbine condition monitoring and fault diagnosis technologies are of great significance for ensuring the safe and stable operation of the unit. Furthermore, with the rapid development of data mining and artificial intelligence, there is a need for the diagnosis and continuous monitoring of the unit's equipment status.

[0025] In this context, this disclosure proposes an intelligent prediction method for turbine operating conditions that integrates multi-source monitoring information. The specific concept is as follows: the initial feature sequence obtained by processing the target fusion information corresponding to the multi-source monitoring information of the unit under test under different operating conditions is input into the trained prediction model. The first target feature sequence is obtained by matching the sample feature sequence in the prediction model. The first target prediction operating condition is determined based on the first target feature sequence to predict the operating status of the turbine in a short period of time and to judge the possible operating conditions and anomalies of the unit under test.

[0026] Reference Figure 1 , Figure 1 This is a schematic diagram of the production equipment structure of the hardware operating environment involved in the embodiments of this application.

[0027] like Figure 1 As shown, the production equipment may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0028] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the production equipment and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0029] like Figure 1 As shown, the memory 1005, which serves as a storage medium, may include an operating system, a data storage module, a network communication module, a user interface module, and electronic programs.

[0030] exist Figure 1 In the production equipment shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the production equipment of the present invention can be set in the production equipment. The production equipment calls the intelligent prediction device for turbine operating conditions that integrates multi-source monitoring information stored in the memory 1005 through the processor 1001, and executes the intelligent prediction method for turbine operating conditions that integrates multi-source monitoring information provided in the embodiments of this application.

[0031] The following detailed description of the intelligent prediction method for turbine operating conditions that integrates multi-source monitoring information, provided in this application, is illustrated through specific embodiments.

[0032] Please see Figure 2 , Figure 2 This is a flowchart illustrating an intelligent prediction method for turbine operating conditions that integrates multi-source monitoring information, according to an exemplary embodiment. Figure 2 As shown, the method includes: S101. Under the current operating conditions of the unit under test, acquire multiple target fusion information of the unit under test. The target fusion information is obtained after processing the multi-source monitoring information.

[0033] In the specific implementation process, the unit under test refers to the turbine unit whose operating conditions need to be predicted. The current operating condition is any operating condition of the unit under test. The flow rate, head, and speed of the turbine are different under different operating conditions. Among them, the operating conditions of the unit under test can include steady-state operating conditions, transient operating conditions, and fault operating conditions. Steady-state operating conditions refer to the unit under test being in a stable operating state. Transient operating conditions can include variable operating conditions and start-up and shutdown operating conditions. Variable operating conditions refer to the operating state of the unit under test when the load deviates from the rated load. Start-up and shutdown operating conditions refer to the operating state of the turbine unit in the process of starting up and about to stop. Fault operating conditions refer to the operating state of the unit under test when a fault occurs.

[0034] Multi-source monitoring information refers to the information generated during the operation of the unit under test (TUD) acquired through various sensors. This generally includes vibration information, swing information, pressure pulsation information, and acoustic information. Specifically, vibration information includes vibration signals from the upper frame, lower frame, and generator stator of the TTD; swing information includes swing signals from the upper guide bearing, lower guide bearing, and water guide bearing of the TTD; pressure pulsation information includes pressure pulsation signals from the tailrace inlet / outlet and volute inlet / outlet of the TTD, with at least four measuring points; and acoustic information includes acoustic monitoring signals from the turbine chamber and generator wind tunnel of the TTD, with at least two measuring points.

[0035] After fusing multi-source monitoring information, target fusion information can be obtained. Generally, existing information fusion technologies can be used to fuse multi-source monitoring information. When fusing multi-source monitoring information, it is necessary to preprocess the collected information and perform time alignment on multiple pieces of information to ensure data synchronization. The target fusion information acquired by the unit under test under the current operating conditions is continuous in time; that is, multiple sets of multi-source monitoring information are acquired continuously within a certain period, and each set of multi-source monitoring information is fused to obtain the corresponding target fusion information.

[0036] S102. Symbolize the fused information of multiple targets to obtain an initial feature sequence.

[0037] In the specific implementation process, the fused information from multiple targets is symbolized, that is, each target fused information is transformed, thereby extracting the symbolic entropy features from each target fused information to transform the signal into features. The multiple symbolic entropy features are arranged sequentially according to the acquisition time of the corresponding target fused information to obtain the initial feature sequence.

[0038] S103. Input the initial feature sequence into the trained prediction model to obtain the first target feature sequence, wherein the first target feature sequence is the sample feature sequence that matches the initial feature sequence among multiple sample feature sequences of the prediction model.

[0039] In the specific implementation process, the prediction model is used to match the input initial feature sequence with multiple sample feature sequences in the prediction model to obtain the first target feature sequence. The first target feature sequence refers to the sample feature sequence in the prediction model that has the highest matching degree with the initial feature sequence.

[0040] In the prediction model, each sub-condition under different operating conditions of the unit under test can correspond to a sample feature sequence. If there is a sequence segment in the sample feature sequence that matches the initial feature sequence, then this sample feature sequence is the first target feature sequence. Alternatively, each sub-condition under different operating conditions of each unit under test can correspond to multiple sample feature sequences, and the length of each sample feature sequence is longer than the length of the initial feature sequence. In this case, multiple sample feature sequences need to be matched, and the sample feature sequence with the highest matching degree to the initial feature sequence is the first target feature sequence.

[0041] As mentioned above, the signal length of the initial feature sequence is less than the length of the sample feature sequence. Generally, the signal length of the initial feature sequence is 1-10 times shorter than the length of the sample feature sequence. For example, if the signal length of the initial feature sequence is 7 times shorter than the length of the sample feature sequence, and the signal length of the initial feature sequence is m, then the signal length of the sample feature sequence is m+7. During the matching process, the entire sequence of the initial feature sequence is compared sequentially with the sequence segments of length m in the sample feature sequence, thereby selecting the sample feature sequence with the highest matching degree with the initial feature sequence as the first target feature sequence.

[0042] Specifically, when each sub-operating condition of the unit under test can correspond to a sample feature sequence, after inputting the initial feature sequence into the trained prediction model, it is necessary to search for a sample feature sequence that matches the initial feature sequence based on the initial sequence features, and to perform matching calculations on the data in the sample feature sequence. Therefore, during the matching process, the R2 coefficients of different sequence segments on the initial feature sequence and the matched sample feature sequence are calculated. After the initial feature sequence has been matched with all sample feature sequences, all obtained R2 coefficients are compared, and the sample feature sequence operating condition corresponding to the largest R2 value is the first target feature sequence.

[0043] Reference Figure 3 The diagram illustrates an exemplary embodiment of a method for intelligent prediction of turbine operating conditions by fusing multi-source monitoring information, showing the matching process between an initial feature sequence and sample feature sequences. The matching process involves inputting the initial feature sequence into a trained prediction model to obtain k sample feature sequences that match the initial feature sequence. The step size during matching is 1, and the R² coefficient is calculated for each match. After matching with all samples, all obtained R² coefficients are compared, and the sample feature sequence corresponding to the largest R² value is the first target feature sequence for this prediction. Based on the first target feature sequence, the corresponding first target predicted operating condition can be obtained. Figure 3The process involves matching one of the sample feature sequences with the initial sample features. The sample feature sequence is matched with the initial sample features point by point. The prediction result is k=5. Assuming the predicted value is 12423, when matching samples in the graph, the R2 coefficients of 12423 and 12423 are calculated for the first time. Then, the step is increased by 1. The R2 coefficients of 12423 and 24237 are calculated for the second time. And so on. The sample feature sequence corresponding to the sequence segment with the largest R2 coefficient is the first target feature sequence.

[0044] S104. Determine the first target prediction condition corresponding to the first target feature sequence.

[0045] In the specific implementation process, after determining the first target feature sequence, the prediction result can be obtained based on the last data of the sequence segment in the first target feature sequence that has the highest matching degree with the initial feature sequence.

[0046] This embodiment of the disclosure symbolizes the target fusion information corresponding to each set of multi-source monitoring information of the unit under test to obtain initial sequence features, and obtains a first target feature sequence matching the initial sequence features through a trained prediction model, thereby determining the first target predicted operating condition corresponding to the first target feature sequence. By fusing the multi-source monitoring information obtained by multiple sensors of the unit under test under the current operating condition into a set of target fusion information, redundant information in the multi-source monitoring information can be eliminated; symbolizing multiple target fusion information to extract features from the target fusion information can increase the noise resistance of the information. When the extracted features are used as input to the trained prediction model, the predicted first target predicted operating condition is more reliable, enhancing the accuracy of the prediction results.

[0047] In some embodiments, the training method for the prediction model includes: Obtain a labeled training sample set, which includes sample feature sequences corresponding to multiple sample fusion information of the unit under test under different operating conditions. The sample fusion information is obtained by processing the multi-source monitoring information of the samples obtained under any operating condition of the unit under test. Based on a labeled training sample set, the pre-defined neural network is iteratively trained to obtain a trained prediction model.

[0048] In the specific implementation process, the structure of the preset neural network includes an input layer, a hidden layer, and an output layer. The input layer has m neurons, which corresponds to the length of the training samples input into the preset neural network (corresponding to the length of the initial feature sequence of the prediction model after input training). The hidden layer contains 20 neurons, and the activation function of the hidden layer is the ReLU function. There is one output neuron, and the output layer uses a linear activation function. The optimizer is Adam.

[0049] Furthermore, in the labeled training sample set, each sample feature sequence contains a label. Since the actual length of the sample feature sequence is m+1, and the information length of the initial feature sequence input to the prediction model is m, the last element of the sample feature sequence is not included in the training when training the preset neural network. That is, the length of the sample feature sequence input to the preset neural network is m. Additionally, the label value is the same as the last element of the sample feature sequence. After inputting the sample feature sequence into the preset neural network, the output result is compared with the label value for adjustment. In other words, during training, for each sample feature sequence, the part used for training is input from the input layer. The difference between the training label and the actual output of the model is the prediction error of the model for each iteration. Through continuous iterative training, the model's output will gradually approach the true value of the prediction, completing the model training.

[0050] The training samples for the prediction model are sample feature sequences corresponding to target fusion information under different working conditions. Each working condition corresponds to multiple sub-working conditions, and the samples used for each sub-working condition are divided into the following categories: ① Steady-state operating conditions, including sub-conditions under different loads and no-load sub-conditions; ② Transient operating conditions, including sub-operating conditions during load increase and load decrease; and sub-operating conditions during start-up and shutdown. ③ Fault conditions: Low-frequency vortex bands in the tailrace pipe, abnormal shafting, and severe cavitation erosion. In some embodiments, before the step of acquiring multiple target fusion information of the unit under test (TUD) under its current operating condition, the method further includes: Within the first preset time period when the unit under test is in the current operating condition, the multi-source monitoring information of the unit under test is acquired once every second preset time period. The information in each group of multi-source monitoring information is time-aligned to obtain the corresponding synchronous multi-source information. The synchronous multi-source information is fused to obtain the target fused information.

[0051] In the specific implementation process, the first preset duration is longer than the second preset duration. This means the ratio of the first preset duration to the second preset duration represents the number of sets of multi-source monitoring information acquired. The information in the multi-source monitoring information is processed to ensure time alignment of each piece of information; that is, each set of multi-source monitoring information has the same number of data points within a certain period. The acquisition time of each piece of information in the multi-source monitoring information is aligned to ensure the time synchronization of each data point within the multi-source monitoring information, so that all data within a set of multi-source monitoring information have the same sampling frequency. Furthermore, under the premise of time synchronization, interpolation operations can be performed on low-sampling-frequency data.

[0052] In some embodiments, the step of fusing synchronous multi-source information to obtain target fused information includes: Sample information is obtained based on the number of information items corresponding to the synchronized multi-source information and the length of each information item; The sample information is standardized to obtain standard information; Sensitive information is extracted from the standard information to obtain target fusion information.

[0053] In the specific implementation process, data fusion is based on the principal components of synchronous multi-source information. This mainly involves retaining sensitive information from multiple sets of original information based on eigenvalues, thereby achieving dataset compression. The process of fusing synchronous multi-source information takes multi-source monitoring information from hydropower units as an example. Assuming the number of signals in each set of multi-source monitoring information is n, and the length of each signal is m, the multi-source monitoring information will constitute an m×n matrix X(m×n). After standardizing the matrix X(m×n), the matrix is ​​obtained. Here, the Z-Score normalization method is used to obtain... The formula is as follows:

[0054] Among them, t i Principal vector, p i This is the load vector.

[0055] Each set of multi-source monitoring signals is fused into a single new signal, i.e., the number of principal components is 1, to obtain the target fusion information. The formula is as follows:

[0056] T1 is the fused principal vector, and p1 is the fused load vector.

[0057] In some embodiments, after symbolizing the fused information of multiple targets to obtain an initial feature sequence, the method further includes: Within a second preset time period following the first preset time period, acquire updated fusion information; The updated fused information is symbolized and the initial feature sequence is updated to obtain the updated feature sequence; The updated feature sequence is input into the trained prediction model to obtain the second target feature sequence; Determine the second target prediction condition corresponding to the second target feature sequence.

[0058] In the specific implementation process, after determining the first target prediction condition, it is necessary to predict the next stage of the first target prediction condition. At this time, within the second preset time after the first preset time, the target fusion information corresponding to the multi-source monitoring information is re-acquired as the updated fusion information. Symbolizing the updated fusion information can obtain a feature data. The first digit of the initial feature sequence corresponding to the first preset time is deleted, and the feature data is added to the last digit of the initial feature sequence to obtain the updated feature sequence. The updated feature sequence is then input into the trained prediction model to obtain the prediction result of the next stage.

[0059] The embodiments disclosed herein are based on the prediction results of the next stage of the first target prediction operating condition of the prediction unit. In this process, the second target prediction operating condition can be obtained by updating according to the initial feature sequence used when predicting the first target prediction operating condition, without having to re-acquire multiple target fusion information for calculation, which can reduce the amount of calculation and increase the efficiency of predicting the second target prediction operating condition.

[0060] In some embodiments, the step of symbolizing the fused information of multiple targets to obtain an initial feature sequence includes: Convert the target fusion information into a binary symbol sequence; Convert a binary symbol sequence into a decimal symbol sequence; Extract the symbol entropy features of the decimal symbol sequence to obtain the initial feature sequence.

[0061] In the specific implementation process, when converting the target fusion information into a binary symbol sequence, the mean of all data in the target fusion information is calculated. Values ​​greater than the mean are recorded as 1, and those less than the mean are recorded as 0. At this time, the target fusion information is converted into a symbol sequence composed of 0 and 1.

[0062] The binary symbol sequence is converted into a decimal symbol sequence by using a coding window of length 3 as the basic unit. The binary symbol sequence is encoded and calculated, and each step is advanced by 1. Each window contains only 3 consecutive data points from the binary symbol sequence. The binary numbers in the coding window are then converted into decimal numbers, thus obtaining a decimal symbol sequence consisting of integers from 0 to 9.

[0063] In some embodiments, the step of extracting the symbol entropy features of the decimal symbol sequence to obtain an initial feature sequence includes: The symbol entropy characteristics of a decimal symbol sequence can be calculated using the following formula:

[0064] Where SE is the symbol entropy feature, M is the number of symbol types in the decimal symbol sequence (M is usually 8), and Pi(n) is the probability of the i-th symbol appearing in the decimal symbol sequence.

[0065] Please see Figure 4 , Figure 4 This is a block diagram illustrating an intelligent prediction device for turbine operating conditions that integrates multi-source monitoring information, according to an exemplary embodiment. Figure 4 As shown, the device includes: The first acquisition module is used to acquire multiple target fusion information of the unit under test under the current operating condition. The target fusion information is obtained by processing multi-source monitoring information. The first acquisition module is used to symbolize the fused information of multiple targets to obtain an initial feature sequence; The second acquisition module is used to input the initial feature sequence into the trained prediction model to obtain the first target feature sequence, wherein the first target feature sequence is the sample feature sequence that matches the initial feature sequence among multiple sample feature sequences of the prediction model; The first determining module is used to determine the first target prediction condition corresponding to the first target feature sequence.

[0066] In some embodiments, the device further includes a training module, the training module comprising: The acquisition submodule is used to acquire a labeled training sample set, which includes sample feature sequences corresponding to multiple sample fusion information of the unit under test under different operating conditions. The sample fusion information is obtained by processing the multi-source monitoring information of the samples acquired under any operating condition of the unit under test. The training submodule is used to iteratively train a pre-defined neural network based on a labeled training sample set to obtain a trained prediction model.

[0067] In some embodiments, the device further includes: The second acquisition module is used to acquire multi-source monitoring information of the unit under test once every second preset time period within a first preset time period when the unit under test is in the current operating condition. The third acquisition module is used to perform time alignment processing on the information in each group of multi-source monitoring information to obtain the corresponding synchronous multi-source information. The fourth acquisition module is used to fuse synchronous multi-source information to obtain target fused information.

[0068] In some embodiments, the fourth obtaining module includes: The first submodule is used to obtain sample information based on the number of information corresponding to the synchronized multi-source information and the length of each information. The second submodule is used to standardize the sample information to obtain standard information; The third submodule is used to extract sensitive information from the standard information to obtain target fusion information.

[0069] In some embodiments, the device further includes: The third acquisition module is used to acquire updated fusion information within a second preset time period after the first preset time period; The fifth module is used to symbolize the updated fusion information and update the initial feature sequence to obtain the updated feature sequence. The sixth module is used to input the updated feature sequence into the trained prediction model to obtain the second target feature sequence; The second determining module is used to determine the second target prediction condition corresponding to the second target feature sequence.

[0070] In some embodiments, the first obtaining module includes: The first conversion submodule is used to convert the target fusion information into a binary symbol sequence; The second conversion submodule is used to convert binary symbol sequences into decimal symbol sequences; The calculation submodule is used to extract the symbol entropy features of the decimal symbol sequence to obtain the initial feature sequence.

[0071] In some embodiments, the computation submodule is specifically used for: The symbol entropy characteristics of a decimal symbol sequence can be calculated using the following formula:

[0072] Where SE is the symbol entropy feature, M is the number of symbol types in the decimal symbol sequence, and Pi(n) is the probability of the i-th symbol appearing in the decimal symbol sequence.

[0073] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0074] In another exemplary embodiment, a non-transitory computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the intelligent prediction method for turbine operating conditions that integrates multi-source monitoring information in the embodiments of this disclosure.

[0075] In another exemplary embodiment, an electronic device is also provided, comprising: A memory on which computer programs are stored; A processor is used to execute a computer program in memory to implement the steps of the intelligent prediction method for turbine operating conditions that integrates multi-source monitoring information in the embodiments of this disclosure.

[0076] In another exemplary embodiment, a computer program product is also provided, which includes a computer program executable by a programmable device, the computer program having a code portion for performing the above-described intelligent prediction method for turbine operating conditions that integrates multi-source monitoring information when executed by the programmable device.

[0077] The preferred embodiments of this disclosure have been described in detail above with reference to the accompanying drawings. However, this disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this disclosure, various simple modifications can be made to the technical solutions of this disclosure, and these simple modifications all fall within the protection scope of this disclosure.

[0078] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, this disclosure will not describe the various possible combinations separately.

[0079] Furthermore, various different embodiments of this disclosure can be combined in any way, as long as they do not violate the spirit of this disclosure, they should also be regarded as the content disclosed in this disclosure.

Claims

1. A method for intelligent prediction of turbine operating conditions by integrating multi-source monitoring information, characterized in that, The method includes: Under the current operating conditions of the unit under test, multiple target fusion information of the unit under test is acquired. The target fusion information is obtained after processing multi-source monitoring information. The fused information from the multiple targets is symbolized to obtain an initial feature sequence; The initial feature sequence is input into the trained prediction model to obtain a first target feature sequence, wherein the first target feature sequence is a sample feature sequence that matches the initial feature sequence among multiple sample feature sequences of the prediction model; Determine the first target prediction condition corresponding to the first target feature sequence; Before the step of acquiring multiple target fusion information of the unit under test under its current operating condition, the method further includes: Within a first preset time period when the unit under test is in its current operating condition, the multi-source monitoring information of the unit under test is acquired once every second preset time period. The information in each group of multi-source monitoring information is time-aligned to obtain the corresponding synchronous multi-source information. The synchronous multi-source information is fused to obtain the target fused information; After symbolizing the fused information of the multiple targets to obtain an initial feature sequence, the method further includes: Within the second preset time period following the first preset time period, update the fusion information; The updated fused information is symbolized and the initial feature sequence is updated to obtain the updated feature sequence; The updated feature sequence is input into the trained prediction model to obtain the second target feature sequence; Determine the second target prediction condition corresponding to the second target feature sequence.

2. The method according to claim 1, characterized in that, The training methods for the prediction model include: A labeled training sample set is obtained, wherein the training sample set includes sample feature sequences corresponding to multiple sample fusion information of the unit under test under different operating conditions, and the sample fusion information is obtained by processing the multi-source monitoring information of the samples obtained under any operating condition of the unit under test. Based on the labeled training sample set, the preset neural network is iteratively trained to obtain the trained prediction model.

3. The method according to claim 1, characterized in that, The step of fusing the synchronous multi-source information to obtain the target fused information includes: Based on the number of information items corresponding to the synchronized multi-source information and the length of each information item, sample information is obtained; The sample information is standardized to obtain standard information; Sensitive information is extracted from the standard information to obtain target fusion information.

4. The method according to claim 1, characterized in that, The step of symbolizing the fused information of the multiple targets to obtain an initial feature sequence includes: The target fusion information is converted into a binary symbol sequence; Convert the binary symbol sequence into a decimal symbol sequence; Extract the symbol entropy features of the decimal symbol sequence to obtain the initial feature sequence.

5. The method according to claim 4, characterized in that, The step of extracting the symbol entropy features of the decimal symbol sequence to obtain the initial feature sequence includes: The symbol entropy feature of the decimal symbol sequence is calculated using the following formula: Where SE is the symbol entropy feature, M is the number of symbol types in the decimal symbol sequence, and P... i (n) represents the probability of the i-th symbol appearing in the decimal symbol sequence.

6. A smart predictive device for turbine operating conditions that integrates multi-source monitoring information, characterized in that, The device includes: The first acquisition module is used to acquire multiple target fusion information of the unit under test within a first preset time period under the current operating condition. The target fusion information is obtained after processing based on multi-source monitoring information. The first acquisition module is used to symbolize the fused information of the multiple targets to obtain an initial feature sequence; The second obtaining module is used to input the initial feature sequence into the trained prediction model to obtain a first target feature sequence, wherein the first target feature sequence is a sample feature sequence that matches the initial feature sequence among multiple sample feature sequences of the prediction model; The first determining module is used to determine the first target prediction condition corresponding to the first target feature sequence; The second acquisition module is used to acquire multi-source monitoring information of the unit under test once every second preset time period within a first preset time period when the unit under test is in the current operating condition. The third acquisition module is used to perform time alignment processing on the information in each group of multi-source monitoring information to obtain the corresponding synchronous multi-source information. The fourth acquisition module is used to fuse synchronous multi-source information to obtain target fused information; The third acquisition module is used to acquire updated fusion information within a second preset time period after the first preset time period; The fifth module is used to symbolize the updated fusion information and update the initial feature sequence to obtain the updated feature sequence. The sixth module is used to input the updated feature sequence into the trained prediction model to obtain the second target feature sequence; The second determining module is used to determine the second target prediction condition corresponding to the second target feature sequence.

7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method described in any one of claims 1-5.

8. An electronic device, characterized in that, include: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method according to any one of claims 1-5.

Images