Hydraulic power plant equipment operation state management system and method based on data analysis

Abstract

The embodiment of the invention relates to the technical field of hydraulic power plant equipment management, provides a hydraulic power plant equipment operation state management system and method based on data analysis, and adopts an artificial intelligence technology based on deep learning to carry out real-time monitoring and data analysis on the operation condition of hydraulic power plant equipment. According to the method, the temperature time sequence change trend and the equipment vibration mode of the hydraulic power plant equipment in the operation process can be captured, then cross-modal gating conjoint analysis is carried out on the temperature characteristics and the vibration characteristics of the equipment, potential abnormal conditions of the equipment are excavated, and therefore abnormal early warning of the operation state of the equipment is achieved. The stability and efficiency of hydraulic power plant equipment operation can be improved, and the operation cost and the maintenance risk can be effectively reduced.

Description

Technical Field

[0001] The present disclosure relates to the technical field of hydropower plant equipment management, and in particular to a hydropower plant equipment operation status management system and method based on data analysis. Background Art

[0002] With the continuous growth of global energy demand and the increasing importance of clean energy development, hydropower, as an important part of renewable energy, has gradually become the focus of the power industry in terms of stability and efficiency. As the core facility of hydropower generation, the operating status of hydropower plant equipment is directly related to power generation efficiency, operating costs and safety. However, due to factors such as long-term operation, complex and changeable environment, and equipment aging, hydropower plant equipment often has various potential faults and abnormal conditions during operation. If these problems are not discovered and handled in time, they will seriously affect the operating efficiency and safety of the hydropower plant.

[0003] Traditional management of the operating status of hydropower plant equipment mainly relies on manual inspections and regular maintenance. This method is not only time-consuming and labor-intensive, but also difficult to achieve real-time monitoring and accurate analysis of the equipment operating status, which can easily lead to untimely detection of equipment failures, thereby increasing the risk of equipment failure and maintenance costs. Summary of the invention

[0004] The present disclosure aims to solve at least one of the problems existing in the prior art and provide a hydropower plant equipment operation status management system and method based on data analysis.

[0005] In one aspect of the present disclosure, a hydropower plant equipment operation status management system based on data analysis is provided, the management system comprising:

[0006] A hydropower plant equipment operating condition monitoring module, used to obtain the operating condition data of the monitored hydropower plant equipment, wherein the operating condition data includes a time queue of real-time temperature values ​​and a vibration signal;

[0007] A temperature local time series feature extraction module is used to extract temperature features based on a local time scale from the time queue of the real-time temperature value to obtain a sequence of implicit feature vectors associated with the real-time temperature local time series;

[0008] A forward propagation aggregation module, used for performing feature forward propagation aggregation on the sequence of the real-time temperature local time series associated implicit feature vectors to obtain a real-time temperature time series associated implicit propagation aggregation representation vector;

[0009] A vibration signal waveform feature extraction module, used to extract the waveform semantic features of the vibration signal to obtain a vibration signal waveform semantic feature map;

[0010] The equipment operating condition abnormality analysis module is used to determine whether to generate an equipment operating condition abnormality warning prompt based on the cross-modal gated joint features of the real-time temperature time series associated implicit propagation aggregation representation vector and the vibration signal waveform semantic feature map.

[0011] Optionally, the temperature local time series feature extraction module is specifically used to:

[0012] The time queue of the real-time temperature value is divided into a predetermined time scale and then input into a sequence encoder based on the RNN model to obtain a sequence of implicit feature vectors associated with the local temporal sequence of the real-time temperature.

[0013] Optionally, the forward propagation aggregation module is specifically used to:

[0014] The sequence of the real-time temperature local temporal association implicit feature vectors is input into a feature forward propagation network based on significant temporal attenuation of node energy to obtain the real-time temperature temporal association implicit propagation aggregate representation vector.

[0015] Optionally, the forward propagation aggregation module includes:

[0016] An energy significant descriptor calculation unit is used to calculate the energy significant descriptor of each real-time temperature local time series associated implicit feature vector based on the mean and variance of each real-time temperature local time series associated implicit feature vector in the sequence of the real-time temperature local time series associated implicit feature vector, so as to obtain a time series of energy significant descriptors;

[0017] A timestamp marking unit, used for extracting the timestamp of each real-time temperature local time series associated implicit feature vector in the sequence of the real-time temperature local time series associated implicit feature vector, so as to obtain a time series of timestamps;

[0018] A historical node feature division unit, used to use the real-time temperature local time series associated implicit feature vector corresponding to the current time point in the sequence of the real-time temperature local time series associated implicit feature vector as the current real-time temperature local time series associated implicit feature vector, and use the real-time temperature local time series associated implicit feature vectors corresponding to other time points as the historical real-time temperature local time series associated implicit feature vectors, so as to obtain a sequence of the current real-time temperature local time series associated implicit feature vector and the historical real-time temperature local time series associated implicit feature vector;

[0019] An attenuation factor calculation unit, used to calculate the energy time attenuation factor of each historical real-time temperature local time series associated implicit feature vector in the sequence of the historical real-time temperature local time series associated implicit feature vector relative to the current real-time temperature local time series associated implicit feature vector, so as to obtain a time series of energy time attenuation factors;

[0020] A historical node propagation aggregation unit is used to perform time series propagation aggregation on the sequence of the local time series associated implicit feature vectors of the historical real-time temperature using the time series of the energy significant descriptor as a positive adjustment factor and the time series of the energy time decay factor as a negative adjustment factor, so as to obtain a historical node real-time temperature time series aggregation feature vector;

[0021] The global time series aggregation unit is used to fuse the historical node real-time temperature time series aggregation feature vector and the current real-time temperature local time series associated implicit feature vector to obtain the real-time temperature time series associated implicit propagation aggregation representation vector.

[0022] Optionally, the energy significant descriptor calculation unit is specifically used to:

[0023] The expected value of the fourth power of the difference between each eigenvalue in the real-time temperature local time series associated implicit feature vector and its eigenmean is calculated, and the expected value is divided by the square of the eigenvariance of the real-time temperature local time series associated implicit feature vector to obtain the energy significant descriptor.

[0024] Optionally, the attenuation factor calculation unit is specifically used to:

[0025] Calculating a timestamp difference between the historical real-time temperature local time series associated implicit feature vector and the current real-time temperature local time series associated implicit feature vector;

[0026] The timestamp difference is rounded down and divided by a preset time decay period and then squared, and the square result is multiplied by a preset decay rate to obtain an initial energy decay coefficient;

[0027] An exponential function value with a natural constant as the base and the initial energy decay coefficient as the exponent is calculated to obtain the energy time decay factor.

[0028] Optionally, the vibration signal waveform feature extraction module is specifically used to:

[0029] The vibration signal is input into a vibration signal waveform semantic feature extractor based on a dilated convolutional neural network model to obtain the vibration signal waveform semantic feature map.

[0030] Optionally, the equipment operating condition abnormality analysis module includes:

[0031] A gated cross-modal joint encoding unit, used for inputting the real-time temperature time series associated implicit propagation aggregation representation vector and the vibration signal waveform semantic feature map into a gated cross-modal joint encoder based on a simulated converter structure to obtain a multi-modal joint representation feature map of the equipment working condition;

[0032] An operation status management result generating unit is used to input the multimodal joint representation feature map of the equipment operating condition into the classifier-based operation status management result generator to obtain an operation status management result, wherein the operation status management result is used to indicate whether an abnormal warning prompt of the equipment operating condition is generated.

[0033] Optionally, the gated cross-modal joint encoding unit is specifically used to:

[0034] Performing point convolution coding and space convolution coding on the vibration signal waveform semantic feature map respectively to obtain a vibration signal waveform space deepening semantic feature map;

[0035] Copying the vibration signal waveform space deepening semantic feature map to obtain a backup vibration signal waveform space deepening semantic feature map;

[0036] Reshape the backup vibration signal waveform space deepening semantic feature map and the vibration signal waveform space deepening semantic feature map respectively, so as to obtain a backup vibration signal waveform space deepening semantic feature matrix and a vibration signal waveform space deepening semantic feature matrix as a simulated query feature matrix and a simulated key feature matrix respectively;

[0037] Calculating the product of the pseudo-query feature matrix and the transposed matrix of the pseudo-key feature matrix to obtain a vibration signal waveform local semantic autocorrelation library matrix;

[0038] Inputting the vibration signal waveform local semantic autocorrelation library matrix into the attention gating unit based on the GELU function to obtain the vibration signal waveform local semantic autocorrelation library weight matrix;

[0039] Taking the real-time temperature time series associated implicit propagation aggregation representation vector as the value feature vector, calculating the product between the vibration signal waveform local semantic autocorrelation library weight matrix and the value feature vector to obtain a cross-modal interaction fusion condition weight vector;

[0040] The cross-modal interactive fusion condition weight vector is used as a weight vector to weight the vibration signal waveform semantic feature map channel by channel to obtain the multi-modal joint representation feature map of the equipment working condition.

[0041] Another aspect of the present disclosure provides a method for managing the operating status of hydropower plant equipment based on data analysis, the management method comprising:

[0042] Acquire the operating condition data of the monitored hydropower plant equipment, wherein the operating condition data includes a time queue of real-time temperature values ​​and a vibration signal;

[0043] Extracting temperature features based on a local time scale from the time queue of the real-time temperature values ​​to obtain a sequence of implicit feature vectors associated with local time series of the real-time temperature;

[0044] Performing feature forward propagation aggregation on the sequence of the real-time temperature local time series associated implicit feature vectors to obtain a real-time temperature time series associated implicit propagation aggregation representation vector;

[0045] Extracting the waveform semantic features of the vibration signal to obtain a vibration signal waveform semantic feature map;

[0046] Based on the cross-modal gated joint features of the real-time temperature time series associated implicit propagation aggregation representation vector and the vibration signal waveform semantic feature map, it is determined whether to generate an abnormal equipment operating condition warning prompt.

[0047] Compared with the prior art, the present invention adopts artificial intelligence technology based on deep learning to perform real-time monitoring and data analysis on the operating conditions of hydropower plant equipment, which can capture the temperature time series change trend and equipment vibration mode of the hydropower plant equipment during operation, and then through cross-modal gating joint analysis of the temperature characteristics and vibration characteristics of the equipment, the potential abnormal conditions of the equipment are excavated, thereby realizing abnormal early warning of the equipment operating status, which can not only improve the stability and efficiency of the operation of hydropower plant equipment, but also effectively reduce the operating cost and maintenance risk. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] One or more embodiments are exemplarily described by pictures in the corresponding drawings, and these exemplifications do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0049] Figure 1 A schematic diagram of the structure of a hydropower plant equipment operation status management system based on data analysis provided in one embodiment of the present disclosure;

[0050] Figure 2 A data flow diagram of a hydropower plant equipment operation status management system based on data analysis provided by another embodiment of the present disclosure;

[0051] Figure 3 A schematic diagram of the structure of a forward propagation aggregation module provided in another embodiment of the present disclosure;

[0052] Figure 4 A schematic diagram of the structure of an equipment operating condition abnormality analysis module provided in another embodiment of the present disclosure;

[0053] Figure 5A flowchart of a method for managing the operating status of hydropower plant equipment based on data analysis is provided in another embodiment of the present disclosure. DETAILED DESCRIPTION

[0054] In recent years, with the rapid development of technologies such as big data and artificial intelligence, intelligent management of the operating status of hydropower plant equipment has become possible. In view of the problems existing in the traditional management method of the operating status of hydropower plant equipment, the present invention provides a hydropower plant equipment operating status management system and method based on data analysis to improve the stability and efficiency of the operation of hydropower plant equipment and effectively reduce the operating cost and maintenance risk. Among them, the technical concept of the present invention is: using artificial intelligence technology based on deep learning to conduct real-time monitoring and data analysis of the operating conditions of hydropower plant equipment, capture the temperature time series change trend and equipment vibration mode of the hydropower plant equipment during operation, and then through the cross-modal gating joint analysis of the temperature characteristics and vibration characteristics of the equipment, dig out the potential abnormal conditions of the equipment, so as to realize the abnormal early warning of the equipment operating status. In this way, not only can the stability and efficiency of the operation of hydropower plant equipment be improved, but also the operating cost and maintenance risk can be effectively reduced.

[0055] In order to make the purpose, technical scheme and advantages of the embodiments of the present disclosure clearer, the embodiments of the present disclosure will be described in detail below in conjunction with the accompanying drawings. However, it can be understood by those skilled in the art that in each embodiment of the present disclosure, many technical details are proposed in order to enable readers to better understand the present disclosure. However, even without these technical details and various changes and modifications based on the following embodiments, the technical scheme claimed for protection in the present disclosure can also be implemented. The division of the following embodiments is for the convenience of description and should not constitute any limitation on the specific implementation of the present disclosure. The various embodiments can be combined and referenced with each other without contradiction.

[0056] An embodiment of the present disclosure relates to a hydropower plant equipment operation status management system 100 based on data analysis, including a hydropower plant equipment operation condition monitoring module 110, a temperature local time series feature extraction module 120, a forward propagation aggregation module 130, a vibration signal waveform feature extraction module 140, and an equipment operation condition abnormality analysis module 150.

[0057] Combined Figure 2 The hydropower plant equipment operating condition monitoring module 110 is used to obtain the operating condition data of the monitored hydropower plant equipment, wherein the operating condition data includes a time queue of real-time temperature values ​​and a vibration signal.

[0058] Specifically, the operating condition data can be collected through the sensor network used to monitor the hydropower plant equipment. It should be understood that temperature and vibration are two important physical quantities in the operation of the equipment, and the two can directly or indirectly reflect the operating status of the equipment. For example, an abnormal increase in temperature may mean that there are problems such as friction, overload or cooling system failure inside the equipment; and changes in vibration patterns may be related to faults such as bearing wear, imbalance or structural looseness. Therefore, in the embodiment of the present disclosure, the real-time temperature and vibration signals of the hydropower plant equipment are used as the main monitoring indicators to identify potential abnormal conditions of the equipment.

[0059] The temperature local time series feature extraction module 120 is used to extract temperature features based on the local time scale from the time queue of the real-time temperature value, so as to obtain a sequence of implicit feature vectors associated with the real-time temperature local time series.

[0060] It should be understood that during the operation of the hydropower plant equipment, the temperature of the equipment is a process that changes continuously over time. Therefore, in order to effectively learn and capture the temporal variation trend of the temperature, the temperature local temporal feature extraction module 120 can be specifically used to: divide the time queue of the real-time temperature value into a predetermined time scale and input it into a sequence encoder based on the RNN (Recurrent Neural Network) model to obtain a sequence of implicit feature vectors associated with the real-time temperature local temporal sequence.

[0061] That is to say, this embodiment first divides the time queue of real-time temperature values ​​into a series of local time windows according to a predetermined time scale, so as to analyze the time series fluctuation of temperature in the local time domain in a more detailed manner. Subsequently, the RNN model is used as a sequence encoder to perform time series modeling on the temperature data in these local time windows. Through the recurrent connection structure of the RNN model, the temporal correlation dependency of the temperature data is captured, and the temporal dynamic change pattern of the temperature is learned to generate a sequence of implicit feature vectors of the local temporal correlation of the real-time temperature.

[0062] The forward propagation aggregation module 130 is used to perform feature forward propagation aggregation on the sequence of real-time temperature local time series associated implicit feature vectors to obtain a real-time temperature time series associated implicit propagation aggregation representation vector.

[0063] It should be understood that in order to mine the temporal variation characteristics of temperature in the global time domain, it is necessary to further integrate the information of the sequence of implicit feature vectors associated with local temporal associations of real-time temperature. In particular, in the evaluation of the operating status of the equipment, the change in temperature data often has a time decay characteristic, that is, the more recent temperature data is more important for the evaluation of the current equipment status, while the more distant temperature data may gradually lose its influence over time. Based on this, the forward propagation aggregation module 130 can be specifically used to: input the sequence of implicit feature vectors associated with local temporal associations of real-time temperature into a feature forward propagation network based on the significant temporal decay of node energy, so as to obtain a real-time temperature temporal association implicit propagation aggregation representation vector.

[0064] That is to say, this embodiment introduces a feature forward propagation network based on the significant time series attenuation of node energy to process the sequence of local time series associated implicit feature vectors of real-time temperature, and realizes the global time series feature extraction of temperature data by simulating the time attenuation effect. Specifically, the feature forward propagation network first calculates the energy significant descriptor according to the mean and variance of each local time series associated implicit feature vector of real-time temperature, quantifies its feature importance, and assigns time information to the local time series associated implicit feature vector of real-time temperature at each time node, and uses the time series attenuation mechanism to generate the corresponding energy time attenuation factor based on the time span between the local time series associated implicit feature vector of real-time temperature at each historical time node and the current (i.e., the end of the time series) local time series associated implicit feature vector of real-time temperature, so that in the process of feature forward propagation, it can aggregate information based on the feature significance of each time node and the time series attenuation characteristics of information to generate a real-time temperature time series associated implicit propagation aggregate representation vector, thereby more accurately describing the operating status of the device.

[0065] For example, Figure 3 As shown, the forward propagation aggregation module 130 includes an energy significant descriptor calculation unit 131, a timestamp marking unit 132, a historical node feature division unit 133, an attenuation factor calculation unit 134, a historical node propagation aggregation unit 135, and a global timing aggregation unit 136.

[0066] The energy significant descriptor calculation unit 131 is used to calculate the energy significant descriptor of each real-time temperature local time series associated implicit feature vector based on the mean and variance of each real-time temperature local time series associated implicit feature vector in the sequence of real-time temperature local time series associated implicit feature vectors to obtain a time series of energy significant descriptors.

[0067] Exemplarily, the energy significant descriptor calculation unit 131 is specifically used to: calculate the expected value of the fourth power of the difference between each eigenvalue in the real-time temperature local time series associated implicit feature vector and its eigenvalue mean, and divide the expected value by the square of the feature variance of the real-time temperature local time series associated implicit feature vector to obtain the energy significant descriptor.

[0068] The timestamp marking unit 132 is used to extract the timestamp of each real-time temperature local time series associated implicit feature vector in the sequence of real-time temperature local time series associated implicit feature vectors to obtain a time series of timestamps.

[0069] The historical node feature division unit 133 is used to use the real-time temperature local time series associated implicit feature vector corresponding to the current time point in the sequence of real-time temperature local time series associated implicit feature vectors as the current real-time temperature local time series associated implicit feature vector, and use the real-time temperature local time series associated implicit feature vectors corresponding to other time points as the historical real-time temperature local time series associated implicit feature vectors, so as to obtain a sequence of the current real-time temperature local time series associated implicit feature vector and the historical real-time temperature local time series associated implicit feature vector.

[0070] The attenuation factor calculation unit 134 is used to calculate the energy time attenuation factor of each historical real-time temperature local time series associated implicit feature vector in the sequence of historical real-time temperature local time series associated implicit feature vectors relative to the current real-time temperature local time series associated implicit feature vector to obtain a time series of energy time attenuation factors.

[0071] Exemplarily, the attenuation factor calculation unit 134 is specifically used to: calculate the timestamp difference between the historical real-time temperature local time series associated implicit feature vector and the current real-time temperature local time series associated implicit feature vector; round down the timestamp difference and divide it by the preset time attenuation period and then square it, and multiply the square result by the preset attenuation rate to obtain the initial energy attenuation coefficient; calculate the exponential function value with the natural constant as the base and the initial energy attenuation coefficient as the exponent to obtain the energy time attenuation factor.

[0072] The historical node propagation aggregation unit 135 is used to perform time series propagation aggregation on the sequence of historical real-time temperature local time series associated implicit feature vectors using the time series of energy significant descriptors as positive adjustment factors and the time series of energy time decay factors as negative adjustment factors to obtain the historical node real-time temperature time series aggregation feature vector.

[0073] The global time series aggregation unit 136 is used to fuse the historical node real-time temperature time series aggregation feature vector and the current real-time temperature local time series associated implicit feature vector to obtain the real-time temperature time series associated implicit propagation aggregation representation vector.

[0074] In the above embodiment, the forward propagation aggregation module 130 processes the sequence of the real-time temperature local time series associated implicit feature vectors through the various units it includes using the following time series propagation fusion formula to obtain the real-time temperature time series associated implicit propagation aggregation representation vector, wherein the time series propagation fusion formula is:

[0075] X={x1,x2,...,x k-1 , x k}

[0076]

[0077]

[0078]

[0079]

[0080] Among them, X represents the sequence of implicit feature vectors associated with the local time series of real-time temperature, the value of k is the number of feature vectors in the sequence of implicit feature vectors associated with the local time series of real-time temperature, and x1, x2, x i 、x k-1 and x k They respectively represent the first, second, i-th, k-1-th and current (i.e., k-th) real-time temperature local time series associated implicit feature vectors in the sequence of real-time temperature local time series associated implicit feature vectors, represents the jth eigenvalue of the i-th real-time temperature local time series associated implicit eigenvector, the value of L is the length of the i-th real-time temperature local time series associated implicit eigenvector, μ i is the characteristic mean of the implicit characteristic vector of the local time series association of the i-th real-time temperature, σ 2 i is the characteristic variance of the implicit characteristic vector associated with the local time series of the i-th real-time temperature, It represents the fourth-order central moment of the implicit eigenvector of the local time series association of the ith real-time temperature, that is, the expected value of the fourth power of the difference between each eigenvalue in the ith real-time temperature local time series association implicit eigenvector and its eigenmean. is the energy significant descriptor of the implicit feature vector of the local temporal association of the i-th real-time temperature, is the energy significant descriptor of the implicit feature vector of the local temporal correlation of the current real-time temperature, t k The timestamp representing the implicit feature vector of the local temporal association of the current real-time temperature, t i The timestamp representing the implicit feature vector associated with the local temporal sequence of the i-th real-time temperature, represents rounding down, β represents the preset decay rate, L represents the preset time decay period, exp(·) represents the exponential function operation, α is the weight hyperparameter, V f Represents the real-time temperature time series associated implicit propagation aggregation representation vector.

[0081] The vibration signal waveform feature extraction module 140 is used to extract the waveform semantic features of the vibration signal to obtain a vibration signal waveform semantic feature map.

[0082] It should be understood that, considering that vibration signals usually have the characteristics of non-stationarity, nonlinearity and time-varying, in order to more effectively analyze the operating vibration mode of the equipment, the vibration signal waveform feature extraction module 140 can be specifically used to: input the vibration signal into a vibration signal waveform semantic feature extractor based on a void convolutional neural network model to obtain a vibration signal waveform semantic feature map.

[0083] That is to say, this embodiment uses a vibration signal waveform semantic feature extractor based on a dilated convolutional neural network model to perform convolution processing on the vibration signal to mine the key semantic information in the vibration signal waveform, thereby obtaining a vibration signal waveform semantic feature map. By introducing a dilated convolution operation, the dilated convolutional neural network model can expand the receptive field without increasing the number of parameters, thereby being able to capture long-distance dependencies and complex patterns in the vibration signal waveform, achieve global analysis of the device vibration pattern, and provide an important basis for subsequent anomaly detection.

[0084] Exemplarily, the atrous convolutional neural network model may include a first convolution branch structure, a second convolution branch structure, a third convolution branch structure, and a fourth convolution branch structure in parallel, and a multi-scale fusion structure connected to the first to fourth convolution branch structures, wherein the first convolution branch uses a first convolution kernel having a first size, the second convolution branch uses a second convolution kernel having a first size and a first atrous rate, the third convolution branch uses a third convolution kernel having a first size and a second atrous rate, and the fourth convolution branch uses a fourth convolution kernel having a first size and a third atrous rate.

[0085] The equipment operating condition abnormality analysis module 150 is used to determine whether to generate an abnormal equipment operating condition warning prompt based on the cross-modal gated joint features of the real-time temperature time series associated implicit propagation aggregation representation vector and the vibration signal waveform semantic feature map. For example, when the cross-modal gated joint features indicate that the equipment operating condition is abnormal, the equipment operating condition abnormality analysis module 150 can issue an abnormal equipment operating condition warning prompt by sound, light, etc. When the cross-modal gated joint features indicate that the equipment operating condition is normal, the equipment operating condition abnormality analysis module 150 does not need to generate an abnormal equipment operating condition warning prompt.

[0086] It should be understood that in order to further conduct a joint analysis of the temperature time series associated implicit propagation aggregation representation vector and the vibration signal waveform semantic feature map, in order to comprehensively utilize the multimodal data features and obtain more comprehensive equipment operation status information, such as Figure 4 As shown, the equipment operating condition abnormality analysis module 150 may include a gated cross-modal joint encoding unit 151 and an operating status management result generation unit 152. In order to achieve effective fusion of temperature features and vibration features, the gated cross-modal joint encoding unit 151 introduces a gated cross-modal joint encoder based on a transformer-like structure, and combines the gated unit by imitating the self-attention mechanism of the Transformer structure to achieve effective fusion of different modal features. That is, the gated cross-modal joint encoding unit 151 is used to input the real-time temperature time series associated implicit propagation aggregation representation vector and the vibration signal waveform semantic feature map into the gated cross-modal joint encoder based on the transformer-like structure to obtain a multi-modal joint representation feature map of the equipment condition.

[0087] Exemplarily, the gated cross-modal joint encoding unit 151 is specifically used to: perform point convolution encoding and space convolution encoding on the vibration signal waveform semantic feature map, respectively, to obtain a vibration signal waveform space deepening semantic feature map; copy the vibration signal waveform space deepening semantic feature map, to obtain a backup vibration signal waveform space deepening semantic feature map; reshape the backup vibration signal waveform space deepening semantic feature map and the vibration signal waveform space deepening semantic feature map, respectively, to obtain a backup vibration signal waveform space deepening semantic feature matrix and a vibration signal waveform space deepening semantic feature matrix as a simulated query feature matrix and a simulated key feature matrix, respectively; calculate the simulated query feature matrix and the simulated key feature matrix. The product of the transposed matrices of the feature matrix is ​​obtained to obtain the local semantic autocorrelation library matrix of the vibration signal waveform; the local semantic autocorrelation library matrix of the vibration signal waveform is input into the attention gating unit based on the GELU function to obtain the local semantic autocorrelation library weight matrix of the vibration signal waveform; the real-time temperature time series associated implicit propagation aggregation representation vector is used as the value feature vector, and the product between the local semantic autocorrelation library weight matrix of the vibration signal waveform and the value feature vector is calculated to obtain the cross-modal interaction fusion conditional weight vector; the cross-modal interaction fusion conditional weight vector is used as the weight vector, and the semantic feature map of the vibration signal waveform is weighted channel by channel to obtain the multi-modal joint representation feature map of the equipment working condition.

[0088] That is to say, the gated cross-modal joint encoder based on the simulated transformer structure first performs point convolution encoding and spatial convolution encoding on the semantic feature map of the vibration signal waveform to extract its deep spatial features and enhance its spatial expression ability, and then constructs the simulated query feature matrix and the simulated key feature matrix. By calculating the product of the query feature matrix and the transposed key feature matrix, the internal correlation is revealed, and the local semantic autocorrelation library matrix of the vibration signal waveform is constructed. Then, the attention gating unit based on the GELU function is used to dynamically adjust the autocorrelation weight to achieve adaptive selection and enhancement of features and generate a weight matrix. Furthermore, by multiplying the weight matrix with the temperature time series associated implicit propagation aggregation representation vector to obtain the cross-modal interactive fusion conditional weight vector, conditional fusion between modalities is achieved, and the cross-modal association between temperature features and vibration features is enhanced. Finally, the cross-modal interactive fusion conditional weight vector is used to weight the semantic feature map of the vibration signal waveform channel by channel to integrate the multimodal features of temperature data and vibration signals and generate a multimodal joint representation feature map of equipment working conditions.

[0089] In the above embodiment, the gated cross-modal joint coding unit 151 is used to process the real-time temperature time series associated implicit propagation aggregation representation vector and the vibration signal waveform semantic feature map according to the following cross-modal joint coding formula to obtain the equipment working condition multi-modal joint representation feature map, wherein the cross-modal joint coding formula is:

[0090] F1=Conv 3×3 (Conv 1×1 (LN(F i )))

[0091] F 2=Copy(F1)

[0092]

[0093]

[0094]

[0095] Among them, F i represents the semantic feature map of the vibration signal waveform, LN represents the layer normalization operation, Conv 1×1 Represents point convolution, Conv 3×3 represents the convolution operation based on 3×3 convolution kernel, F1 represents the spatial deepening semantic feature map of vibration signal waveform, Copy(·) represents the copy operation, F2 represents the spatial deepening semantic feature map of backup vibration signal waveform, reshape(·) represents the feature shape reshaping, (·) T represents the transpose of a matrix, represents matrix multiplication operation, Gelu represents Gaussian error linear unit, Ma Represents the weight matrix of the local semantic autocorrelation library of the vibration signal waveform, V f represents the real-time temperature time series associated implicit propagation aggregation representation vector, V a represents the cross-modal interaction fusion condition weight vector, F i ′ represents the multi-modal joint representation feature diagram of equipment operating conditions.

[0096] The operation status management result generation unit 152 is used to input the multi-modal joint representation feature map of the equipment working condition into the operation status management result generator based on the classifier to obtain the operation status management result, and the operation status management result is used to indicate whether an abnormal equipment working condition warning prompt is generated. For example, when the operation status management result generator based on the classifier identifies that the equipment working condition is abnormal, it can output an operation status management result including an abnormal equipment working condition warning prompt, and the operation status management result generation unit 152 can issue an abnormal equipment working condition warning prompt through sound, light, etc. on this basis. When the operation status management result generator based on the classifier identifies that the equipment working condition is normal, it can output an operation status management result including no abnormal equipment working condition warning prompt, and on this basis, the operation status management result generation unit 152 does not need to generate an abnormal equipment working condition warning prompt.

[0097] It should be understood that during the operation of hydropower plant equipment, temperature and vibration are important indicators that reflect the health status of the equipment. Relying on any one feature alone may lead to inaccurate anomaly detection. By performing cross-modal gating joint analysis of temperature features and vibration features, the operating status of the equipment can be captured more comprehensively, thereby improving the accuracy of anomaly detection. This multi-modal joint representation feature map of the equipment operating condition is then input into a classifier-based operating status management result generator. This classifier can identify normal and abnormal equipment operating states based on the trained model, thereby obtaining an operating status management result.

[0098] Compared with the prior art, the hydropower plant equipment operation status management system based on data analysis provided in the embodiment of the present disclosure adopts artificial intelligence technology based on deep learning to perform real-time monitoring and data analysis on the operation conditions of hydropower plant equipment, capture the temperature time series change trend and equipment vibration mode of the hydropower plant equipment during operation, and then through cross-modal gating joint analysis of the temperature characteristics and vibration characteristics of the equipment, dig out the potential abnormal conditions of the equipment, thereby realizing abnormal early warning of the equipment operation status, which can not only improve the stability and efficiency of the operation of hydropower plant equipment, but also effectively reduce the operation cost and maintenance risk.

[0099] Another embodiment of the present disclosure relates to a method for managing the operating status of hydropower plant equipment based on data analysis, such as Figure 5 As shown, including:

[0100] S1, obtaining the operating condition data of the monitored hydropower plant equipment, wherein the operating condition data includes a time queue of real-time temperature values ​​and a vibration signal. Specifically, when the operating condition data can be acquired through a sensor network used to monitor the hydropower plant equipment, S1 can specifically include: obtaining the operating condition data of the monitored hydropower plant equipment acquired by the sensor network, wherein the operating condition data includes a time queue of real-time temperature values ​​and a vibration signal.

[0101] S2, extracting temperature features based on a local time scale from the time queue of the real-time temperature values ​​to obtain a sequence of implicit feature vectors associated with the local time series of the real-time temperature.

[0102] S3, performing feature forward propagation aggregation on the sequence of the real-time temperature local time series associated implicit feature vectors to obtain a real-time temperature time series associated implicit propagation aggregation representation vector.

[0103] S4, extracting waveform semantic features of the vibration signal to obtain a vibration signal waveform semantic feature map.

[0104] S5, determining whether to generate an abnormal equipment operating condition warning prompt based on the cross-modal gated joint features of the real-time temperature time series associated implicit propagation aggregation representation vector and the vibration signal waveform semantic feature map.

[0105] It should be noted that the specific operations of each step in the hydropower plant equipment operation status management method based on data analysis provided in this embodiment have been described in detail in the description of the hydropower plant equipment operation status management system based on data analysis provided in the above embodiment, and will not be repeated here.

[0106] Compared with the prior art, the method for managing the operating status of hydropower plant equipment based on data analysis provided in the embodiment of the present disclosure adopts artificial intelligence technology based on deep learning to perform real-time monitoring and data analysis on the operating conditions of hydropower plant equipment, capture the temperature time series change trend and equipment vibration mode of the hydropower plant equipment during operation, and then through cross-modal gating joint analysis of the temperature characteristics and vibration characteristics of the equipment, dig out the potential abnormal conditions of the equipment, thereby realizing abnormal early warning of the equipment operating status, which can not only improve the stability and efficiency of the operation of hydropower plant equipment, but also effectively reduce the operating cost and maintenance risk.

[0107] The basic principles of the technical solutions provided by the present disclosure are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present disclosure are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. must be possessed by each embodiment of the present disclosure. In addition, the specific details of the above embodiments are only for the purpose of illustration and ease of understanding, rather than limitation, and the above details do not limit the present disclosure to being implemented by adopting the above specific details.

[0108] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided in the present disclosure, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiment described above is only schematic. For example, the module division is only a logical function division, and there may be other division methods in actual implementation. The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0109] In addition, each functional module in each embodiment of the present disclosure may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0110] Those skilled in the art will appreciate that the above-mentioned embodiments are specific embodiments for implementing the present disclosure, and in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present disclosure.

[0111] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units included in the system can also be implemented by one unit through software or hardware.

[0112] Finally, it should be noted that the above description has been given for the purpose of illustration and description. In addition, the above embodiments are only used to illustrate the technical solution of the present disclosure and are not intended to limit it. Although the present disclosure is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present disclosure can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present disclosure.

Claims

1. A hydropower plant equipment operation status management system based on data analysis, characterized in that: The management system comprises: A hydropower plant equipment operating condition monitoring module, used to obtain the operating condition data of the monitored hydropower plant equipment, wherein the operating condition data includes a time queue of real-time temperature values ​​and a vibration signal; A temperature local time series feature extraction module is used to extract temperature features based on a local time scale from the time queue of the real-time temperature value to obtain a sequence of implicit feature vectors associated with the real-time temperature local time series; A forward propagation aggregation module, used for performing feature forward propagation aggregation on the sequence of the real-time temperature local time series associated implicit feature vectors to obtain a real-time temperature time series associated implicit propagation aggregation representation vector; A vibration signal waveform feature extraction module, used to extract the waveform semantic features of the vibration signal to obtain a vibration signal waveform semantic feature map; The equipment operating condition abnormality analysis module is used to determine whether to generate an equipment operating condition abnormality warning prompt based on the cross-modal gated joint features of the real-time temperature time series associated implicit propagation aggregation representation vector and the vibration signal waveform semantic feature map.

2. The management system according to claim 1, characterized in that: The temperature local time series feature extraction module is specifically used for: The time queue of the real-time temperature value is segmented into a predetermined time scale and then input into a sequence encoder based on an RNN model to obtain a sequence of implicit feature vectors associated with the local temporal sequence of the real-time temperature.

3. The management system according to claim 1, characterized in that: The forward propagation aggregation module is specifically used for: The sequence of the real-time temperature local temporal association implicit feature vectors is input into a feature forward propagation network based on significant temporal attenuation of node energy to obtain the real-time temperature temporal association implicit propagation aggregate representation vector.

4. The management system according to claim 3, characterized in that: The forward propagation aggregation module comprises: An energy significant descriptor calculation unit is used to calculate the energy significant descriptor of each real-time temperature local time series associated implicit feature vector based on the mean and variance of each real-time temperature local time series associated implicit feature vector in the sequence of the real-time temperature local time series associated implicit feature vector, so as to obtain a time series of energy significant descriptors; A timestamp marking unit, used for extracting the timestamp of each real-time temperature local time series associated implicit feature vector in the sequence of the real-time temperature local time series associated implicit feature vector, so as to obtain a time series of timestamps; A historical node feature division unit, used to use the real-time temperature local time series associated implicit feature vector corresponding to the current time point in the sequence of the real-time temperature local time series associated implicit feature vector as the current real-time temperature local time series associated implicit feature vector, and use the real-time temperature local time series associated implicit feature vectors corresponding to other time points as the historical real-time temperature local time series associated implicit feature vectors, so as to obtain a sequence of the current real-time temperature local time series associated implicit feature vector and the historical real-time temperature local time series associated implicit feature vector; An attenuation factor calculation unit, used to calculate the energy time attenuation factor of each historical real-time temperature local time series associated implicit feature vector in the sequence of the historical real-time temperature local time series associated implicit feature vector relative to the current real-time temperature local time series associated implicit feature vector, so as to obtain a time series of energy time attenuation factors; A historical node propagation aggregation unit is used to perform time series propagation aggregation on the sequence of the local time series associated implicit feature vectors of the historical real-time temperature using the time series of the energy significant descriptor as a positive adjustment factor and the time series of the energy time decay factor as a negative adjustment factor, so as to obtain a historical node real-time temperature time series aggregation feature vector; The global time series aggregation unit is used to fuse the historical node real-time temperature time series aggregation feature vector and the current real-time temperature local time series associated implicit feature vector to obtain the real-time temperature time series associated implicit propagation aggregation representation vector.

5. The management system according to claim 4, characterized in that: The energy significant descriptor calculation unit is specifically used to: The expected value of the fourth power of the difference between each eigenvalue in the real-time temperature local time series associated implicit feature vector and its eigenmean is calculated, and the expected value is divided by the square of the eigenvariance of the real-time temperature local time series associated implicit feature vector to obtain the energy significant descriptor.

6. The management system according to claim 4, characterized in that: The attenuation factor calculation unit is specifically used for: Calculating a timestamp difference between the historical real-time temperature local time series associated implicit feature vector and the current real-time temperature local time series associated implicit feature vector; The timestamp difference is rounded down and divided by a preset time decay period and then squared, and the square result is multiplied by a preset decay rate to obtain an initial energy decay coefficient; An exponential function value with a natural constant as the base and the initial energy decay coefficient as the exponent is calculated to obtain the energy time decay factor.

7. The management system according to claim 1, characterized in that: The vibration signal waveform feature extraction module is specifically used for: The vibration signal is input into a vibration signal waveform semantic feature extractor based on a dilated convolutional neural network model to obtain the vibration signal waveform semantic feature map.

8. The management system according to claim 1, characterized in that: The equipment operation condition abnormality analysis module includes: A gated cross-modal joint encoding unit, used for inputting the real-time temperature time series associated implicit propagation aggregation representation vector and the vibration signal waveform semantic feature map into a gated cross-modal joint encoder based on a simulated converter structure to obtain a multi-modal joint representation feature map of the equipment working condition; An operation status management result generating unit is used to input the multimodal joint representation feature map of the equipment operating condition into the classifier-based operation status management result generator to obtain an operation status management result, wherein the operation status management result is used to indicate whether an abnormal warning prompt of the equipment operating condition is generated.

9. The management system according to claim 8, characterized in that: The gated cross-modal joint encoding unit is specifically used for: Performing point convolution coding and space convolution coding on the vibration signal waveform semantic feature map respectively to obtain a vibration signal waveform space deepening semantic feature map; Copying the vibration signal waveform space deepening semantic feature map to obtain a backup vibration signal waveform space deepening semantic feature map; Reshape the backup vibration signal waveform space deepening semantic feature map and the vibration signal waveform space deepening semantic feature map respectively, so as to obtain a backup vibration signal waveform space deepening semantic feature matrix and a vibration signal waveform space deepening semantic feature matrix as a simulated query feature matrix and a simulated key feature matrix respectively; Calculating the product of the pseudo-query feature matrix and the transposed matrix of the pseudo-key feature matrix to obtain a vibration signal waveform local semantic autocorrelation library matrix; Inputting the vibration signal waveform local semantic autocorrelation library matrix into the attention gating unit based on the GELU function to obtain the vibration signal waveform local semantic autocorrelation library weight matrix; Taking the real-time temperature time series associated implicit propagation aggregation representation vector as the value feature vector, calculating the product between the vibration signal waveform local semantic autocorrelation library weight matrix and the value feature vector to obtain a cross-modal interaction fusion condition weight vector; The cross-modal interactive fusion condition weight vector is used as a weight vector to weight the vibration signal waveform semantic feature map channel by channel to obtain the multi-modal joint representation feature map of the equipment working condition.

10. A method for managing the operating status of hydropower plant equipment based on data analysis, characterized in that: The management method comprises: Acquire the operating condition data of the monitored hydropower plant equipment, wherein the operating condition data includes a time queue of real-time temperature values ​​and a vibration signal; Extracting temperature features based on a local time scale from the time queue of the real-time temperature values ​​to obtain a sequence of implicit feature vectors associated with local time series of the real-time temperature; Performing feature forward propagation aggregation on the sequence of the real-time temperature local time series associated implicit feature vectors to obtain a real-time temperature time series associated implicit propagation aggregation representation vector; Extracting the waveform semantic features of the vibration signal to obtain a vibration signal waveform semantic feature map; Based on the cross-modal gated joint features of the real-time temperature time series associated implicit propagation aggregation representation vector and the vibration signal waveform semantic feature map, it is determined whether to generate an abnormal equipment operating condition warning prompt.

Images