Automatic monitoring system and method for hydraulic power plant

Abstract

The embodiment of the invention relates to the technical field of hydraulic power plant monitoring, and provides an automatic monitoring system and method for a hydraulic power plant, and the system employs a sensing technology to carry out the data collection of the operation state of a water turbine, and employs a deep learning algorithm to carry out the time sequence analysis of the collected operation state data set of the water turbine. According to the method, the time sequence change mode characteristics of all the operation state parameters of the water turbine are extracted, then dominant aggregation is carried out on all the operation state parameter characteristics to obtain multi-mode information of the operation state of the water turbine, and therefore intelligent identification and early warning of the abnormal operation state of the water turbine are achieved. The real-time performance and accuracy of hydraulic power plant equipment monitoring are effectively improved, the phenomena of false alarm and missing alarm are reduced, and the safety and stability of a hydraulic power plant are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of hydropower plant monitoring, and in particular to an automated monitoring system and method for a hydropower plant. Background Art

[0002] As energy demand continues to grow and the efficient use of water resources has become a global focus, hydropower, as a clean and renewable form of energy, has become increasingly important in the energy structure. As the core facility for hydropower energy conversion, the operating efficiency, safety and stability of hydropower plants are directly related to the overall performance of the power system and the reliability of energy supply. The key equipment in hydropower plants, turbines, are often affected by various factors such as water flow conditions, mechanical wear, and electrical failures during operation. Therefore, effective monitoring and management of the operating status of turbines is of great significance for preventing failures, reducing downtime, and reducing operation and maintenance costs.

[0003] Traditional monitoring methods for hydropower plants mainly rely on manual inspections and regular maintenance. However, this method is not only inefficient, but also often causes equipment damage and even safety accidents due to the inability to detect and respond to sudden failures in a timely manner. At present, although some intelligent monitoring systems for hydropower plants have introduced automated detection technology, most of them are based on preset thresholds to judge equipment status and abnormal warnings, lacking the ability to conduct in-depth mining and comprehensive evaluation of the collected data, and may still lead to false alarms or missed reports. Summary of the invention

[0004] The present disclosure aims to solve at least one of the problems existing in the prior art and provides a hydropower plant automation monitoring system and method.

[0005] In one aspect of the present disclosure, a method for automated monitoring of a hydropower plant is provided, the method comprising:

[0006] Acquire the operating status data of the monitored turbine in the hydropower plant collected by the sensor group; the operating status data includes bearing temperature value, oil temperature value, oil pressure value, cooling water temperature value, flow value and vibration amplitude value;

[0007] Using a wireless communication module, transmitting the operating status data to a central control room;

[0008] In the central control room, the operating status data is time-series encoded based on the parameter sample granularity to obtain the corresponding bearing temperature time-series correlation feature vector, oil temperature time-series correlation feature vector, oil pressure time-series correlation feature vector, cooling water temperature time-series correlation feature vector, flow rate time-series correlation feature vector and vibration amplitude time-series correlation feature vector;

[0009] In the central control room, the bearing temperature time series correlation feature vector, the oil temperature time series correlation feature vector, the oil pressure time series correlation feature vector, the cooling water temperature time series correlation feature vector, the flow rate time series correlation feature vector and the vibration amplitude time series correlation feature vector are subjected to feature significant aggregation to obtain a multi-modal time series significant aggregation representation vector of the turbine operation state;

[0010] In the central control room, it is determined whether to generate an abnormal operating status warning prompt based on the multi-modal time series significant aggregation representation vector of the turbine operating status.

[0011] Optionally, the operating status data is time-series encoded based on the parameter sample granularity to obtain corresponding bearing temperature time-series correlation feature vectors, oil temperature time-series correlation feature vectors, oil pressure time-series correlation feature vectors, cooling water temperature time-series correlation feature vectors, flow rate time-series correlation feature vectors, and vibration amplitude time-series correlation feature vectors, including:

[0012] The operating status data is sorted according to the time dimension and the parameter sample dimension to obtain the corresponding time series of bearing temperature values, time queue of oil temperature values, time queue of oil pressure values, time queue of cooling water temperature values, time queue of flow values, and time queue of vibration amplitude values;

[0013] The time series of the bearing temperature values, the time queue of the oil temperature values, the time queue of the oil pressure values, the time queue of the cooling water temperature values, the time queue of the flow value and the time queue of the vibration amplitude value are respectively input into the operating status time series feature extractor based on the Bi-LSTM model to obtain the corresponding bearing temperature time series association feature vector, the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow rate time series association feature vector and the vibration amplitude time series association feature vector.

[0014] Optionally, the bearing temperature time series correlation feature vector, the oil temperature time series correlation feature vector, the oil pressure time series correlation feature vector, the cooling water temperature time series correlation feature vector, the flow rate time series correlation feature vector and the vibration amplitude time series correlation feature vector are subjected to feature significant aggregation to obtain a multi-modal time series significant aggregation representation vector of the turbine operation state, including:

[0015] Performing cluster analysis on the bearing temperature time series correlation feature vector, the oil temperature time series correlation feature vector, the oil pressure time series correlation feature vector, the cooling water temperature time series correlation feature vector, the flow rate time series correlation feature vector, and the vibration amplitude time series correlation feature vector to obtain a self-supervised cluster representation vector of the operating state;

[0016] Based on the potential clustering contribution scores of the bearing temperature time series association feature vector, the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow rate time series association feature vector and the vibration amplitude time series association feature vector relative to the operating state self-supervised clustering representation vector, feature explicit aggregation is performed to obtain the multimodal time series significant aggregation representation vector of the turbine operating state.

[0017] Optionally, cluster analysis is performed on the bearing temperature time series association feature vector, the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow rate time series association feature vector, and the vibration amplitude time series association feature vector to obtain a self-supervised cluster representation vector of the operating state, including:

[0018] Calculate the positional mean vector among the bearing temperature time series association feature vector, the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow rate time series association feature vector and the vibration amplitude time series association feature vector, and use the positional mean vector as the operating status self-supervised clustering representation vector.

[0019] Optionally, based on the potential clustering contribution scores of the bearing temperature time series association feature vector, the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow rate time series association feature vector, and the vibration amplitude time series association feature vector relative to the operating state self-supervised clustering representation vector, feature explicit aggregation is performed to obtain the multimodal time series significant aggregation representation vector of the turbine operating state, including:

[0020] Calculate the potential clustering contribution scores of the bearing temperature time series association feature vector, the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow rate time series association feature vector, and the vibration amplitude time series association feature vector relative to the operating state self-supervised clustering representation vector to obtain a set of potential clustering contribution scores of operating state parameter features;

[0021] Arranging the set of running state parameter feature potential cluster contribution scores into a running state parameter feature potential cluster contribution field distribution vector;

[0022] Performing explicit modeling on the distribution vector of the potential clustering contribution field of the operating state parameter feature based on the self-attention mechanism to obtain a modulation weight vector of the potential clustering contribution field of the operating state parameter feature;

[0023] Taking each eigenvalue in the modulation weight vector of the potential clustering contribution field of the operating state parameter features as the weight, the bearing temperature time series association feature vector, the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow rate time series association feature vector and the vibration amplitude time series association feature vector are weightedly aggregated to obtain the multi-modal time series significant aggregation representation vector of the turbine operating state.

[0024] Optionally, calculating the potential clustering contribution scores of the bearing temperature time series association feature vector, the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow rate time series association feature vector, and the vibration amplitude time series association feature vector relative to the operating state self-supervised clustering representation vector to obtain a set of potential clustering contribution scores of operating state parameter features, including:

[0025] Calculate the dotted vector between the bearing temperature time series associated feature vector and the running state self-supervised clustering representation vector, and calculate the base-two logarithm of the absolute value of each eigenvalue in the dotted vector to obtain the bearing temperature potential clustering contribution weight vector;

[0026] Calculate the dot product vector between the bearing temperature potential clustering contribution weight vector and the bearing temperature time series associated eigenvector, and calculate the exponential function value with e as the base and the sum of the eigenvalues ​​of the dot product vector as the exponent to obtain the bearing temperature feature potential clustering contribution score included in the set of operating state parameter feature potential clustering contribution scores.

[0027] Optionally, the running state parameter feature potential cluster contribution field distribution vector is explicitly modeled based on a self-attention mechanism to obtain a running state parameter feature potential cluster contribution field modulation weight vector, including:

[0028] Using the query embedding matrix, the key embedding matrix and the value embedding matrix respectively, the potential clustering contribution field distribution vector of the operating state parameter feature is embedded and transformed to obtain the corresponding query vector, key vector and value vector;

[0029] Multiply the query vector by the transposed vector of the key vector and divide by the square root of the scale of the key vector to obtain an attention score matrix;

[0030] The attention score matrix is ​​operated by a softmax function and then multiplied by the value vector to obtain the potential clustering contribution field modulation weight vector of the operating state parameter features.

[0031] Optionally, determining whether to generate an abnormal operation state warning prompt based on the multi-modal time series significant aggregation representation vector of the operation state of the turbine includes:

[0032] Inputting the multimodal time series significant aggregation representation vector of the turbine operating state into a turbine state monitoring module based on a classifier to obtain a state monitoring result, wherein the state monitoring result is used to indicate whether the operating state of the monitored turbine is abnormal;

[0033] In response to the status monitoring result indicating that the operating status of the monitored turbine is abnormal, an abnormal operating status warning prompt is generated.

[0034] Optionally, the multi-modal time series significant aggregation representation vector of the turbine operation state is input into a classifier-based turbine state monitoring module to obtain a state monitoring result, including:

[0035] Use the fully connected layer of the classifier to perform fully connected encoding on the multi-modal temporal significant aggregation representation vector of the operating state of the turbine to obtain a fully connected encoded multi-modal temporal significant aggregation representation vector of the operating state of the turbine;

[0036] The fully connected coded turbine operating state multimodal time series significant aggregation representation vector is input into the Softmax classification function of the classifier to obtain the state monitoring result.

[0037] Another aspect of the present disclosure provides a hydropower plant automation monitoring system, the hydropower plant automation monitoring system comprising:

[0038] An operating status data acquisition module is used to acquire operating status data of the monitored turbine in the hydropower plant collected by the sensor group; the operating status data includes bearing temperature value, oil temperature value, oil pressure value, cooling water temperature value, flow value and vibration amplitude value;

[0039] An operation status data transmission module, used to transmit the operation status data to a central control room using a wireless communication module;

[0040] An operation status data time series encoding module is used to perform time series encoding on the operation status data based on parameter sample granularity in the central control room to obtain corresponding bearing temperature time series correlation feature vectors, oil temperature time series correlation feature vectors, oil pressure time series correlation feature vectors, cooling water temperature time series correlation feature vectors, flow rate time series correlation feature vectors and vibration amplitude time series correlation feature vectors;

[0041] An operating status data feature significant aggregation module is used to perform feature significant aggregation on the bearing temperature time series correlation feature vector, the oil temperature time series correlation feature vector, the oil pressure time series correlation feature vector, the cooling water temperature time series correlation feature vector, the flow rate time series correlation feature vector and the vibration amplitude time series correlation feature vector in the central control room to obtain a multi-modal time series significant aggregation representation vector of the turbine operating status;

[0042] The abnormal operation state warning prompt generation module is used to determine whether to generate an abnormal operation state warning prompt in the central control room based on the multi-modal time series significant aggregation representation vector of the turbine operation state.

[0043] Compared with the prior art, the present invention adopts sensing technology to collect data on the operating status of the turbine, and uses a deep learning algorithm to perform time series analysis on the collected operating status data set of the turbine, so as to extract the time series change pattern characteristics of various operating status parameters of the turbine, and then obtain multimodal information of the operating status of the turbine by explicitly aggregating the characteristics of various operating status parameters, thereby realizing intelligent identification and early warning of abnormal operating status of the turbine, effectively improving the real-time and accuracy of equipment monitoring in hydropower plants, reducing false alarms and missed alarms, and improving the safety and stability of hydropower plants. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] 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.

[0045] Figure 1 A flow chart of a hydropower plant automation monitoring method provided in one embodiment of the present disclosure;

[0046] Figure 2 A data flow diagram of a hydropower plant automation monitoring method provided by another embodiment of the present disclosure;

[0047] Figure 3 A flowchart of step S3 in a hydropower plant automation monitoring method provided by another embodiment of the present disclosure;

[0048] Figure 4 A flowchart of step S4 in a hydropower plant automation monitoring method provided by another embodiment of the present disclosure;

[0049] Figure 5 A schematic structural diagram of an automated monitoring system for a hydropower plant provided in accordance with another embodiment of the present disclosure. DETAILED DESCRIPTION

[0050] 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.

[0051] Unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0052] Although the present disclosure makes various references to certain modules in the system according to embodiments of the present disclosure, any number of different modules can be used and run on a user terminal and / or server. The modules are illustrative only, and different aspects of the systems and methods can use different modules.

[0053] The present disclosure uses a flow chart to illustrate the operations performed according to the method of the embodiments of the present disclosure. It should be understood that the preceding or following operations are not necessarily performed precisely in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. At the same time, other operations may also be added to these processes, or one or more operations may be removed from these processes.

[0054] In view of the problems existing in the existing monitoring methods for hydropower plants, the present invention provides an automated monitoring system and method for hydropower plants, the technical concept of which is: using sensing technology to collect data on the operating status of the turbine, and using a deep learning algorithm to perform time series analysis on the collected data set of the operating status of the turbine, so as to extract the time series change pattern characteristics of various operating status parameters of the turbine, and then through explicit aggregation of the characteristics of various operating status parameters, to obtain multi-modal information of the operating status of the turbine, so as to realize intelligent identification and early warning of abnormal operating status of the turbine. In this way, the real-time and accuracy of the equipment monitoring of the hydropower plant can be effectively improved, the false alarm and missed alarm phenomenon can be reduced, and the safety and stability of the hydropower plant can be improved.

[0055] One embodiment of the present disclosure relates to a method for automated monitoring of a hydropower plant, the process of which is as follows: Figure 1 As shown, the data flow is as follows Figure 2 shown.

[0056] Combined Figure 1 and Figure 2 The hydropower plant automation monitoring method provided in this embodiment includes the following steps:

[0057] Step S1, obtaining the operating status data of the monitored turbine in the hydropower plant collected by the sensor group; the operating status data includes bearing temperature value, oil temperature value, oil pressure value, cooling water temperature value, flow value and vibration amplitude value.

[0058] Specifically, as the core equipment of a hydropower plant, the operating status of the turbine directly affects the power generation efficiency and safety of the entire hydropower plant. Parameters such as bearing temperature, oil temperature, oil pressure, cooling water temperature, flow rate and vibration amplitude are important indicators reflecting the operating status of the turbine, which respectively reveal the bearing wear of the turbine, the performance of the lubricating oil, the oil supply capacity of the lubrication system, the efficiency of the cooling system, the flow rate change of the water flow, and the size of the mechanical vibration. By real-time monitoring and data analysis of these key parameters, the main characteristics of the turbine operation process can be fully and accurately grasped to ensure the comprehensiveness of monitoring.

[0059] Step S2: using the wireless communication module to transmit the operation status data to the central control room.

[0060] Specifically, after obtaining the operating status data of the monitored turbine in the hydropower plant, step S2 uses the wireless communication module to transmit the operating status data of the monitored turbine in the hydropower plant to the central control room, so as to use the rich computing resources of the central control room to further process and analyze the operating status data, thereby realizing the monitoring of the turbine operating status and abnormality identification.

[0061] Step S3, in the central control room, the operating status data is time-series encoded based on the parameter sample granularity to obtain the corresponding bearing temperature time-series correlation feature vector, oil temperature time-series correlation feature vector, oil pressure time-series correlation feature vector, cooling water temperature time-series correlation feature vector, flow rate time-series correlation feature vector and vibration amplitude time-series correlation feature vector.

[0062] Specifically, considering that the turbine operation status data has obvious time series characteristics, therefore, by performing time series encoding on the operation status data based on the parameter sample granularity, it can help to better understand and analyze the time dependency in these data, so as to more accurately extract the characteristics reflecting the turbine operation status. Here, by arranging the operation status data according to the time and parameter sample granularity, independent time series sequences can be formed, such as the time series of bearing temperature values, the time series of oil temperature values, the time series of oil pressure values, the time series of cooling water temperature values, the time series of flow values, and the time series of vibration amplitude values, etc., which not only retains the relationship between the changes of various parameters over time, but also ensures the synchronization between different parameters, thereby improving the accuracy of monitoring the operation status of the turbine.

[0063] Exemplary, combined Figure 3 In step S3, the operating status data is time-series encoded based on the parameter sample granularity to obtain the corresponding bearing temperature time-series correlation feature vector, oil temperature time-series correlation feature vector, oil pressure time-series correlation feature vector, cooling water temperature time-series correlation feature vector, flow rate time-series correlation feature vector and vibration amplitude time-series correlation feature vector, including step S31 and step S32.

[0064] Step S31, sorting the operating status data according to the time dimension and the parameter sample dimension to obtain the corresponding time series of bearing temperature values, time queue of oil temperature values, time queue of oil pressure values, time queue of cooling water temperature values, time queue of flow values ​​and time queue of vibration amplitude values.

[0065] Specifically, considering that the operating status data of the turbine is multidimensional, nonlinear time-series variation data, in order to more accurately capture the time-series variation characteristics of various operating status parameters, step S31 organizes the operating status data according to the time dimension and the parameter sample dimension, so as to organize the various operating status parameters therein into independent time series data, while maintaining the time sequence relationship of various operating status parameters and the time correspondence between the parameters, thereby obtaining the corresponding time series of bearing temperature values, time queue of oil temperature values, time queue of oil pressure values, time queue of cooling water temperature values, time queue of flow values ​​and time queue of vibration amplitude values, so as to facilitate the data time series feature extraction in subsequent steps.

[0066] Step S32, input the time series of bearing temperature values, the time queue of oil temperature values, the time queue of oil pressure values, the time queue of cooling water temperature values, the time queue of flow values ​​and the time queue of vibration amplitude values ​​into the operating status time series feature extractor based on the Bi-LSTM model respectively, and obtain the corresponding bearing temperature time series association feature vector, oil temperature time series association feature vector, oil pressure time series association feature vector, cooling water temperature time series association feature vector, flow rate time series association feature vector and vibration amplitude time series association feature vector.

[0067] Specifically, in order to ensure the comprehensiveness and accuracy of data timing analysis, this embodiment adopts a Bi-LSTM model as an operating status timing feature extractor, and performs feature extraction processing on the time series of bearing temperature values, the time queue of oil temperature values, the time queue of oil pressure values, the time queue of cooling water temperature values, the time queue of flow values, and the time queue of vibration amplitude values, respectively, so as to utilize the bidirectional memory capability of the Bi-LSTM model, and perform bidirectional timing modeling on the input time queue data through forward and backward LSTM units, so as to fully capture the bidirectional timing dependency in the time queue data, so as to more comprehensively learn the timing change patterns of various operating status parameters of the turbine, and obtain the corresponding bearing temperature timing association feature vector, oil temperature timing association feature vector, oil pressure timing association feature vector, cooling water temperature timing association feature vector, flow timing association feature vector, and vibration amplitude timing association feature vector.

[0068] Among them, the Bi-LSTM model is a neural network structure based on a bidirectional long short-term memory (Bi-directional Long Short-Term Memory) network, which can simultaneously process the forward and backward information of sequence data. It consists of two independent LSTM units, one is a forward LSTM unit for processing the positive order of the sequence, and the other is a backward LSTM unit for processing the reverse order of the sequence. The outputs of these two units are combined as the final prediction result of the model output.

[0069] Step S4, in the central control room, perform feature significant aggregation on the bearing temperature time series correlation feature vector, the oil temperature time series correlation feature vector, the oil pressure time series correlation feature vector, the cooling water temperature time series correlation feature vector, the flow rate time series correlation feature vector and the vibration amplitude time series correlation feature vector to obtain a multi-modal time series significant aggregation representation vector of the turbine operating state.

[0070] Specifically, in order to grasp the overall operating state of the turbine from a global perspective, step S4 further fuses the time series feature information of various operating parameters of the turbine to obtain a multi-modal comprehensive feature representation of the turbine operating state.

[0071] In particular, this embodiment provides a feature significance aggregation method based on cluster analysis. First, the bearing temperature time series correlation feature vector, oil temperature time series correlation feature vector, oil pressure time series correlation feature vector, cooling water temperature time series correlation feature vector, flow time series correlation feature vector and vibration amplitude time series correlation feature vector are clustered by unsupervised learning method to reveal the intrinsic correlation distribution between the time series features of each operating state parameter, and generate a self-supervised cluster representation vector of the operating state, so as to represent the main operating state mode of the turbine. Then, the potential cluster contribution score of the time series correlation feature vector of each operating state parameter relative to the self-supervised cluster representation vector of the operating state is calculated to reflect its contribution to the cluster feature and quantify its role in the clustering process, thereby providing a basis for subsequent feature selection and weight allocation. Then, the calculated potential cluster contribution scores are integrated into vector form, and the self-attention mechanism is used to explicitly model them. A more targeted weight vector is generated through global autocorrelation analysis, thereby realizing weighted aggregation of the time series features of each operating state parameter, so as to obtain a multi-modal time series significant aggregation representation vector of the operating state of the turbine. In this way, it can be ensured that the importance and uniqueness of various parameters in reflecting the operating status of the turbine are fully considered during the aggregation process, thereby improving the accuracy and effectiveness of multimodal information fusion.

[0072] In other words, combined Figure 4 In step S4, the bearing temperature time series correlation feature vector, the oil temperature time series correlation feature vector, the oil pressure time series correlation feature vector, the cooling water temperature time series correlation feature vector, the flow rate time series correlation feature vector and the vibration amplitude time series correlation feature vector are subjected to feature significant aggregation to obtain a multimodal time series significant aggregation representation vector of the turbine operating state, which may include step S41 and step S42.

[0073] Step S41, cluster analysis is performed on the bearing temperature time series correlation feature vector, the oil temperature time series correlation feature vector, the oil pressure time series correlation feature vector, the cooling water temperature time series correlation feature vector, the flow rate time series correlation feature vector and the vibration amplitude time series correlation feature vector to obtain a self-supervised cluster representation vector of the operating status.

[0074] Among them, step S41 specifically includes: calculating the position mean vector between the bearing temperature timing association feature vector, the oil temperature timing association feature vector, the oil pressure timing association feature vector, the cooling water temperature timing association feature vector, the flow timing association feature vector and the vibration amplitude timing association feature vector, and using the position mean vector as the operating status self-supervised clustering representation vector.

[0075] Step S42, based on the potential clustering contribution scores of the bearing temperature time series association feature vector, the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow rate time series association feature vector and the vibration amplitude time series association feature vector relative to the operating state self-supervised clustering representation vector, feature explicit aggregation is performed to obtain a multimodal time series significant aggregation representation vector of the turbine operating state.

[0076] Among them, step S42 specifically includes the following sub-steps: calculating the potential clustering contribution scores of the bearing temperature time series association feature vector, the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow time series association feature vector and the vibration amplitude time series association feature vector relative to the operating state self-supervised clustering representation vector, and obtaining a set of operating state parameter feature potential clustering contribution scores. The set of operating state parameter feature potential clustering contribution scores is arranged as an operating state parameter feature potential clustering contribution field distribution vector. The operating state parameter feature potential clustering contribution field distribution vector is explicitly modeled based on the self-attention mechanism to obtain the operating state parameter feature potential clustering contribution field modulation weight vector. Using each eigenvalue in the operating state parameter feature potential clustering contribution field modulation weight vector as the weight, the bearing temperature time series association feature vector, the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow time series association feature vector and the vibration amplitude time series association feature vector are weightedly aggregated to obtain the multi-modal time series significant aggregation representation vector of the turbine operating state.

[0077] In the sub-steps included in step S42, the potential clustering contribution scores of the bearing temperature time series association feature vector, the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow rate time series association feature vector and the vibration amplitude time series association feature vector relative to the operating state self-supervised clustering representation vector are calculated to obtain a set of potential clustering contribution scores of operating state parameter features, including: calculating the dot division vector between the bearing temperature time series association feature vector and the operating state self-supervised clustering representation vector, and calculating the base-two logarithmic value of the absolute value of each eigenvalue in the dot division vector to obtain the bearing temperature potential clustering contribution weight vector; calculating the dot product vector between the bearing temperature potential clustering contribution weight vector and the bearing temperature time series association feature vector, and calculating the exponential function value with base e and the sum of each eigenvalue of the dot product vector as the exponent to obtain the bearing temperature feature potential clustering contribution score included in the set of potential clustering contribution scores of operating state parameter features.

[0078] It should be noted that by replacing the bearing temperature time series association feature vector with the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow rate time series association feature vector and the vibration amplitude time series association feature vector, respectively, the corresponding oil temperature potential clustering contribution weight vector, oil pressure potential clustering contribution weight vector, cooling water temperature potential clustering contribution weight vector, flow rate potential clustering contribution weight vector and vibration amplitude potential clustering contribution weight vector can be obtained, and then the set of operating state parameter feature potential clustering contribution scores including the oil temperature feature potential clustering contribution score, the oil pressure feature potential clustering contribution score, the cooling water temperature feature potential clustering contribution score, the flow rate feature potential clustering contribution score and the vibration amplitude feature potential clustering contribution score can be obtained respectively.

[0079] In the sub-steps included in step S42, explicit modeling is performed on the distribution vector of the potential clustering contribution field of the operating state parameter features based on the self-attention mechanism to obtain the modulation weight vector of the potential clustering contribution field of the operating state parameter features, including: using the query embedding matrix, the key embedding matrix and the value embedding matrix to embed the distribution vector of the potential clustering contribution field of the operating state parameter features to obtain the corresponding query vector, key vector and value vector; multiplying the query vector by the transposed vector of the key vector and dividing it by the square root of the scale of the key vector to obtain an attention score matrix; multiplying the attention score matrix by the value vector after passing the softmax function operation to obtain the modulation weight vector of the potential clustering contribution field of the operating state parameter features.

[0080] In other words, step S4 can perform feature significant aggregation on the bearing temperature time series correlation feature vector, the oil temperature time series correlation feature vector, the oil pressure time series correlation feature vector, the cooling water temperature time series correlation feature vector, the flow time series correlation feature vector and the vibration amplitude time series correlation feature vector through the following feature significant aggregation formula, thereby obtaining a multi-modal time series significant aggregation representation vector of the turbine operation state, wherein the feature significant aggregation formula includes:

[0081] X={x 1 , x 2 , ..., x i , ..., x 6}

[0082]

[0083] v d ={d 1 ;d 2 ;...;d i ;...;d 6}

[0084]

[0085] v q =W q v d

[0086] v k =W k v d

[0087] v v =W b v d

[0088]

[0089] Where X represents the bearing temperature time series correlation feature vector x 1 , oil temperature time series correlation feature vector x 2 , oil pressure time series correlation feature vector x 3 , cooling water temperature time series correlation feature vector x 4 , traffic time series correlation feature vector x 5 and the vibration amplitude time series correlation feature vector x 6 The set composed of x i represents each eigenvector in set X, i ranges from 1 to 6, x c represents the self-supervised clustering representation vector of the running status, x ij represents the feature vector x i The jth eigenvalue of cj Represents the running state self-supervised clustering representation vector x c The j-th eigenvalue of , |·| represents the absolute value, log represents the logarithmic function with base 2, and m represents the eigenvector x i The length of , exp(·) represents the exponential function with e as the base, d i Represents x i The corresponding running state parameter feature potential cluster contribution score in the set of running state parameter feature potential cluster contribution scores, d 1 , d 2 , ..., d 6 represents the potential clustering contribution score of each operating state parameter feature in the set of potential clustering contribution scores of operating state parameter features, namely, the potential clustering contribution score of bearing temperature feature, the potential clustering contribution score of oil temperature feature, the potential clustering contribution score of oil pressure feature, the potential clustering contribution score of cooling water temperature feature, the potential clustering contribution score of flow feature and the potential clustering contribution score of vibration amplitude feature, v d represents the potential clustering contribution field distribution vector of the operating state parameter characteristics, W q , W k and W bdenote the query embedding matrix, key embedding matrix and value embedding matrix respectively, and v q 、v k and v v denote the query vector, key vector, and value vector respectively, (·) T represents the transpose of the vector, Softmax represents the normalized exponential function, i.e., the softmax function, represents matrix multiplication, v w represents the modulation weight vector of the potential clustering contribution field of the operating state parameter characteristics, represents the ith eigenvalue in the modulation weight vector of the potential clustering contribution field of the operating state parameter feature, x f Represents the multi-modal time series significant aggregation representation vector of the turbine operating status.

[0090] Step S5: In the central control room, based on the multi-modal time series significant aggregation representation vector of the turbine operation status, determine whether to generate an abnormal operation status warning prompt.

[0091] Specifically, by analyzing the characteristic information in the multimodal time series significant aggregation representation vector of the turbine operating status, it is possible to reveal the complex patterns of the various operating parameters of the turbine over time, and based on this, determine whether the turbine is in an abnormal operating state, and thus decide whether it is necessary to generate an abnormal operating status warning.

[0092] Exemplarily, in step S5, based on the multimodal time series significant aggregation representation vector of the turbine operating status, it is determined whether to generate an abnormal operating status warning prompt, including: inputting the multimodal time series significant aggregation representation vector of the turbine operating status into a turbine status monitoring module based on a classifier to obtain a status monitoring result, and the status monitoring result is used to indicate whether there is an abnormality in the operating status of the monitored turbine; in response to the status monitoring result indicating that the operating status of the monitored turbine is abnormal, generating an abnormal operating status warning prompt.

[0093] Specifically, this embodiment adopts a pre-trained classifier model to construct a turbine state monitoring module, and uses the turbine state monitoring module to perform feature pattern learning on the multimodal time series significant aggregation representation vector of the turbine operating state, so as to classify and judge the current operating state of the turbine according to the learned feature information, and output the corresponding state monitoring result.

[0094] Exemplarily, the multimodal temporal significant aggregation representation vector of the operating state of the turbine is input into a turbine state monitoring module based on a classifier to obtain a state monitoring result, including: using the fully connected layer of the classifier to fully connect the multimodal temporal significant aggregation representation vector of the operating state of the turbine to obtain a fully connected encoded multimodal temporal significant aggregation representation vector of the operating state of the turbine; inputting the fully connected encoded multimodal temporal significant aggregation representation vector of the operating state of the turbine into the Softmax classification function of the classifier to obtain a state monitoring result.

[0095] Specifically, when the status monitoring results indicate that the operating status of the monitored turbine is abnormal, an abnormal operating status warning prompt is generated to promptly notify the staff to take appropriate measures to intervene, thereby effectively avoiding potential safety risks and equipment failures.

[0096] Compared with the prior art, the method for automated monitoring of a hydropower plant provided in the embodiment of the present disclosure adopts sensing technology to collect data on the operating status of a turbine, and uses a deep learning algorithm to perform time series analysis on the collected operating status data set of the turbine, so as to extract the time series change pattern characteristics of various operating status parameters of the turbine, and then obtain multimodal information of the operating status of the turbine by explicitly aggregating the characteristics of various operating status parameters, thereby realizing intelligent identification and early warning of abnormal operating status of the turbine, effectively improving the real-time and accuracy of equipment monitoring in the hydropower plant, reducing false alarms and missed alarms, and improving the safety and stability of the hydropower plant.

[0097] Another embodiment of the present disclosure relates to a hydropower plant automation monitoring system 100, such as Figure 5 As shown, it includes an operation status data acquisition module 110, an operation status data transmission module 120, an operation status data time series encoding module 130, an operation status data feature significant aggregation module 140, and an operation status abnormal warning prompt generation module 150.

[0098] The operating status data acquisition module 110 is used to acquire the operating status data of the monitored turbine in the hydropower plant collected by the sensor group; the operating status data includes bearing temperature value, oil temperature value, oil pressure value, cooling water temperature value, flow value and vibration amplitude value;

[0099] The operation status data transmission module 120 is used to transmit the operation status data to the central control room using the wireless communication module;

[0100] The operation status data time series encoding module 130 is used to perform time series encoding on the operation status data based on the parameter sample granularity in the central control room, and obtain the corresponding bearing temperature time series correlation feature vector, oil temperature time series correlation feature vector, oil pressure time series correlation feature vector, cooling water temperature time series correlation feature vector, flow rate time series correlation feature vector and vibration amplitude time series correlation feature vector;

[0101] The operation status data feature significant aggregation module 140 is used to perform feature significant aggregation on the bearing temperature time series correlation feature vector, the oil temperature time series correlation feature vector, the oil pressure time series correlation feature vector, the cooling water temperature time series correlation feature vector, the flow time series correlation feature vector and the vibration amplitude time series correlation feature vector in the central control room to obtain a multi-modal time series significant aggregation representation vector of the turbine operation status;

[0102] The abnormal operation state warning prompt generation module 150 is used to determine whether to generate an abnormal operation state warning prompt in the central control room based on the multi-modal time series significant aggregation representation vector of the turbine operation state.

[0103] The specific implementation method of the hydropower plant automation monitoring system provided in the embodiment of the present disclosure can be found in the description of the hydropower plant automation monitoring method provided in the embodiment of the present disclosure, and will not be repeated here.

[0104] Compared with the prior art, the hydropower plant automation monitoring system provided by the embodiment of the present disclosure adopts sensing technology to collect data on the operating status of the turbine, and uses a deep learning algorithm to perform time series analysis on the collected operating status data set of the turbine to extract the time series change pattern characteristics of various operating status parameters of the turbine, and then obtains multimodal information of the operating status of the turbine by explicitly aggregating the characteristics of various operating status parameters, thereby realizing intelligent identification and early warning of abnormal operating status of the turbine, effectively improving the real-time and accuracy of equipment monitoring in the hydropower plant, reducing false alarms and missed alarms, and improving the safety and stability of the hydropower plant.

[0105] 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.

Claims

1. A method for automated monitoring of a hydropower plant, characterized in that: The hydropower plant automation monitoring method comprises: Acquire the operating status data of the monitored turbine in the hydropower plant collected by the sensor group; the operating status data includes bearing temperature value, oil temperature value, oil pressure value, cooling water temperature value, flow value and vibration amplitude value; Using a wireless communication module, transmitting the operating status data to a central control room; In the central control room, the operating status data is time-series encoded based on the parameter sample granularity to obtain the corresponding bearing temperature time-series correlation feature vector, oil temperature time-series correlation feature vector, oil pressure time-series correlation feature vector, cooling water temperature time-series correlation feature vector, flow rate time-series correlation feature vector and vibration amplitude time-series correlation feature vector; In the central control room, the bearing temperature time series correlation feature vector, the oil temperature time series correlation feature vector, the oil pressure time series correlation feature vector, the cooling water temperature time series correlation feature vector, the flow rate time series correlation feature vector and the vibration amplitude time series correlation feature vector are subjected to feature significant aggregation to obtain a multi-modal time series significant aggregation representation vector of the turbine operation state; In the central control room, it is determined whether to generate an abnormal operating status warning prompt based on the multi-modal time series significant aggregation representation vector of the turbine operating status.

2. The hydropower plant automation monitoring method according to claim 1, characterized in that: The operating status data is time-series encoded based on the parameter sample granularity to obtain the corresponding bearing temperature time-series correlation feature vector, oil temperature time-series correlation feature vector, oil pressure time-series correlation feature vector, cooling water temperature time-series correlation feature vector, flow rate time-series correlation feature vector and vibration amplitude time-series correlation feature vector, including: The operating status data is sorted according to the time dimension and the parameter sample dimension to obtain the corresponding time series of bearing temperature values, time queue of oil temperature values, time queue of oil pressure values, time queue of cooling water temperature values, time queue of flow values, and time queue of vibration amplitude values; The time series of the bearing temperature values, the time queue of the oil temperature values, the time queue of the oil pressure values, the time queue of the cooling water temperature values, the time queue of the flow value and the time queue of the vibration amplitude value are respectively input into the operating status time series feature extractor based on the Bi-LSTM model to obtain the corresponding bearing temperature time series association feature vector, the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow rate time series association feature vector and the vibration amplitude time series association feature vector.

3. The hydropower plant automation monitoring method according to claim 1, characterized in that: The bearing temperature time series correlation feature vector, the oil temperature time series correlation feature vector, the oil pressure time series correlation feature vector, the cooling water temperature time series correlation feature vector, the flow rate time series correlation feature vector and the vibration amplitude time series correlation feature vector are subjected to feature significant aggregation to obtain a multi-modal time series significant aggregation representation vector of the turbine operation state, including: Performing cluster analysis on the bearing temperature time series correlation feature vector, the oil temperature time series correlation feature vector, the oil pressure time series correlation feature vector, the cooling water temperature time series correlation feature vector, the flow rate time series correlation feature vector, and the vibration amplitude time series correlation feature vector to obtain a self-supervised cluster representation vector of the operating state; Based on the potential clustering contribution scores of the bearing temperature time series association feature vector, the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow rate time series association feature vector and the vibration amplitude time series association feature vector relative to the operating state self-supervised clustering representation vector, feature explicit aggregation is performed to obtain the multimodal time series significant aggregation representation vector of the turbine operating state.

4. The hydropower plant automation monitoring method according to claim 3, characterized in that: Cluster analysis is performed on the bearing temperature time series correlation feature vector, the oil temperature time series correlation feature vector, the oil pressure time series correlation feature vector, the cooling water temperature time series correlation feature vector, the flow rate time series correlation feature vector, and the vibration amplitude time series correlation feature vector to obtain a self-supervised cluster representation vector of the operating state, including: Calculate the positional mean vector among the bearing temperature time series association feature vector, the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow rate time series association feature vector and the vibration amplitude time series association feature vector, and use the positional mean vector as the operating status self-supervised clustering representation vector.

5. The hydropower plant automation monitoring method according to claim 3, characterized in that: Based on the potential clustering contribution scores of the bearing temperature time series association feature vector, the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow rate time series association feature vector, and the vibration amplitude time series association feature vector relative to the operating state self-supervised clustering representation vector, the multi-modal time series significant aggregation representation vector of the turbine operating state is obtained, including: Calculate the potential clustering contribution scores of the bearing temperature time series association feature vector, the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow rate time series association feature vector, and the vibration amplitude time series association feature vector relative to the operating state self-supervised clustering representation vector to obtain a set of potential clustering contribution scores of operating state parameter features; Arranging the set of running state parameter feature potential cluster contribution scores into a running state parameter feature potential cluster contribution field distribution vector; Performing explicit modeling on the distribution vector of the potential cluster contribution field of the operating state parameter feature based on the self-attention mechanism to obtain a modulation weight vector of the potential cluster contribution field of the operating state parameter feature; Taking each eigenvalue in the modulation weight vector of the potential clustering contribution field of the operating state parameter features as the weight, the bearing temperature time series association feature vector, the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow rate time series association feature vector and the vibration amplitude time series association feature vector are weightedly aggregated to obtain the multi-modal time series significant aggregation representation vector of the turbine operating state.

6. The hydropower plant automation monitoring method according to claim 5, characterized in that: Calculating the potential clustering contribution scores of the bearing temperature time series association feature vector, the oil temperature time series association feature vector, the oil pressure time series association feature vector, the cooling water temperature time series association feature vector, the flow rate time series association feature vector, and the vibration amplitude time series association feature vector relative to the operating state self-supervised clustering representation vector, and obtaining a set of potential clustering contribution scores of operating state parameter features, including: Calculate the dotted vector between the bearing temperature time series associated feature vector and the running state self-supervised clustering representation vector, and calculate the base-two logarithm of the absolute value of each eigenvalue in the dotted vector to obtain the bearing temperature potential clustering contribution weight vector; Calculate the dot product vector between the bearing temperature potential clustering contribution weight vector and the bearing temperature time series associated eigenvector, and calculate the exponential function value with e as the base and the sum of the eigenvalues ​​of the dot product vector as the exponent to obtain the bearing temperature feature potential clustering contribution score included in the set of operating state parameter feature potential clustering contribution scores.

7. The method for automatic monitoring of a hydropower plant according to claim 5, characterized in that: The running state parameter feature potential cluster contribution field distribution vector is explicitly modeled based on the self-attention mechanism to obtain the running state parameter feature potential cluster contribution field modulation weight vector, including: Using the query embedding matrix, the key embedding matrix and the value embedding matrix respectively, the potential clustering contribution field distribution vector of the operating state parameter feature is embedded and transformed to obtain the corresponding query vector, key vector and value vector; Multiply the query vector by the transposed vector of the key vector and divide by the square root of the scale of the key vector to obtain an attention score matrix; The attention score matrix is ​​operated by a softmax function and then multiplied by the value vector to obtain the potential clustering contribution field modulation weight vector of the operating state parameter features.

8. The method for automatic monitoring of a hydropower plant according to claim 1, characterized in that: Determining whether to generate an abnormal operation state warning prompt based on the multi-modal time series significant aggregation representation vector of the operation state of the turbine includes: Inputting the multimodal time series significant aggregation representation vector of the turbine operating state into a turbine state monitoring module based on a classifier to obtain a state monitoring result, wherein the state monitoring result is used to indicate whether the operating state of the monitored turbine is abnormal; In response to the status monitoring result indicating that the operating status of the monitored turbine is abnormal, an abnormal operating status warning prompt is generated.

9. The hydropower plant automation monitoring method according to claim 8, characterized in that: The multi-modal time series significant aggregation representation vector of the turbine operation state is input into a turbine state monitoring module based on a classifier to obtain a state monitoring result, including: Use the fully connected layer of the classifier to perform fully connected encoding on the multi-modal temporal significant aggregation representation vector of the operating state of the turbine to obtain a fully connected encoded multi-modal temporal significant aggregation representation vector of the operating state of the turbine; The fully connected coded turbine operating state multimodal time series significant aggregation representation vector is input into the Softmax classification function of the classifier to obtain the state monitoring result.

10. A hydropower plant automation monitoring system, characterized in that: The hydropower plant automation monitoring system comprises: An operating status data acquisition module is used to acquire operating status data of the monitored turbine in the hydropower plant collected by the sensor group; the operating status data includes bearing temperature value, oil temperature value, oil pressure value, cooling water temperature value, flow value and vibration amplitude value; An operation status data transmission module, used to transmit the operation status data to a central control room using a wireless communication module; An operation status data time series encoding module is used to perform time series encoding on the operation status data based on parameter sample granularity in the central control room to obtain corresponding bearing temperature time series correlation feature vectors, oil temperature time series correlation feature vectors, oil pressure time series correlation feature vectors, cooling water temperature time series correlation feature vectors, flow rate time series correlation feature vectors and vibration amplitude time series correlation feature vectors; An operating status data feature significant aggregation module is used to perform feature significant aggregation on the bearing temperature time series correlation feature vector, the oil temperature time series correlation feature vector, the oil pressure time series correlation feature vector, the cooling water temperature time series correlation feature vector, the flow rate time series correlation feature vector and the vibration amplitude time series correlation feature vector in the central control room to obtain a multi-modal time series significant aggregation representation vector of the turbine operating status; The abnormal operation state warning prompt generation module is used to determine whether to generate an abnormal operation state warning prompt in the central control room based on the multi-modal time series significant aggregation representation vector of the turbine operation state.

Images