Substation equipment state intelligent sensing and early warning method based on multi-source data

Abstract

The invention discloses a substation equipment state intelligent sensing and early warning method based on multi-source data. The method comprises the steps that real-time operation data and historical abnormal state data of substation equipment are acquired; obtaining time sequence data based on all historical abnormal state data to form a reference template; comparing the real-time operation data of the substation equipment with the reference data based on a dynamic time warping algorithm, and calculating the similarity; judging whether the current running state is abnormal or not based on a similarity result, if yes, classifying an abnormal type, and analyzing and predicting a future running trend; and uploading the newly appearing exception to a database, and optimizing the algorithm. The method has the advantages that the real-time data and the historical abnormal data are combined, the operation state of the equipment is monitored in real time, the abnormity is accurately identified and classified based on the dynamic time warping algorithm, potential faults are predicted in advance, scientific basis is provided for equipment maintenance and management, and sudden faults are reduced.

Description

Technical Field

[0001] The present invention relates to real-time monitoring, early warning and prediction technology, and in particular to a method for intelligent perception and early warning of substation equipment status based on multi-source data. Background Art

[0002] The intelligent perception and early warning method for substation equipment status based on multi-source data uses a dynamic time warping algorithm to compare the real-time data of the current equipment with the reference template and calculate the similarity between the two. Based on the similarity results, the system determines whether the current operating status of the equipment is abnormal. If an abnormality is detected, the system will further classify the abnormality type, analyze the abnormal development trend, and predict the future equipment operation. This method can effectively identify potential faults, provide intelligent early warnings, and help operation and maintenance personnel take measures in advance to reduce the risk of equipment failure and improve the stability and reliability of the equipment. By continuously optimizing algorithms and updating data, the system can adapt to changes in the environment and equipment and continuously improve the accuracy and sensitivity of early warnings.

[0003] The current substation equipment status perception and early warning methods on the market mainly rely on a variety of sensors and monitoring technologies, combined with big data analysis and artificial intelligence algorithms to achieve real-time monitoring of equipment operating status and fault early warning. Common methods include anomaly detection based on sensor data such as vibration, temperature, current, and voltage, and data analysis and pattern recognition through machine learning, deep learning and other algorithms to detect potential equipment failures or performance degradation. Some advanced early warning systems also combine equipment health index, remaining service life prediction and other technologies to perform fault diagnosis and future trend prediction, thereby providing more accurate maintenance decision support. These methods can detect equipment failures at an early stage, optimize operation and maintenance strategies, improve the reliability and stability of substations, and reduce the probability of sudden accidents. Summary of the invention

[0004] In order to improve the existing substation equipment status perception and early warning method, an intelligent perception and early warning method for substation equipment status based on multi-source data is provided. By collecting real-time operation data and historical abnormal data, a time series reference template is constructed, and the dynamic time warping algorithm is used to compare and calculate the similarity, so as to judge whether the equipment operation status is abnormal and predict the future operation trend.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0006] The intelligent perception and early warning method of substation equipment status based on multi-source data is characterized by comprising:

[0007] Obtain real-time operation data and historical abnormal status data of substation equipment;

[0008] Based on all historical abnormal status data, obtain time series data to form a reference template;

[0009] Based on the dynamic time warping algorithm, the real-time operation data of substation equipment is compared with the reference data to calculate the similarity;

[0010] Based on the similarity results, determine whether the current operating status is abnormal. If abnormal, classify the abnormal type and analyze and predict future operating trends;

[0011] Upload new anomalies to the database and optimize the algorithm.

[0012] Preferably, the obtaining of real-time operation data and historical abnormal status data of substation equipment specifically includes:

[0013] Obtain real-time data on various indicators of the main operating equipment in the substation, including the oil temperature, winding temperature, gas composition in the oil, noise and vibration of the main transformer, the opening and closing operation performance of the circuit breaker, and contact resistance;

[0014] Based on the database, the operation data of the substation equipment under historical abnormal conditions is obtained.

[0015] Preferably, the acquiring of time series data based on all historical abnormal state data to form a reference template specifically includes:

[0016] Based on historical abnormal status data, determine the time granularity and standardize the format of all timestamps;

[0017] Sort the data based on timestamps to generate standardized sequences to form time series data;

[0018] Based on the time series data formed by all historical abnormal status data, it is summarized to obtain a reference template.

[0019] Preferably, based on the dynamic time warping algorithm, the real-time operation data of the substation equipment is compared with the reference data, and the similarity calculation specifically includes:

[0020] Based on the real-time data of various indicators of the main operating equipment in the substation, it is converted into time series data in the dynamic time warping algorithm;

[0021] Based on the acquired real-time time series X and the compared reference template time series Y, normalization operations on length and time axis are performed to align the two;

[0022] Based on the Euclidean distance formula, the distance between each pair of time points in the alignment process is calculated as the local distance, and the formula is:

[0023]

[0024] Among them, d(x i ,yj ) is the local distance, i and j are two time points;

[0025] Based on the cumulative distance matrix, the minimum cumulative distance required to reach each point in the time series is obtained. The update rule of the matrix is:

[0026] D[i,j]=d(x i ,y j )+min{D[i-1,j],D[i,j-1],D[i-1,j-1]}

[0027] Among them, D[i-1,j] means moving forward one time step along sequence X, D[i,j-1] means moving forward one time step along sequence Y, and D[i-1,j] means moving forward one time step along both sequences X and Y.

[0028] The optimal alignment path is obtained based on the back-tracking cumulative distance matrix to match the nonlinear relationship between the two time series;

[0029] Based on the calculation of the cumulative distance matrix D[m,n], the total distance between the two sequences is obtained, which is expressed as the similarity between the two.

[0030] Preferably, obtaining the optimal alignment path based on the backtracking cumulative distance matrix and matching the nonlinear relationship between the two time series specifically includes:

[0031] Based on the path from the matrix starting point (1,1) to the end point (n,m), each pair of coordinates on the path (i k ,j k ) indicates that the corresponding time points in sequence X and sequence Y are aligned.

[0032] The coordinates in the path satisfy i1≤i2≤...≤i k , and j1≤j≤...≤j k , the path cannot be reversed.

[0033] Each step of the path moves to the right, down, or diagonally from the current position.

[0034] Based on the alignment path, the dynamic time planning algorithm can bend the time axis of a sequence to align the two sequences and calculate their similarity more accurately.

[0035] Preferably, judging whether the current operating state is abnormal based on the similarity result, classifying the abnormal type if abnormal, and analyzing and predicting the future operating trend specifically include:

[0036] Based on the historical abnormal state data, a threshold τ is defined. When the matching distance is greater than the threshold, the current state of the device is considered abnormal, otherwise it is considered normal.

[0037] Based on the reference template generated by historical abnormal state data, different abnormal templates are divided into i , where each template is for a specific type of abnormal condition;

[0038] Based on the reference template of the historical abnormal state data matching the real-time time series X, calculate each template Y i The matching distance D(X,Y i ), select the template with the smallest matching distance as the prediction result, and obtain the abnormal state represented by different real-time time series;

[0039] Predict the future operation of the equipment based on the trends of the reference template.

[0040] Preferably, the uploading of the newly appeared anomaly to the database, the optimization algorithm specifically includes:

[0041] Upload new abnormal status data to the database, compare and analyze it with historical abnormal status data, and train to improve the matching efficiency of the algorithm.

[0042] Compared with the prior art, the advantages of the present invention are: it can realize real-time monitoring and dynamic prediction of equipment status. By collecting real-time operation data and historical abnormal data of substation equipment and combining big data analysis, the system can continuously learn and update the operation model of the equipment, and improve the accuracy and sensitivity of fault warning. The dynamic time warping algorithm is used to compare the real-time data of the equipment with the historical abnormal state template, so that the system can identify small abnormal changes and improve the accuracy of abnormal detection. It can classify and trend the operation status of the equipment. Once an abnormality is detected, the system can not only identify whether there is a fault, but also further judge the type of abnormality by analyzing the similarity between historical data and real-time data, and predict future operation trends. This provides more comprehensive and accurate fault diagnosis and prediction information for substation operation and maintenance personnel, so that preventive measures can be taken in advance to avoid sudden equipment failures and reduce downtime and maintenance costs. The method is based on time series data for processing, and can take into account the long-term change trend of equipment operation, so it has a strong early warning capability for long-term problems such as equipment aging and performance degradation. Through accurate fault warning and state prediction, the operation reliability and safety of the substation can be effectively improved, the operation and maintenance strategy can be optimized, and the equipment maintenance cost can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A flow chart of the device status perception steps of the method proposed in the present invention;

[0044] Figure 2 A flow chart of data acquisition of substation equipment according to the method proposed by the present invention;

[0045] Figure 3A flow chart of obtaining a time series template for the method proposed by the present invention;

[0046] Figure 4 A flow chart of a dynamic time warping algorithm of the method proposed in the present invention;

[0047] Figure 5 The most appropriate path calculation flow chart of the method proposed by the present invention;

[0048] Figure 6 This is a flow chart of abnormal state trend prediction of the method proposed by the present invention;

[0049] Figure 7 This is a schematic diagram of data uploading of the method proposed in the present invention. DETAILED DESCRIPTION

[0050] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0051] The intelligent perception and early warning method of substation equipment status based on multi-source data includes:

[0052] Step 1: Obtain real-time operation data and historical abnormal status data of substation equipment;

[0053] Step 2: Based on all historical abnormal status data, obtain time series data to form a reference template;

[0054] Step 3: Based on the dynamic time warping algorithm, the real-time operation data of the substation equipment is compared with the reference data to calculate the similarity;

[0055] Step 4: Based on the similarity results, determine whether the current operating status is abnormal. If abnormal, classify the abnormal type and analyze and predict future operating trends;

[0056] Step 5: Upload new anomalies to the database and optimize the algorithm.

[0057] See also Figure 2 As shown, obtaining real-time operation data and historical abnormal status data of substation equipment specifically includes:

[0058] Obtain real-time data on various indicators of the main operating equipment in the substation, including the oil temperature, winding temperature, gas composition in the oil, noise and vibration of the main transformer, the opening and closing operation performance of the circuit breaker, and contact resistance;

[0059] Based on the database, the operation data of the substation equipment under historical abnormal conditions is obtained.

[0060] See also Figure 3As shown, based on all historical abnormal status data, time series data is obtained to form a reference template, which specifically includes:

[0061] Based on historical abnormal status data, determine the time granularity and standardize the format of all timestamps;

[0062] Sort the data based on timestamps to generate standardized sequences to form time series data;

[0063] Based on the time series data formed by all historical abnormal status data, it is summarized to obtain a reference template.

[0064] Specifically, when processing historical abnormal status data, how to determine the appropriate time granularity (such as minutes, hours, days, etc.) is an important issue. If the granularity is too small, the data will be too detailed, cumbersome to process and may cause redundancy; if the granularity is too large, key information may be lost, resulting in poor detection results. Therefore, a dynamic granularity adjustment method can be adopted according to actual conditions. For example, a smaller granularity is used during time periods when abnormalities occur frequently, and a larger granularity is used during normal operation. This can more effectively capture fault characteristics while avoiding computing pressure caused by excessive data volume.

[0065] See also Figure 4 As shown in the figure, based on the dynamic time warping algorithm, the real-time operation data of the substation equipment is compared with the reference data, and the similarity calculation specifically includes:

[0066] Based on the real-time data of various indicators of the main operating equipment in the substation, it is converted into time series data in the dynamic time warping algorithm;

[0067] Based on the acquired real-time time series X and the compared reference template time series Y, normalization operations on length and time axis are performed to align the two;

[0068] Based on the Euclidean distance formula, the distance between each pair of time points in the alignment process is calculated as the local distance, and the formula is:

[0069]

[0070] Among them, d(x i ,y j ) is the local distance, i and j are two time points;

[0071] Based on the cumulative distance matrix, the minimum cumulative distance required to reach each point in the time series is obtained. The update rule of the matrix is:

[0072] D[i,j]=d(x i ,y j)+min{D[i-1,j],D[i,j-1],D[i-1,j-1]}

[0073] Among them, D[i-1,j] means moving forward one time step along sequence X, D[i,j-1] means moving forward one time step along sequence Y, and D[i-1,j] means moving forward one time step along both sequences X and Y.

[0074] The optimal alignment path is obtained based on the back-tracking cumulative distance matrix to match the nonlinear relationship between the two time series;

[0075] Based on the calculation of the cumulative distance matrix D[m,n], the total distance between the two sequences is obtained, which is expressed as the similarity between the two.

[0076] It is understandable that the core of the dynamic time warping algorithm is to align two time series, but since there may be differences in the time scale and length between the real-time data and the reference template, simple alignment may cause data misalignment, affecting subsequent distance calculations and similarity assessments. Therefore, before alignment, the time axes of the two time series need to be normalized so that they have the same time span. Interpolation technology can be used to fill the original data to ensure that each time point of the time series is in the same time interval. It is also possible to use adaptive time warping technology to adjust the time axis alignment strategy according to the characteristics of the data. For example, the local time alignment method is used to reduce the impact of time misalignment by adjusting the weight or sampling frequency of each time point.

[0077] See also Figure 5 As shown in the figure, the optimal alignment path is obtained based on the back-tracking cumulative distance matrix, and the nonlinear relationship between the two time series is matched, which specifically includes:

[0078] Based on the path from the matrix starting point (1,1) to the end point (n,m), each pair of coordinates on the path (i k ,j k ) indicates that the corresponding time points in sequence X and sequence Y are aligned.

[0079] The coordinates in the path satisfy i1≤i2≤...≤i k , and j1≤j≤...≤j k , the path cannot be reversed.

[0080] Each step of the path moves to the right, down, or diagonally from the current position.

[0081] Based on the alignment path, the dynamic time planning algorithm can bend the time axis of a sequence to align the two sequences and calculate their similarity more accurately.

[0082] Specifically, since the dynamic time planning algorithm needs to store the entire cumulative distance matrix D(i,j), the memory consumption and storage requirements may be very large when processing long time series. Dynamic programming can be used to save memory by only retaining the current row or current column instead of the entire matrix. For example, only the i-th row and the i+1-th row are stored, and they are continuously updated iteratively, so that the space complexity can be reduced from O(n×m) to O(min(n,m)). Divide the large time series into several segments, perform dynamic time planning alignment on each segment, and finally merge the results. This method is particularly suitable for very large-scale data sets.

[0083] See also Figure 6 As shown in the figure, based on the similarity results, it is determined whether the current operating status is abnormal. If abnormal, the abnormal type is classified, and the future operating trend is analyzed and predicted, including:

[0084] Based on the historical abnormal state data, a threshold τ is defined. When the matching distance is greater than the threshold, the current state of the device is considered abnormal, otherwise it is considered normal.

[0085] Based on the reference template generated by historical abnormal state data, different abnormal templates Y are divided i , where each template is for a specific type of abnormal condition;

[0086] Based on the reference template of the historical abnormal state data matching the real-time time series X, calculate each template Y i The matching distance D(X,Y i ), select the template with the smallest matching distance as the prediction result, and obtain the abnormal state represented by different real-time time series;

[0087] Predict the future operation of the equipment based on the trends of the reference template.

[0088] Specifically, if the historical template shows that the device has experienced a trend of temperature increase, load increase, etc. under certain conditions, then during the real-time data matching process, it can also be inferred that the device may experience similar changes in the future. For example, if the real-time data has the highest matching degree with the "overload failure" template, it means that the current load of the device is close to the limit, and there is a high risk of overload failure in the future. Therefore, we can predict that the device may need to increase the load or take cooling measures in the future, and prepare for maintenance in advance.

[0089] Reference Figure 7 As shown, the newly appeared anomalies are uploaded to the database. The optimization algorithm specifically includes:

[0090] Upload new abnormal status data to the database, compare and analyze it with historical abnormal status data, and train to improve the matching efficiency of the algorithm.

[0091] In summary, the advantages of the present invention are: by combining real-time operation data with historical abnormal status data, it can fully reflect the working status of the equipment and enhance the accuracy and real-time performance of the early warning system. By extracting historical abnormal status data to generate a time series reference template, a reliable comparison basis can be provided for subsequent abnormality detection to ensure the accuracy of the judgment. Secondly, the dynamic time warping algorithm is used for data comparison to improve the adaptability and accuracy to complex operating modes. Based on similarity-based abnormal judgment, potential faults can be discovered in a timely manner, human intervention can be reduced, and the automation and intelligence level of the early warning system can be improved. At the same time, the classification of abnormal types and trend prediction functions provide operation and maintenance personnel with clear troubleshooting directions and future operation and maintenance strategies, and optimize equipment management. Overall, this method not only improves the equipment management efficiency of substations and reduces the risk of failures through intelligent perception and precise early warning, but also contributes to the long-term safe operation of equipment.

[0092] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for intelligent perception and early warning of substation equipment status based on multi-source data, characterized in that: include: Obtain real-time operation data and historical abnormal status data of substation equipment; Based on all historical abnormal status data, obtain time series data to form a reference template; Based on the dynamic time warping algorithm, the real-time operation data of substation equipment is compared with the reference data to calculate the similarity; Based on the similarity results, determine whether the current operating status is abnormal. If abnormal, classify the abnormal type and analyze and predict future operating trends; Upload new anomalies to the database and optimize the algorithm.

2. The method for intelligent perception and early warning of substation equipment status based on multi-source data according to claim 1 is characterized in that: The acquisition of real-time operation data and historical abnormal status data of substation equipment specifically includes: Obtain real-time data on various indicators of the main operating equipment in the substation, including the oil temperature, winding temperature, gas composition in the oil, noise and vibration of the main transformer, the opening and closing operation performance of the circuit breaker, and contact resistance; Based on the database, the operation data of the substation equipment under historical abnormal conditions is obtained.

3. The method for intelligent perception and early warning of substation equipment status based on multi-source data according to claim 1 is characterized in that: The method of obtaining time series data based on all historical abnormal status data to form a reference template specifically includes: Based on historical abnormal status data, determine the time granularity and standardize the format of all timestamps; Sort the data based on timestamps to generate standardized sequences to form time series data; Based on the time series data formed by all historical abnormal status data, it is summarized to obtain a reference template.

4. The method for intelligent perception and early warning of substation equipment status based on multi-source data according to claim 1 is characterized in that: Based on the dynamic time warping algorithm, the real-time operation data of the substation equipment is compared with the reference data, and the similarity calculation specifically includes: Based on the real-time data of various indicators of the main operating equipment in the substation, it is converted into time series data in the dynamic time warping algorithm; Based on the acquired real-time time series X and the compared reference template time series Y, normalization operations on length and time axis are performed to align the two; Based on the Euclidean distance formula, the distance between each pair of time points in the alignment process is calculated as the local distance, and the formula is: Among them, d(x i ,y j ) is the local distance, i and j are two time points; Based on the cumulative distance matrix, the minimum cumulative distance required to reach each point in the time series is obtained. The update rule of the matrix is: D[i,j]=d(x i ,y j )+min{D[i-1,j],D[i,j-1],D[i-1,j-1]} Among them, D[i-1,j] means moving forward one time step along sequence X, D[i,j-1] means moving forward one time step along sequence Y, and D[i-1,j] means moving forward one time step along both sequences X and Y. The optimal alignment path is obtained based on the back-tracking cumulative distance matrix to match the nonlinear relationship between the two time series; Based on the calculation of the cumulative distance matrix D[m,n], the total distance between the two sequences is obtained, which is expressed as the similarity between the two.

5. The method for intelligent perception and early warning of substation equipment status based on multi-source data according to claim 1 is characterized in that: The method of obtaining the optimal alignment path based on the backtracking cumulative distance matrix and matching the nonlinear relationship between the two time series specifically includes: Based on the path from the matrix starting point (1,1) to the end point (n,m), each pair of coordinates on the path (i k ,j k ) indicates that the corresponding time points in sequence X and sequence Y are aligned. The coordinates in the path satisfy i1≤i2≤...≤i k , and j1≤j≤...≤j k , the path cannot be reversed. Each step of the path moves to the right, down, or diagonally from the current position. Based on the alignment path, the dynamic time planning algorithm can bend the time axis of a sequence to align the two sequences and calculate their similarity more accurately.

6. The method for intelligent perception and early warning of substation equipment status based on multi-source data according to claim 1 is characterized in that: Based on the similarity results, the current operating state is judged to be abnormal. If abnormal, the abnormal type is classified, and the future operating trend is analyzed and predicted. Specifically, the following are included: Based on the historical abnormal state data, a threshold τ is defined. When the matching distance is greater than the threshold, the current state of the device is considered abnormal, otherwise it is considered normal. Based on the reference template generated by historical abnormal state data, different abnormal templates Y are divided i , where each template is for a specific type of abnormal condition; Based on the reference template of the historical abnormal state data matching the real-time time series X, calculate each template Y i The matching distance D(X,Y i ), select the template with the smallest matching distance as the prediction result, and obtain the abnormal state represented by different real-time time series; Predict the future operation of the equipment based on the trends of the reference template.

7. The method for intelligent perception and early warning of substation equipment status based on multi-source data according to claim 1 is characterized in that: The optimization algorithm for uploading the newly appeared anomaly to the database specifically includes: Upload new abnormal status data to the database, compare and analyze it with historical abnormal status data, and train to improve the matching efficiency of the algorithm.

Images