A method and system for monitoring power consumption at transformer end

Abstract

The present invention relates to the field of transformer energy consumption monitoring, and in particular to a method and system for monitoring transformer-end power consumption. The method comprises the following steps: obtaining real-time transformer operation monitoring parameters; performing time-series window segmentation on the real-time transformer operation monitoring parameters and performing time-series load characteristic calculation to obtain transformer load characteristics for multiple time periods; constructing a time-series load characteristic curve based on the transformer load characteristics for multiple time periods; performing load distribution trend evolution on the time-series load characteristic curve to obtain load distribution change trend data; performing deep dynamic trend mining on the real-time transformer operation monitoring parameters based on the load distribution change trend data to obtain a dynamic trend representation of power loss; obtaining transformer temperature monitoring data and a transformer surface image; performing multi-dimensional uniform interpolation on the transformer surface image based on the transformer temperature monitoring data to construct a three-dimensional temperature distribution trend field. The present invention achieves efficient and accurate energy consumption monitoring.

Description

Technical Field

[0001] The present invention relates to the field of transformer energy consumption monitoring, and in particular to a transformer-end power consumption monitoring method and system. Background Art

[0002] With the continuous improvement of the automation and intelligence level of industrial production, the role of transformers in power systems is becoming more and more important. Transformers are key energy conversion equipment in power systems, and their operating status directly affects the stability and energy utilization efficiency of the entire power grid. However, in long-term continuous operation, the power consumption of transformers may be affected by various internal and external factors, such as load fluctuations, winding aging, cooling system failures, etc. These factors will cause abnormal energy consumption of transformers, thereby affecting their operating efficiency and economy. Traditional transformer power consumption monitoring methods mainly rely on manual inspections and regular testing, which have problems such as low detection accuracy and low analysis efficiency. Therefore, an intelligent transformer energy consumption monitoring method is needed. Summary of the Invention

[0003] In order to solve the above technical problems, the present invention proposes a transformer-end power consumption monitoring method and system to solve at least one of the above technical problems.

[0004] To achieve the above object, the present invention provides a method for monitoring power consumption at a transformer end, comprising the following steps:

[0005] Step S1: obtaining transformer real-time operation monitoring parameters; performing time series window segmentation on the transformer real-time operation monitoring parameters, and performing time series load characteristic calculation to obtain transformer load characteristics for multiple time periods;

[0006] Step S2: constructing a time-series load characteristic curve based on the transformer load characteristics of multiple time periods; performing load distribution trend evolution on the time-series load characteristic curve to obtain load distribution change trend data;

[0007] Step S3: performing deep dynamic trend mining on the transformer real-time operation monitoring parameters based on the load distribution change trend data, thereby obtaining a dynamic trend representation of the power loss;

[0008] Step S4: Obtain transformer temperature monitoring data and a transformer surface image; perform multi-dimensional uniform interpolation on the transformer surface image according to the transformer temperature monitoring data, and reconstruct a three-dimensional temperature distribution to construct a three-dimensional temperature distribution trend field;

[0009] Step S5: performing iterative correlation evolution learning on the three-dimensional temperature distribution trend field based on the dynamic trend representation of the power loss, and constructing a power-temperature situation map;

[0010] Step S6: Based on the power-temperature situation map, locate the abnormal energy consumption nodes of the transformer, perform intelligent load parameter fine-tuning, and build an intelligent load fine-tuning engine.

[0011] The present invention helps to capture the dynamic changes of the real-time operating status of the transformer through time series window segmentation and feature calculation, and provides basic data for subsequent analysis. The transformer load characteristics of multiple time periods can reflect the changes in the load in different time periods, and provide detailed information for energy consumption monitoring. By constructing a time series load characteristic curve and load distribution trend evolution, the dynamic changes of the load distribution can be intuitively understood, providing a basis for energy consumption trend analysis. Obtaining load distribution change trend data helps to discover the regularity of load changes and provides support for subsequent dynamic trend mining. Through deep dynamic trend mining, potential energy consumption trend laws can be discovered, which helps to predict future electricity consumption. Obtaining the dynamic trend characterization of loss power can more accurately describe the changes in transformer power loss, and provide more refined data for energy consumption analysis. Multi-dimensional uniform interpolation and three-dimensional temperature distribution reconstruction can accurately reflect the distribution of transformer surface temperature and provide visualization support for temperature monitoring. Constructing a three-dimensional temperature distribution trend field helps track temperature change trends and provide important information for energy consumption and safety analysis. Constructing a power-temperature situation map through iterative correlation evolution learning can comprehensively consider the relationship between power and temperature and help understand the linkage law between the two. Based on the map, the situation changes of power and temperature can be more intuitively displayed, which helps to discover abnormal energy consumption nodes and conduct further analysis. By locating abnormal energy consumption nodes and fine-tuning intelligent load parameters, energy consumption anomalies can be found in time and load strategies can be adjusted to effectively reduce energy consumption costs. Building an intelligent load fine-tuning engine can automatically optimize load configuration, improve power efficiency and extend equipment life.

[0012] Preferably, step S1 includes the following steps:

[0013] Step S11: continuously monitoring the normal operation status of the transformer based on multiple sensors and extracting real-time operation monitoring parameters of the transformer;

[0014] Step S12: performing multi-scale adaptive filtering on the transformer real-time operation monitoring parameters to generate filtered optimized monitoring parameters;

[0015] Step S13: performing time series window segmentation on the filter optimization monitoring parameters to obtain monitoring parameters for multiple time periods;

[0016] Step S14: performing time series load characteristic calculation on the monitoring parameters of the multiple time periods to obtain transformer load characteristics of the multiple time periods.

[0017] The present invention can fully understand the operating status of the transformer through continuous monitoring by multiple sensors, and the extracted real-time monitoring parameters contain rich information. Data extraction based on multiple sensors helps to accurately capture multiple aspects of the transformer operating status and provide sufficient support for subsequent data processing. Multi-scale adaptive filtering can effectively remove noise and interference in the monitoring parameters, improve the accuracy and reliability of the data, and generate filtered optimized monitoring parameters to help improve data quality and provide a more reliable basis for subsequent analysis and decision-making. Time series window segmentation can divide the monitoring parameters by time, making the data easier to process and analyze, while retaining the time series information of the data. Obtaining monitoring parameters for multiple time periods helps to have a more detailed understanding of the changes in the transformer operating status in different time periods, and provides more dimensions for subsequent analysis. By calculating the time series load characteristics of the monitoring parameters for multiple time periods, the load situation of the transformer can be more comprehensively understood, and the regularity of load changes can be revealed. Obtaining transformer load characteristics for multiple time periods can provide more detailed and accurate data support for energy consumption monitoring and analysis, and help discover potential problems and optimize operation.

[0018] Preferably, the specific steps of step S12 are:

[0019] Calculate the mean value of the transformer's real-time operation monitoring parameters to obtain the normalized range of the monitoring parameters;

[0020] Based on the normalized range of monitoring parameters, the abnormal outliers of the real-time operation monitoring parameters of the transformer are eliminated to obtain the abnormal value elimination parameters;

[0021] Perform multi-scale wavelet transform decomposition on the outlier removal parameters to extract detailed feature information in different frequency domain bands;

[0022] Discrete frequency components are calculated for the detailed feature information of different frequency domain bands to obtain the frequency component value of each frequency band;

[0023] Adaptively optimize the filter coefficients of the detail feature information of different frequency bands based on the frequency component value of each frequency band to generate multiple frequency band filter data;

[0024] The filter data of multiple frequency bands are reconstructed by nonlinear transformation to generate filter optimization monitoring parameters.

[0025] The present invention can help determine the monitoring parameter range under normal operating conditions by calculating parameter means and establishing normalized ranges, providing benchmark reference values. Establishing normalized ranges helps identify abnormal values ​​in subsequent monitoring parameters, laying the foundation for abnormality detection and processing. Eliminating abnormal outliers can improve data quality and accuracy, ensuring the effectiveness of subsequent analysis and processing. Through outlier elimination processing, the interference of abnormal data on subsequent feature extraction and analysis can be reduced, making the data more reliable. Wavelet transform can effectively extract the time-frequency characteristics of signals, help capture signal change information at different scales, and extract detailed feature information of different frequency domain bands to more comprehensively describe the changing laws of monitoring parameters, provide more details for subsequent analysis, and calculate frequency component values. It can help understand the energy distribution of monitoring parameters in different frequency bands and reveal potential frequency domain characteristics. The extraction of frequency component values ​​can help to more finely analyze the frequency domain characteristics of monitoring parameters and provide important information for subsequent processing. Through adaptive filter coefficient optimization, the filter parameters can be adjusted according to the characteristics of the frequency band to improve the filtering effect. Generating filter data in multiple frequency bands can help to more clearly display the characteristics of monitoring parameters in different frequency domains and provide more dimensional information for subsequent analysis. Through nonlinear transformation reconstruction, the expression form of monitoring parameters can be further optimized to highlight the characteristic information of the data. Generating filter-optimized monitoring parameters can help reduce noise interference, highlight the potential characteristics of monitoring parameters, and provide a more accurate data basis for subsequent analysis and application.

[0026] Preferably, the specific steps of step S2 are:

[0027] Step S21: performing load serialization fitting on the transformer load characteristics of multiple time periods to construct a time series load characteristic curve;

[0028] Step S22: performing load change analysis on the time series load characteristic curve to obtain time series load change data;

[0029] Step S23: performing load parameter distribution identification on the transformer load characteristics in multiple time periods to generate load distribution data for different time periods;

[0030] Step S24: performing load distribution trend evolution on the time series load change data based on the load distribution data of different time periods, thereby obtaining load distribution change trend data.

[0031] By serializing and fitting the transformer load characteristics of multiple time periods, the present invention can better understand the regularity of load changes and provide a more accurate data basis for subsequent analysis. Constructing a time-series load characteristic curve helps to intuitively display the changing trend of load characteristics over time. By performing load change analysis on the time-series load characteristic curve, we can deeply understand the fluctuation of load in different time periods, providing a basis for abnormal detection and optimization. Obtaining time-series load change data helps to discover periodic or sudden changes in load, and helps predict future load change trends. Identifying the load parameter distribution in different time periods can reveal the characteristics and changing laws of load in different time periods, and provide important information for load management and prediction. Generating load distribution data for different time periods helps to establish a comprehensive load characteristic archive. By performing trend evolution analysis on load distribution data, we can understand the evolution of load distribution over time, and provide a basis for formulating long-term energy consumption management strategies. Obtaining load distribution change trend data helps to discover the overall changing trend of load distribution, and provides support for adjusting load configuration and predicting energy consumption changes.

[0032] Preferably, the specific steps of step S3 are:

[0033] Step S31: extracting transformer real-time current and voltage data based on transformer real-time operation monitoring parameters;

[0034] Step S32: performing multi-time point power calculation on the transformer load characteristics of multiple time periods according to the real-time current and voltage data of the transformer to generate load power parameters of the multiple time periods;

[0035] Step S33: performing reactive power calculation on the load power parameters of multiple time periods to generate load reactive power parameters for each time period;

[0036] Step S34: performing actual loss calculation on the load power parameters of multiple time periods and the load reactive power parameters of each time period, thereby obtaining load loss power data of different time periods;

[0037] Step S35: performing deep dynamic trend mining on the load loss power data of different time periods according to the load distribution change trend data, thereby obtaining a dynamic trend representation of the loss power.

[0038] By extracting real-time current and voltage data of the transformer, the present invention can monitor the operating status of the power system in real time and provide basic data for subsequent energy consumption analysis. Real-time data extraction helps to timely discover abnormal conditions of the power system and improve the reliability and safety of equipment operation. Using real-time current and voltage data to perform load power calculations in multiple time periods can accurately reflect the energy consumption level of the transformer under different load conditions. Generating load power parameters for multiple time periods helps to understand the impact of load changes on energy consumption and provide detailed data support for energy consumption assessment. Performing reactive power calculation on load power parameters for multiple time periods can evaluate the reactive power demand of the system and help optimize the power factor of the power system. Generating load reactive power parameters for each time period helps to analyze the power factor of the load and provide a basis for energy saving and power quality improvement. Through actual loss calculation, the load loss power in different time periods can be accurately assessed, helping to identify the causes of energy consumption fluctuations and potential energy-saving opportunities. Obtaining load loss power data helps to quantify energy consumption losses and provide support for optimizing energy utilization and cost control. Deep dynamic trend mining of loss power data can discover potential patterns and changing trends of load loss power and help to provide early warning of energy consumption anomalies.

[0039] Preferably, the specific steps of step S4 are:

[0040] Step S41: Obtain transformer temperature monitoring data and transformer surface image;

[0041] Step S42: performing three-dimensional meshing on the transformer surface image to obtain a transformer surface network;

[0042] Step S43: extracting multiple grid points based on the transformer surface network;

[0043] Step S44: performing multi-dimensional uniform interpolation on multiple grid points according to the transformer temperature monitoring data, thereby obtaining a transformer temperature network;

[0044] Step S45: performing a three-dimensional temperature distribution analysis on the transformer temperature network to obtain three-dimensional grid temperature distribution data;

[0045] Step S46: performing multi-time temperature change fluctuation analysis on the three-dimensional grid temperature distribution data to generate three-dimensional temperature change fluctuation data;

[0046] Step S47: reconstructing the three-dimensional temperature distribution of the three-dimensional temperature change fluctuation data to construct a three-dimensional temperature distribution trend field.

[0047] By acquiring transformer temperature monitoring data and surface images, the present invention can fully understand the operating status and thermal load of the transformer. The temperature monitoring data and surface images provide real-time temperature information, which helps monitor the heat distribution and thermal changes of the transformer. Performing three-dimensional meshing on the transformer surface image helps to more precisely describe the structure and characteristics of the transformer surface. The three-dimensional meshing provides a basic mesh structure, providing a data foundation for subsequent temperature analysis and interpolation. By extracting multiple grid points, representative locations can be selected for temperature monitoring and analysis, accurately reflecting the temperature distribution of the entire transformer surface. Multi-dimensional uniform interpolation using the transformer temperature monitoring data can fill in missing data and obtain a more continuous and accurate temperature network. A comprehensive transformer temperature network can be obtained through interpolation, providing data support for subsequent temperature distribution analysis. Performing three-dimensional temperature distribution analysis on the transformer temperature network can provide in-depth understanding of the temperature distribution pattern and heat transfer inside the transformer. The generated three-dimensional mesh temperature distribution data helps to evaluate the thermal characteristics of the transformer and provide a basis for temperature control and equipment protection. Multi-time temperature change fluctuation analysis and three-dimensional temperature distribution reconstruction can reveal the dynamic change pattern and trend of the transformer temperature over time. Constructing a three-dimensional temperature distribution trend field helps predict temperature change trends, promptly identify potential fault risks, and ensure the safe and stable operation of the transformer.

[0048] Preferably, the specific steps of step S5 are:

[0049] Step S51: mining implicit correlation features of the three-dimensional temperature distribution trend field based on the power loss dynamic trend characterization to obtain power loss-temperature distribution correlation data;

[0050] Step S52: Discretize the time series temperature characteristics of the three-dimensional temperature distribution trend field and extract the three-dimensional temperature distribution vector;

[0051] Step S53: performing feature space mapping processing on the dynamic trend representation of power loss and the three-dimensional temperature distribution vector to construct a potential feature space model;

[0052] Step S54: performing iterative correlation evolution learning on the potential feature space model according to the loss power-temperature distribution correlation data to construct a power-temperature situation map.

[0053] The present invention can deeply explore the potential correlation features between the dynamic trend of power loss and the three-dimensional temperature distribution through implicit correlation feature mining, and obtain the power loss-temperature distribution correlation data, which is helpful for understanding the relationship between power loss and temperature distribution. By discretizing the three-dimensional temperature distribution and extracting the three-dimensional temperature distribution vector, the complex temperature data can be converted into an easy-to-process feature vector. The three-dimensional temperature distribution vector extraction helps to more effectively describe and analyze the temperature characteristics, and provides a data basis for subsequent feature space mapping processing. Through feature space mapping processing, the power loss dynamic trend characterization and the three-dimensional temperature distribution vector can be mapped to the potential feature space. Constructing a potential feature space model helps to convert the correlation relationship between different features into a visual and analytical form, and helps to understand the complex relationship between power and temperature. Through iterative correlation evolution learning, a power-temperature situation map can be constructed based on the power loss-temperature distribution correlation data. Constructing a power-temperature situation map helps to reveal the complex relationship and change law between power and temperature, and provides a deeper understanding and analysis for transformer-end power consumption monitoring.

[0054] Preferably, the specific steps of step S6 are:

[0055] Step S61: monitoring abnormal energy consumption parameter deviations of the transformer based on the power-temperature situation map, and marking abnormal energy consumption deviation parameters;

[0056] Step S62: locating abnormal nodes on the transformer surface network according to the abnormal energy consumption deviation parameter, and extracting abnormal energy consumption grid points;

[0057] Step S63: performing abnormal type attribution identification on the abnormal energy consumption grid point to obtain abnormal type attribution data;

[0058] Step S64: Perform intelligent load parameter fine-tuning based on the abnormal type attribution data and build an intelligent load fine-tuning engine.

[0059] The present invention can monitor the abnormal energy consumption parameter deviation of the transformer through the power-temperature situation map, timely identify potential energy consumption anomalies, mark the abnormal energy consumption deviation parameters to help accurately capture the abnormal energy consumption of the transformer, and provide important information for subsequent abnormal problem diagnosis and processing. According to the abnormal energy consumption deviation parameters, the abnormal nodes of the transformer surface network are located, the abnormal energy consumption grid points are extracted, and the energy consumption anomaly points are accurately located. The extraction of abnormal energy consumption grid points helps to focus on the specific location where the anomaly exists, and provides positioning support for further abnormal type attribution identification and processing. Abnormal type attribution identification of abnormal energy consumption grid points can analyze the root cause of the anomaly and help understand the specific type and characteristics of the abnormal situation. Obtaining abnormal type attribution data helps to deeply understand the abnormal situation and provides a basis for identifying anomalies for subsequent intelligent load parameter fine-tuning. Based on the abnormal type attribution data, an intelligent load parameter fine-tuning engine can be constructed to realize intelligent fine-tuning and optimization of transformer load parameters. The intelligent load fine-tuning engine helps to automatically adjust load parameters according to abnormal situations, improve energy utilization efficiency, reduce energy consumption losses, and ensure equipment operation stability.

[0060] In this specification, a transformer-end power consumption monitoring system is provided, which is used to execute the transformer-end power consumption monitoring method described above, including:

[0061] The time series segmentation module is used to obtain the real-time operation monitoring parameters of the transformer; the real-time operation monitoring parameters of the transformer are segmented into time series windows, and the time series load characteristics are calculated to obtain the transformer load characteristics of multiple time periods;

[0062] The load distribution module is used to construct a time-series load characteristic curve based on the transformer load characteristics of multiple time periods; the load distribution trend evolution of the time-series load characteristic curve is performed to obtain load distribution change trend data;

[0063] Dynamic trend mining module, used to conduct deep dynamic trend mining on transformer real-time operation monitoring parameters based on load distribution change trend data, thereby obtaining dynamic trend representation of power loss;

[0064] The three-dimensional temperature distribution module is used to obtain transformer temperature monitoring data and transformer surface images; based on the transformer temperature monitoring data, the transformer surface image is subjected to multi-dimensional uniform interpolation, and the three-dimensional temperature distribution is reconstructed to construct a three-dimensional temperature distribution trend field;

[0065] The situation map module is used to perform iterative correlation evolution learning on the three-dimensional temperature distribution trend field based on the dynamic trend representation of power loss, and construct a power-temperature situation map;

[0066] The parameter fine-tuning module is used to locate abnormal energy consumption nodes of the transformer based on the power-temperature situation map, perform intelligent load parameter fine-tuning, and build an intelligent load fine-tuning engine.

[0067] The present invention helps to capture the time series characteristics of the transformer operating status by obtaining real-time monitoring parameters and performing time series window segmentation, providing a real-time data basis. The time series load characteristic calculation can help analyze the load characteristics of different time periods, and provide key data for subsequent load distribution analysis and dynamic trend mining. Constructing a time series load characteristic curve based on the load characteristics of multiple time periods helps to reveal the changing trend of load characteristics over time, and provides visualization support for the evolution of load distribution trends. Analyzing the load distribution change trend data helps to understand the laws and trends of load changes, and provides important clues for subsequent dynamic trend mining. Through deep dynamic trend mining, the dynamic trends of the transformer real-time monitoring parameters can be revealed according to the load distribution change trend data, including the dynamic characterization of loss power. Obtaining the dynamic trend characterization of loss power helps to understand the laws and trends of power changes, and provides in-depth information for energy consumption monitoring and analysis. Insight: Through multi-dimensional uniform interpolation and three-dimensional temperature distribution reconstruction, an accurate three-dimensional temperature distribution trend field can be constructed to reveal the spatial distribution characteristics of the transformer surface temperature. Constructing a three-dimensional temperature distribution trend field helps to understand the dynamic changes of temperature distribution and provides a basis for power-temperature correlation analysis and situation map construction. Through iterative correlation evolution learning, a power-temperature situation map is constructed to reveal the correlation between power and temperature, providing a comprehensive perspective for energy consumption monitoring and anomaly detection. Constructing a power-temperature situation map helps to fully understand the dynamic situation of power consumption at the transformer end and provide a basis for subsequent parameter fine-tuning. Through abnormal energy consumption node positioning and intelligent load parameter fine-tuning, intelligent adjustment of the transformer can be achieved, energy utilization efficiency can be improved, and energy consumption losses can be reduced. Building an intelligent load fine-tuning engine helps to adjust load parameters in real time, ensure safe and stable operation of equipment, and improve equipment life and performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 A schematic flow chart of the steps of a method for monitoring power consumption at a transformer end according to the present invention;

[0069] Figure 2 Detailed implementation flow chart of step S1;

[0070] Figure 3 Detailed implementation flow chart of step S2;

[0071] Figure 4 Schematic diagram of the detailed implementation steps of step S3. DETAILED DESCRIPTION

[0072] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0073] This application provides a method and system. The execution entities of the method and system include, but are not limited to, mechanical equipment, data processing platforms, cloud server nodes, network upload devices, etc. that are equipped with the system, which can be regarded as general computing nodes of this application. The data processing platform includes, but is not limited to, at least one of an audio and image management system, an information management system, and a cloud data management system.

[0074] See also Figures 1 to 4 The present invention provides a method for monitoring power consumption at a transformer end, the method comprising the following steps:

[0075] Step S1: obtaining transformer real-time operation monitoring parameters; performing time series window segmentation on the transformer real-time operation monitoring parameters, and performing time series load characteristic calculation to obtain transformer load characteristics for multiple time periods;

[0076] Step S2: constructing a time-series load characteristic curve based on the transformer load characteristics of multiple time periods; performing load distribution trend evolution on the time-series load characteristic curve to obtain load distribution change trend data;

[0077] Step S3: performing deep dynamic trend mining on the transformer real-time operation monitoring parameters based on the load distribution change trend data, thereby obtaining a dynamic trend representation of the power loss;

[0078] Step S4: Obtain transformer temperature monitoring data and a transformer surface image; perform multi-dimensional uniform interpolation on the transformer surface image according to the transformer temperature monitoring data, and reconstruct a three-dimensional temperature distribution to construct a three-dimensional temperature distribution trend field;

[0079] Step S5: performing iterative correlation evolution learning on the three-dimensional temperature distribution trend field based on the dynamic trend representation of the power loss, and constructing a power-temperature situation map;

[0080] Step S6: Based on the power-temperature situation map, locate the abnormal energy consumption nodes of the transformer, perform intelligent load parameter fine-tuning, and build an intelligent load fine-tuning engine.

[0081] The present invention helps to capture the dynamic changes of the real-time operating status of the transformer through time series window segmentation and feature calculation, and provides basic data for subsequent analysis. The transformer load characteristics of multiple time periods can reflect the changes in the load in different time periods, and provide detailed information for energy consumption monitoring. By constructing a time series load characteristic curve and load distribution trend evolution, the dynamic changes of the load distribution can be intuitively understood, providing a basis for energy consumption trend analysis. Obtaining load distribution change trend data helps to discover the regularity of load changes and provides support for subsequent dynamic trend mining. Through deep dynamic trend mining, potential energy consumption trend laws can be discovered, which helps to predict future electricity consumption. Obtaining the dynamic trend characterization of loss power can more accurately describe the changes in transformer power loss, and provide more refined data for energy consumption analysis. Multi-dimensional uniform interpolation and three-dimensional temperature distribution reconstruction can accurately reflect the distribution of transformer surface temperature and provide visualization support for temperature monitoring. Constructing a three-dimensional temperature distribution trend field helps track temperature change trends and provide important information for energy consumption and safety analysis. Constructing a power-temperature situation map through iterative correlation evolution learning can comprehensively consider the relationship between power and temperature and help understand the linkage law between the two. Based on the map, the situation changes of power and temperature can be more intuitively displayed, which is helpful to discover abnormal energy consumption nodes and conduct further analysis. By locating abnormal energy consumption nodes and fine-tuning intelligent load parameters, energy consumption anomalies can be found in time and load strategies can be adjusted to effectively reduce energy consumption costs. Building an intelligent load fine-tuning engine can automatically optimize load configuration, improve power efficiency and extend equipment life.

[0082] In the embodiment of the present invention, see Figure 1 , is a schematic flow chart of the steps of a method for monitoring power consumption at the transformer end of the present invention. In this example, the steps of the method for monitoring power consumption at the transformer end include:

[0083] Step S1: obtaining transformer real-time operation monitoring parameters; performing time series window segmentation on the transformer real-time operation monitoring parameters, and performing time series load characteristic calculation to obtain transformer load characteristics for multiple time periods;

[0084] In this embodiment, sensors (such as current sensors, temperature sensors, and voltage sensors) are installed at key locations of the transformer. The sampling frequency of the sensors is configured to ensure real-time monitoring of the transformer's operating status. Sensor data is regularly acquired through a data acquisition system (such as a PLC or SCADA system) to ensure real-time and accurate data acquisition. Monitoring parameters (such as current, temperature, and voltage) are recorded at each time point. The size of the time series window (such as hourly or minutely) and the overlap step size (such as 0 or 50%) are determined. These parameters affect the accuracy and temporal resolution of the analysis. A sliding window technique is used to segment the real-time monitoring parameters, dividing the continuous time series data into multiple time series windows. Each window contains data at a certain number of time points for load characteristic calculation. The load characteristic indicators to be calculated, such as average load, maximum load, load factor, and power factor, are determined. Load characteristic calculation is performed on the data in each time series window to generate a load characteristic value for each window. The load characteristic data for each time period is recorded to form a complete data set for subsequent analysis.

[0085] Step S2: constructing a time-series load characteristic curve based on the transformer load characteristics of multiple time periods; performing load distribution trend evolution on the time-series load characteristic curve to obtain load distribution change trend data;

[0086] In this embodiment, the load characteristic data (such as average load, maximum load, etc.) of multiple time periods are sorted into a table to ensure that the data of each time period is clearly visible and the load characteristics are arranged in chronological order. A suitable drawing tool or library (such as Matplotlib, Seaborn or Excel) is selected to draw a time series load characteristic curve. The time series load characteristic curve is drawn according to the sorted load characteristic data. Usually, the X-axis is time and the Y-axis is the load characteristic value. The drawn curve is labeled and a legend, title and coordinate axis labels are added to improve readability. The load distribution characteristics that need to be analyzed are determined, such as mean, variance, skewness, kurtosis, etc. These characteristics can reflect the load change trend. Appropriate statistical analysis methods (such as moving average, trend line regression analysis, time series analysis, etc.) are used to analyze changes in load distribution. Based on the data characteristics, an appropriate model (such as linear regression, exponential smoothing, or ARIMA model) is selected for trend analysis. The time series load characteristic curve is analyzed to identify the evolution trend of the load distribution. The selected model is used to fit the data, extract trend information, and calculate the rate of change of load characteristics over time, such as growth rate or fluctuation amplitude, to reflect the dynamic changes in load distribution. The analyzed load distribution change trend data is organized into tables or charts, recording the changes in characteristics such as mean and variance over time. The load distribution change trend data is plotted as a curve to facilitate the intuitive display of the evolution process of the load distribution.

[0087] Step S3: performing deep dynamic trend mining on the transformer real-time operation monitoring parameters based on the load distribution change trend data, thereby obtaining a dynamic trend representation of the power loss;

[0088] In this embodiment, a suitable dynamic trend mining algorithm is selected based on data characteristics, such as time series analysis (ARIMA, SARIMA), machine learning methods (LSTM, random forest), or deep learning models. The dynamic correlation characteristics between the two are analyzed to identify monitoring parameters closely related to transformer power loss, such as load current, winding temperature, and other indicators. These indicators are used as the main features for subsequent power loss dynamic trend mining. In combination with the physical characteristics and loss mechanism of the transformer, a mathematical model for calculating transformer power loss is established to provide a computational basis for subsequent dynamic trend analysis. The selected model is trained using load distribution change trend data and extracted real-time monitoring parameter features. Training and test sets are divided to ensure the generalization ability of the model. The model is evaluated using appropriate evaluation indicators (such as mean square error (MSE) and root mean square error (RMSE)) to ensure the model's prediction accuracy. The trained model is used to predict real-time monitoring parameters, identify the dynamic trend of power loss, perform time series prediction of power loss, generate power loss dynamic trend data for a period of time in the future, calculate the rate of change of power loss, identify mutation points and abnormal fluctuations, and provide in-depth analysis.

[0089] Step S4: Obtain transformer temperature monitoring data and a transformer surface image; perform multi-dimensional uniform interpolation on the transformer surface image according to the transformer temperature monitoring data, and reconstruct a three-dimensional temperature distribution to construct a three-dimensional temperature distribution trend field;

[0090] In this embodiment, real-time temperature data is regularly acquired from the temperature sensor device of the transformer to ensure that the temperature information at different locations is accurately recorded. The data includes but is not limited to the oil temperature and winding temperature of the transformer. A high-resolution camera is used to photograph the surface of the transformer to ensure that the image is clear and covers the entire transformer surface. The timestamp of the image capture is recorded for subsequent alignment with the temperature data. The acquired temperature monitoring data is organized into a structured format (such as CSV or Excel) to ensure that the temperature data of each sensor has a clear timestamp. The surface image is named and stored in an easily accessible folder. An appropriate interpolation method is selected, such as inverse distance weighted (IDW), kriging interpolation, or spline interpolation, to interpolate the temperature distribution on the surface image. The transformer surface image is generated. Form a three-dimensional grid and ensure that the resolution of the grid is high enough to capture temperature changes more accurately. Each grid point will be used for subsequent temperature interpolation calculations. Use the selected interpolation algorithm to interpolate the temperature of each grid point in the three-dimensional grid according to the temperature monitoring data, record the interpolation results of each grid point to form a complete three-dimensional temperature distribution data set, and construct a three-dimensional temperature distribution model based on the interpolated temperature grid point data. Use three-dimensional visualization tools (such as Matplotlib, ParaView or Mayavi) to perform three-dimensional reconstruction to display the temperature distribution on the surface and inside of the transformer. Perform trend analysis on the reconstructed three-dimensional temperature distribution, identify temperature change trends and hot spots, and record key features of the temperature distribution, such as the highest temperature point and temperature gradient.

[0091] Step S5: performing iterative correlation evolution learning on the three-dimensional temperature distribution trend field based on the dynamic trend representation of the power loss, and constructing a power-temperature situation map;

[0092] In this embodiment, deep learning and other methods are used to mine implicit correlation features between the two, establish correlation data reflecting the relationship between power loss and temperature distribution, discretize temperature features in the time dimension, and convert them into three-dimensional temperature distribution vectors to provide input data for subsequent feature space mapping processing. The integrated data is divided into a training set and a test set to ensure the effectiveness of model training and evaluation. The selected model is trained using the training set. The model parameters are iteratively optimized to reduce the loss function (such as mean squared error). Based on the performance during training, the model hyperparameters (such as learning rate, batch size, number of layers, etc.) are adjusted to optimize model performance. The trained model is evaluated using the test set. The accuracy of the predicted power loss and temperature distribution is calculated, the difference between the model prediction results and the actual values ​​is analyzed, and the performance of the model under specific conditions is identified. The trained model is used to predict the power loss and temperature distribution in future time periods, and a power-temperature situation map is generated. The power-temperature situation map is visualized using a visualization tool (such as Matplotlib, Plotly, or Tableau) to show the relationship between power and temperature under different conditions.

[0093] Step S6: Based on the power-temperature situation map, locate the abnormal energy consumption nodes of the transformer, perform intelligent load parameter fine-tuning, and build an intelligent load fine-tuning engine.

[0094] In this embodiment, power-temperature situation map data and real-time monitoring parameters (such as current, voltage, temperature, etc.) are collected to ensure that the timestamps of the data are consistent for subsequent analysis. A suitable anomaly detection method is selected, such as Z-score, IQR (interquartile range) method or machine learning method (such as isolation forest or LOF). The selected algorithm is used to analyze the power-temperature situation map to identify abnormal energy consumption nodes (such as nodes that are significantly higher than the normal range). An intelligent load fine-tuning strategy is formulated, and machine learning or optimization algorithms (such as genetic algorithms, particle swarm optimization) are combined to determine the optimal load parameters. According to the characteristics of the abnormal energy consumption nodes, the load parameter values ​​that need to be fine-tuned are calculated to ensure that they return to the normal range, and the calculated The load parameters are applied to the control system of the transformer, and the fine-tuning effect is monitored in real time. The load parameters and energy consumption data after fine-tuning are observed to ensure that energy consumption returns to normal levels. The changes before and after fine-tuning are recorded. The architecture of the intelligent load fine-tuning engine is designed, including the data input module, analysis and decision-making module, execution module and feedback module. The various modules of the fine-tuning engine are developed to ensure that they can work together efficiently to achieve automated load fine-tuning. The fine-tuning engine is comprehensively tested to ensure its stability and reliability under different conditions. The intelligent load fine-tuning engine is deployed in the production environment and monitored in real time to ensure its effective operation. Based on actual operation data and feedback, the algorithms and strategies of the fine-tuning engine are continuously optimized to improve its intelligence level.

[0095] In this embodiment, refer to Figure 2 , is a flowchart of the detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include:

[0096] Step S11: continuously monitoring the normal operation status of the transformer based on multiple sensors and extracting real-time operation monitoring parameters of the transformer;

[0097] Step S12: performing multi-scale adaptive filtering on the transformer real-time operation monitoring parameters to generate filtered optimized monitoring parameters;

[0098] Step S13: performing time series window segmentation on the filter optimization monitoring parameters to obtain monitoring parameters for multiple time periods;

[0099] Step S14: performing time series load characteristic calculation on the monitoring parameters of the multiple time periods to obtain transformer load characteristics of the multiple time periods.

[0100] In this embodiment, multiple sensors are deployed in the transformer system, including current sensors, temperature sensors, vibration sensors, etc., to continuously collect multi-dimensional operating parameter data of the transformer under normal operating conditions, such as current, temperature, vibration, etc. Various important operating status parameter indicators are extracted from the collected raw monitoring data as real-time operating monitoring parameters of the transformer. Filter parameters (such as thresholds, scaling factors, etc.) are adjusted according to the characteristics of the monitoring parameters. The real-time operating monitoring parameters are filtered to eliminate noise and extract more accurate monitoring parameters. The size and overlapping step of the timing window are determined, and an appropriate time period (such as every hour, every minute, etc.) is selected. The filtered and optimized monitoring parameters are segmented using a sliding window method to generate monitoring parameter data sets for multiple time periods. The load characteristic indicators to be calculated, such as average load, maximum load, load factor, and load fluctuation, are determined. The statistical characteristics of the transformer load in the corresponding time period are calculated, including indicators such as average value, variance, peak factor, and fluctuation range, thereby obtaining the load characteristic parameters of the transformer in different time periods.

[0101] In this embodiment, the specific steps of step S12 are:

[0102] Calculate the mean value of the transformer's real-time operation monitoring parameters to obtain the normalized range of the monitoring parameters;

[0103] Based on the normalized range of monitoring parameters, the abnormal outliers of the real-time operation monitoring parameters of the transformer are eliminated to obtain the abnormal value elimination parameters;

[0104] Perform multi-scale wavelet transform decomposition on the outlier removal parameters to extract detailed feature information in different frequency domain bands;

[0105] Discrete frequency components are calculated for the detailed feature information of different frequency domain bands to obtain the frequency component value of each frequency band;

[0106] Adaptively optimize the filter coefficients of the detail feature information of different frequency bands based on the frequency component value of each frequency band to generate multiple frequency band filter data;

[0107] The filter data of multiple frequency bands are reconstructed by nonlinear transformation to generate filter optimization monitoring parameters.

[0108] In this embodiment, the monitoring parameter data is filtered and optimized, and the statistical mean and standard deviation of each type of monitoring parameter are calculated to determine the normalization range of the monitoring parameter under normal operating conditions. The monitoring parameters are then detected for outliers to determine whether each data point is outside the normalization range. Monitoring parameters that exceed the normalization range are considered outliers and eliminated, while retaining normal data. A suitable mother wavelet (such as Haar wavelet or Daubechies wavelet) is selected, and a wavelet transform is performed based on the data characteristics. The monitoring parameters after eliminating the outliers are then subjected to a wavelet transform to extract detailed feature information at different scales. A suitable adaptive filtering method (such as the LMS algorithm or the RLS algorithm) is selected to optimize the filter coefficients. Based on the frequency component values ​​of each frequency band, the detailed features of different frequency domain bands are adaptively filtered to optimize the signal. The optimized filtered data of multiple frequency bands are organized into a data set. The data is fused through nonlinear transformation (such as wavelet reconstruction) to reconstruct and generate the final filtered optimized monitoring parameters. These multi-optimized monitoring parameters will provide a high-quality data foundation for subsequent anomaly detection.

[0109] In this embodiment, refer to Figure 3 , is a flowchart of the detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include:

[0110] Step S21: performing load serialization fitting on the transformer load characteristics of multiple time periods to construct a time series load characteristic curve;

[0111] Step S22: performing load change analysis on the time series load characteristic curve to obtain time series load change data;

[0112] Step S23: performing load parameter distribution identification on the transformer load characteristics in multiple time periods to generate load distribution data for different time periods;

[0113] Step S24: performing load distribution trend evolution on the time series load change data based on the load distribution data of different time periods, thereby obtaining load distribution change trend data.

[0114] In this embodiment, the load characteristic data of the transformer is extracted from multiple time periods to ensure the integrity and continuity of the data. An appropriate serialization method (such as linear interpolation, spline interpolation or polynomial fitting) is selected to convert the load characteristic data into a time series format. The serialized load data is fitted to generate a time series load characteristic curve. The fitting effect is optimized using techniques such as the least squares method. An appropriate load change analysis method is determined, such as the difference method, moving average or exponential smoothing. The selected analysis method is used to process the time series load characteristic curve, calculate the load change rate or change amount, organize the obtained time series load change data into a structured format, and record the load changes in different time periods. Changes in load parameters are determined, such as mean, variance, skewness, and kurtosis. Appropriate statistical distribution models (such as normal distribution, Log-Normal distribution, or Gamma distribution) are selected to identify the load parameter distribution. Load characteristic data for each time period are analyzed to identify the corresponding load distribution. Load distribution data for different time periods are generated. Appropriate trend analysis methods are selected, such as time series analysis, regression analysis, or machine learning algorithms. Load distribution data for different time periods are integrated with time-series load change data in preparation for trend evolution analysis. The integrated data are analyzed to identify the evolution pattern of load distribution over time and record the change trend.

[0115] In this embodiment, refer to Figure 4 , is a flowchart of the detailed implementation steps of step S3. In this embodiment, the detailed implementation steps of step S3 include:

[0116] Step S31: extracting transformer real-time current and voltage data based on transformer real-time operation monitoring parameters;

[0117] Step S32: performing multi-time point power calculation on the transformer load characteristics of multiple time periods according to the real-time current and voltage data of the transformer to generate load power parameters of the multiple time periods;

[0118] Step S33: performing reactive power calculation on the load power parameters of multiple time periods to generate load reactive power parameters for each time period;

[0119] Step S34: performing actual loss calculation on the load power parameters of multiple time periods and the load reactive power parameters of each time period, thereby obtaining load loss power data of different time periods;

[0120] Step S35: performing deep dynamic trend mining on the load loss power data of different time periods according to the load distribution change trend data, thereby obtaining a dynamic trend representation of the loss power.

[0121] In this embodiment, current and voltage data are extracted from the real-time monitoring system of the transformer to ensure the real-time and accuracy of the data. The extracted current and voltage data are verified to ensure that they are within the normal range and to exclude abnormal values. Power calculation is performed on the current and voltage data of each time period to generate corresponding load power parameters. Reactive power calculation is performed on the current and voltage data of each time period to generate corresponding reactive power parameters. Combined with the loss characteristics of the transformer, the actual power loss of the transformer in different time periods is calculated to obtain the load loss power data of each time period. A suitable dynamic trend mining method is selected, such as time series analysis, machine learning algorithm (such as LSTM) or regression analysis, to integrate the load loss power data with the load distribution change trend data in preparation for in-depth analysis. The integrated data is analyzed to identify the dynamic change pattern of the power loss and record the trend characteristics.

[0122] In this embodiment, step S4 includes the following steps:

[0123] Step S41: Obtain transformer temperature monitoring data and transformer surface image;

[0124] Step S42: performing three-dimensional meshing on the transformer surface image to obtain a transformer surface network;

[0125] Step S43: extracting multiple grid points based on the transformer surface network;

[0126] Step S44: performing multi-dimensional uniform interpolation on multiple grid points according to the transformer temperature monitoring data, thereby obtaining a transformer temperature network;

[0127] Step S45: performing a three-dimensional temperature distribution analysis on the transformer temperature network to obtain three-dimensional grid temperature distribution data;

[0128] Step S46: performing multi-time temperature change fluctuation analysis on the three-dimensional grid temperature distribution data to generate three-dimensional temperature change fluctuation data;

[0129] Step S47: reconstructing the three-dimensional temperature distribution of the three-dimensional temperature change fluctuation data to construct a three-dimensional temperature distribution trend field.

[0130] In this embodiment, temperature sensors are installed at key locations of the transformer to regularly collect temperature monitoring data. A high-resolution camera is used to photograph the transformer surface to ensure image clarity and integrity. The acquired surface image is preprocessed, including noise reduction and contrast enhancement, to improve subsequent processing effects. A suitable meshing algorithm, such as Delaunay triangulation or Voronoi diagram, is selected. Three-dimensional modeling and meshing are performed on the preprocessed surface image to generate a three-dimensional grid structure covering the transformer surface. A standard for extracting grid points is determined, such as evenly spaced points or points within a specific area. Multiple grid points are extracted from the generated surface network, their three-dimensional coordinates are recorded, and a suitable interpolation algorithm, such as Kriging interpolation, inverse distance weighted (IDW), or spline interpolation, is selected. , use the temperature monitoring data to perform multi-dimensional uniform interpolation on the extracted grid points to generate a temperature network on the transformer surface, select a suitable three-dimensional temperature distribution analysis method, such as heat map generation or three-dimensional visualization tools (such as Matplotlib, ParaView), analyze the generated temperature network, generate temperature distribution data of the three-dimensional grid, analyze the changes in temperature distribution between different moments, including indicators such as temperature fluctuation amplitude and rate, generate fluctuation data describing the three-dimensional temperature change trend, select a suitable reconstruction algorithm, such as inverse distance weighting or wavelet reconstruction, use the three-dimensional temperature change fluctuation data to reconstruct the temperature distribution, generate a three-dimensional temperature distribution trend field, visualize the reconstructed three-dimensional temperature distribution trend field, and show the dynamic change trend of the temperature distribution.

[0131] In this embodiment, step S5 includes the following steps:

[0132] Step S51: mining implicit correlation features of the three-dimensional temperature distribution trend field based on the power loss dynamic trend characterization to obtain power loss-temperature distribution correlation data;

[0133] Step S52: Discretize the time series temperature characteristics of the three-dimensional temperature distribution trend field and extract the three-dimensional temperature distribution vector;

[0134] Step S53: performing feature space mapping processing on the dynamic trend representation of power loss and the three-dimensional temperature distribution vector to construct a potential feature space model;

[0135] Step S54: performing iterative correlation evolution learning on the potential feature space model according to the loss power-temperature distribution correlation data to construct a power-temperature situation map.

[0136] In this embodiment, a suitable association mining method is selected, such as the Pearson correlation coefficient, Spearman rank correlation or multiple regression analysis, and the selected method is used to analyze the implicit association between the loss power and the temperature distribution to generate loss power-temperature distribution association data. The temperature characteristics are discretized from the time dimension, the three-dimensional temperature distribution trend field is discretized, and a representative three-dimensional temperature distribution vector is extracted. A suitable feature space mapping method is selected, such as principal component analysis (PCA), t-SNE or autoencoder, to map the loss power dynamic trend representation and the three-dimensional temperature distribution vector, to construct a latent feature space model, and a suitable iterative learning algorithm is selected, such as the gradient descent method, the back propagation algorithm or the reinforcement learning method. The latent feature space model is trained using the loss power-temperature distribution association data, the model parameters are updated, the association features are optimized, and a power-temperature situation map is generated according to the updated model to display the dynamic relationship between the loss power and the temperature distribution. The constructed power-temperature situation map is visualized to ensure that the information is clear and easy to understand.

[0137] In this embodiment, step S6 includes the following steps:

[0138] Step S61: monitoring abnormal energy consumption parameter deviations of the transformer based on the power-temperature situation map, and marking abnormal energy consumption deviation parameters;

[0139] Step S62: locating abnormal nodes on the transformer surface network according to the abnormal energy consumption deviation parameter, and extracting abnormal energy consumption grid points;

[0140] Step S63: performing abnormal type attribution identification on the abnormal energy consumption grid point to obtain abnormal type attribution data;

[0141] Step S64: Perform intelligent load parameter fine-tuning based on the abnormal type attribution data and build an intelligent load fine-tuning engine.

[0142] In this embodiment, the transformer's power loss and temperature distribution data are monitored in real time and compared with the normal operating range in the situation map to identify abnormal power loss and temperature deviation parameters. The marked abnormal energy consumption deviation parameters are integrated with the transformer surface network data to ensure data consistency. A suitable abnormal node location method, such as distance-based cluster analysis or threshold-based screening, is selected. Based on the abnormal energy consumption deviation parameters, abnormal nodes are located in the surface network and corresponding abnormal energy consumption grid points are extracted. A suitable attribution identification method, such as a decision tree, random forest, or neural network, is selected to organize the feature data of the abnormal energy consumption grid points in preparation for attribution analysis. The selected attribution identification method is used to analyze the abnormal energy consumption grid points to identify the abnormality type and cause. Based on the abnormality type attribution data, load parameters that require fine-tuning, such as current, voltage, or power settings, are determined. An intelligent load fine-tuning strategy is developed, and a machine learning algorithm (such as reinforcement learning) is combined to optimize the load parameters. The developed strategy is applied to fine-tune the load parameters, and the fine-tuning effect is monitored in real time. The fine-tuning process and algorithm are integrated to build an intelligent load fine-tuning engine to ensure subsequent automated execution.

[0143] In this embodiment, a transformer-end power consumption monitoring system is provided, which is used to execute the transformer-end power consumption monitoring method described above, including:

[0144] The time series segmentation module is used to obtain the real-time operation monitoring parameters of the transformer; the real-time operation monitoring parameters of the transformer are segmented into time series windows, and the time series load characteristics are calculated to obtain the transformer load characteristics of multiple time periods;

[0145] The load distribution module is used to construct a time-series load characteristic curve based on the transformer load characteristics of multiple time periods; the load distribution trend evolution of the time-series load characteristic curve is performed to obtain load distribution change trend data;

[0146] Dynamic trend mining module, used to conduct deep dynamic trend mining on transformer real-time operation monitoring parameters based on load distribution change trend data, thereby obtaining dynamic trend representation of power loss;

[0147] The three-dimensional temperature distribution module is used to obtain transformer temperature monitoring data and transformer surface images; based on the transformer temperature monitoring data, the transformer surface image is subjected to multi-dimensional uniform interpolation, and the three-dimensional temperature distribution is reconstructed to construct a three-dimensional temperature distribution trend field;

[0148] The situation map module is used to perform iterative correlation evolution learning on the three-dimensional temperature distribution trend field based on the dynamic trend representation of power loss, and construct a power-temperature situation map;

[0149] The parameter fine-tuning module is used to locate abnormal energy consumption nodes of the transformer based on the power-temperature situation map, perform intelligent load parameter fine-tuning, and build an intelligent load fine-tuning engine.

[0150] The present invention helps to capture the time series characteristics of the transformer operating status by obtaining real-time monitoring parameters and performing time series window segmentation, providing a real-time data basis. The time series load characteristic calculation can help analyze the load characteristics of different time periods, and provide key data for subsequent load distribution analysis and dynamic trend mining. Constructing a time series load characteristic curve based on the load characteristics of multiple time periods helps to reveal the changing trend of load characteristics over time, and provides visualization support for the evolution of load distribution trends. Analyzing the load distribution change trend data helps to understand the laws and trends of load changes, and provides important clues for subsequent dynamic trend mining. Through deep dynamic trend mining, the dynamic trends of the transformer real-time monitoring parameters can be revealed according to the load distribution change trend data, including the dynamic characterization of loss power. Obtaining the dynamic trend characterization of loss power helps to understand the laws and trends of power changes, and provides in-depth information for energy consumption monitoring and analysis. Insight: Through multi-dimensional uniform interpolation and three-dimensional temperature distribution reconstruction, an accurate three-dimensional temperature distribution trend field can be constructed to reveal the spatial distribution characteristics of the transformer surface temperature. Constructing a three-dimensional temperature distribution trend field helps to understand the dynamic changes of temperature distribution and provides a basis for power-temperature correlation analysis and situation map construction. Through iterative correlation evolution learning, a power-temperature situation map is constructed to reveal the correlation between power and temperature, providing a comprehensive perspective for energy consumption monitoring and anomaly detection. Constructing a power-temperature situation map helps to fully understand the dynamic situation of power consumption at the transformer end and provide a basis for subsequent parameter fine-tuning. Through abnormal energy consumption node positioning and intelligent load parameter fine-tuning, intelligent adjustment of the transformer can be achieved, energy utilization efficiency can be improved, and energy consumption losses can be reduced. Building an intelligent load fine-tuning engine helps to adjust load parameters in real time, ensure safe and stable operation of equipment, and improve equipment life and performance.

[0151] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0152] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A method for monitoring power consumption at a transformer end, characterized in that: The following steps are involved: Step S1: Obtaining transformer real-time operation monitoring parameters; The real-time operation monitoring parameters of the transformer are divided into time series windows and time series load characteristics are calculated to obtain the transformer load characteristics of multiple time periods; Step S2: constructing a time series load characteristic curve based on the transformer load characteristics of multiple time periods; Perform load distribution trend evolution on the time series load characteristic curve to obtain load distribution change trend data; Step S3: performing deep dynamic trend mining on the transformer real-time operation monitoring parameters based on the load distribution change trend data, thereby obtaining a dynamic trend representation of the power loss; Step S4: Obtain transformer temperature monitoring data and transformer surface image; Based on the transformer temperature monitoring data, the transformer surface image is interpolated in multiple dimensions, and the three-dimensional temperature distribution is reconstructed to construct a three-dimensional temperature distribution trend field. Step S5: performing iterative correlation evolution learning on the three-dimensional temperature distribution trend field based on the dynamic trend representation of the power loss, and constructing a power-temperature situation map; Step S6: Based on the power-temperature situation map, locate the abnormal energy consumption node of the transformer, perform intelligent load parameter fine-tuning, and build an intelligent load fine-tuning engine; Among them, the specific steps of step S1 are: Step S11: continuously monitoring the normal operation status of the transformer based on multiple sensors and extracting real-time operation monitoring parameters of the transformer; Step S12: performing multi-scale adaptive filtering on the transformer real-time operation monitoring parameters to generate filtered optimized monitoring parameters; Step S13: performing time series window segmentation on the filter optimization monitoring parameters to obtain monitoring parameters for multiple time periods; Step S14: performing time series load characteristic calculation on the monitoring parameters of the multiple time periods to obtain transformer load characteristics of the multiple time periods; Among them, the specific steps of step S12 are: Calculate the mean value of the transformer's real-time operation monitoring parameters to obtain the normalized range of the monitoring parameters; Based on the normalized range of monitoring parameters, the abnormal outliers of the real-time operation monitoring parameters of the transformer are eliminated to obtain the abnormal value elimination parameters; Perform multi-scale wavelet transform decomposition on the outlier removal parameters to extract detailed feature information in different frequency domain bands; Discrete frequency components are calculated for the detailed feature information of different frequency domain bands to obtain the frequency component value of each frequency band; Adaptively optimize the filter coefficients of the detail feature information of different frequency bands based on the frequency component value of each frequency band to generate multiple frequency band filter data; The filter data of multiple frequency bands are reconstructed by nonlinear transformation to generate filter optimization monitoring parameters.

2. The method for monitoring power consumption at the transformer end according to claim 1, characterized in that: The specific steps of step S2 are: Step S21: performing load serialization fitting on the transformer load characteristics of multiple time periods to construct a time series load characteristic curve; Step S22: performing load change analysis on the time series load characteristic curve to obtain time series load change data; Step S23: performing load parameter distribution identification on the transformer load characteristics in multiple time periods to generate load distribution data for different time periods; Step S24: performing load distribution trend evolution on the time series load change data based on the load distribution data of different time periods, thereby obtaining load distribution change trend data.

3. The method according to claim 1, characterized in that The specific steps of step S3 are: Step S31: extracting transformer real-time current and voltage data based on transformer real-time operation monitoring parameters; Step S32: performing multi-time point power calculation on the transformer load characteristics of multiple time periods according to the real-time current and voltage data of the transformer to generate load power parameters of the multiple time periods; Step S33: performing reactive power calculation on the load power parameters of multiple time periods to generate load reactive power parameters for each time period; Step S34: performing actual loss calculation on the load power parameters of multiple time periods and the load reactive power parameters of each time period, thereby obtaining load loss power data of different time periods; Step S35: performing deep dynamic trend mining on the load loss power data of different time periods according to the load distribution change trend data, thereby obtaining a dynamic trend representation of the loss power.

4. The method for monitoring power consumption at the transformer end according to claim 1, wherein: The specific steps of step S4 are: Step S41: Obtain transformer temperature monitoring data and transformer surface image; Step S42: performing three-dimensional meshing on the transformer surface image to obtain a transformer surface network; Step S43: extracting multiple grid points based on the transformer surface network; Step S44: performing multi-dimensional uniform interpolation on multiple grid points according to the transformer temperature monitoring data, thereby obtaining a transformer temperature network; Step S45: performing a three-dimensional temperature distribution analysis on the transformer temperature network to obtain three-dimensional grid temperature distribution data; Step S46: performing multi-time temperature change fluctuation analysis on the three-dimensional grid temperature distribution data to generate three-dimensional temperature change fluctuation data; Step S47: reconstructing the three-dimensional temperature distribution of the three-dimensional temperature change fluctuation data to construct a three-dimensional temperature distribution trend field.

5. The method for monitoring power consumption at the transformer end according to claim 1, characterized in that: The specific steps of step S5 are: Step S51: mining implicit correlation features of the three-dimensional temperature distribution trend field based on the power loss dynamic trend characterization to obtain power loss-temperature distribution correlation data; Step S52: Discretize the time series temperature characteristics of the three-dimensional temperature distribution trend field and extract the three-dimensional temperature distribution vector; Step S53: performing feature space mapping processing on the dynamic trend representation of power loss and the three-dimensional temperature distribution vector to construct a potential feature space model; Step S54: performing iterative correlation evolution learning on the potential feature space model according to the loss power-temperature distribution correlation data to construct a power-temperature situation map.

6. The method for monitoring power consumption at the transformer end according to claim 1, characterized in that: The specific steps of step S6 are: Step S61: monitoring abnormal energy consumption parameter deviations of the transformer based on the power-temperature situation map, and marking abnormal energy consumption deviation parameters; Step S62: locating abnormal nodes on the transformer surface network according to the abnormal energy consumption deviation parameter, and extracting abnormal energy consumption grid points; Step S63: performing abnormal type attribution identification on the abnormal energy consumption grid point to obtain abnormal type attribution data; Step S64: Perform intelligent load parameter fine-tuning based on the abnormal type attribution data and build an intelligent load fine-tuning engine.

7. A transformer-end power consumption monitoring system, characterized in that: The method for monitoring power consumption at the transformer end according to claim 1 comprises: The time series segmentation module is used to obtain the real-time operation monitoring parameters of the transformer; the real-time operation monitoring parameters of the transformer are segmented into time series windows, and the time series load characteristics are calculated to obtain the transformer load characteristics of multiple time periods; The load distribution module is used to construct a time-series load characteristic curve based on the transformer load characteristics of multiple time periods; the load distribution trend evolution of the time-series load characteristic curve is performed to obtain load distribution change trend data; Dynamic trend mining module, used to conduct deep dynamic trend mining on transformer real-time operation monitoring parameters based on load distribution change trend data, thereby obtaining dynamic trend representation of power loss; The three-dimensional temperature distribution module is used to obtain transformer temperature monitoring data and transformer surface images; based on the transformer temperature monitoring data, the transformer surface image is subjected to multi-dimensional uniform interpolation, and the three-dimensional temperature distribution is reconstructed to construct a three-dimensional temperature distribution trend field; The situation map module is used to perform iterative correlation evolution learning on the three-dimensional temperature distribution trend field based on the dynamic trend representation of power loss, and construct a power-temperature situation map; The parameter fine-tuning module is used to locate abnormal energy consumption nodes of the transformer based on the power-temperature situation map, perform intelligent load parameter fine-tuning, and build an intelligent load fine-tuning engine.

Images