An intelligent monitoring and control system for power transformers

Abstract

The present invention discloses an intelligent monitoring and control system for power transformers, which relates to the technical field of intelligent monitoring and control of power transformers. The system includes an initial automatic adjustment module, a data acquisition and transmission module, a data preprocessing and feature extraction module, a feature analysis and prediction module, a load classification module, a conventional load control module, and a high-frequency load adaptive control module. The system collects load data in real time through the Internet of Things (IoT) and uses machine learning to predict load change types. For conventional loads, the system maintains the initial frequency to reduce wear. For high-frequency loads, the system adaptively increases the tap changer frequency to ensure rapid transformer response and avoid equipment failures caused by voltage instability. This ensures efficient transformer operation under different load scenarios, reduces the risk of overvoltage or undervoltage, provides reliable power supply for industrial equipment and medical systems, and improves grid stability and security.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring and control of power transformers, and in particular to an intelligent monitoring and control system for power transformers. Background Art

[0002] Intelligent monitoring and control of power transformers involves real-time monitoring and data collection of key transformer operating parameters (such as temperature, current, voltage, oil level, and humidity) through sensors, IoT technology, and intelligent algorithms. This data is then processed and predicted using big data analytics and machine learning models to identify potential abnormal trends or failure risks. The system automatically adjusts transformer operating parameters based on load changes and operating conditions, such as through intelligent control of the cooling system, open-loop load changer (OLTC), and load distribution strategies. This ensures efficient and safe operation of the transformer, extending equipment life and reducing maintenance costs. This process not only improves power system reliability but also enables precise control without human intervention.

[0003] When it comes to intelligent monitoring and regulation of power transformers, automatic tap changer adjustment is a key technology for achieving grid voltage regulation by controlling the tap position of the transformer windings under dynamic conditions such as load changes and grid voltage fluctuations. This ensures the safe and efficient operation of the grid and its electrical equipment. This adjustment process can be performed without power outages, making it a crucial technical tool for achieving grid voltage regulation.

[0004] The existing technology has the following deficiencies:

[0005] The automatic frequency adjustment of existing tap changers is typically fixed and incapable of adaptively adjusting to the dynamic changes in the power transformer load. This is because traditional systems rely primarily on preset voltage thresholds and time intervals to determine whether to switch, in order to avoid mechanical wear and equipment failure caused by frequent switching. When the automatic frequency adjustment of the tap changer cannot adapt to the high-frequency dynamic changes in the power transformer load, the secondary-side output voltage fluctuates frequently and cannot be adjusted in a timely manner, causing the power supply equipment to operate outside the rated voltage range for a long time. Overvoltage can accelerate the aging or burning of insulation materials in sensitive equipment such as motors and capacitors, increasing the risk of short circuits and fires. Undervoltage, on the other hand, can cause unstable operation of critical equipment (such as industrial automation systems and medical equipment), even leading to sudden shutdowns, directly impacting production processes and medical care, and seriously threatening life safety and economic operations.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0007] The purpose of the present invention is to provide an intelligent monitoring and control system for power transformers, which collects load data in real time through Internet of Things technology, and uses machine learning models to analyze and predict the type of dynamic load changes. For conventional dynamically changing loads, the initial adjustment frequency is maintained to reduce mechanical wear and maintenance costs; for high-frequency dynamically changing loads, the tap changer frequency is adaptively increased to ensure that the transformer responds quickly to fluctuations and avoids voltage instability causing equipment failures, thereby ensuring efficient operation of the transformer in different load scenarios, reducing the risk of overvoltage or undervoltage, providing reliable power supply support for key industrial equipment and medical systems, and improving the safety and stability of the power grid to solve the problems in the above-mentioned background technology.

[0008] To achieve the above-mentioned object, the present invention provides the following technical solutions: an intelligent monitoring and control system for power transformers, comprising an initial automatic adjustment module, a data acquisition and transmission module, a data preprocessing and feature extraction module, a feature analysis and prediction module, a load classification module, a conventional load control module, and a high-frequency load adaptive control module;

[0009] Initial automatic adjustment module: In the initial stage, the tap changer is set to the initial automatic adjustment frequency to control the tap position of the transformer winding to ensure that the secondary side output voltage is stable within the predetermined range;

[0010] The data acquisition and transmission module collects the load data of the power transformer in real time during operation and transmits the collected data to the central monitoring system through the Internet of Things technology;

[0011] The data preprocessing and feature extraction module preprocesses the collected load change information and extracts key features that reflect the dynamic changes in the load of the power transformer;

[0012] The feature analysis and prediction module analyzes the extracted features within the monitoring window and inputs the analyzed data into a pre-trained machine learning model. The machine learning model uses the input feature data to predict the future dynamic changes in the power transformer's load.

[0013] The load classification module classifies the load changes of power transformers within the monitoring window into two categories: high-frequency dynamic change type and conventional dynamic change type based on the prediction results of the machine learning model;

[0014] The conventional load control module continues to maintain the initial automatic adjustment of the frequency control transformer winding tap position for load changes classified as conventional dynamic changes;

[0015] The high-frequency load adaptive control module adjusts the actual automatic adjustment frequency of the tap changer based on the characteristics of the current load change for load changes classified as high-frequency dynamic changes. Specifically, it increases the adjustment frequency of the tap changer based on the initial adjustment frequency so that the adjusted actual automatic adjustment frequency responds more quickly to the high-frequency dynamic changes of the load.

[0016] Preferably, the initial automatic frequency adjustment is based on historical grid load data and control strategies, and is applicable to conventional load change scenarios.

[0017] Preferably, key features reflecting the dynamic changes in the load of the power transformer are extracted, wherein the extracted key features include the pulsation of the reactive power demand in the monitoring window and the offset degree and offset frequency of the load frequency in the monitoring window. After acquisition, the pulsation of the reactive power demand in the monitoring window and the offset degree and offset frequency of the load frequency in the monitoring window are analyzed to generate a reactive demand pulsation index and a load frequency offset index respectively. The reactive demand pulsation index is used to quantify the pulsation frequency and amplitude changes of the reactive power demand in the monitoring window, reflecting the degree of fluctuation and instability of the reactive demand in the power grid. The load frequency offset index is used to quantify the offset degree and offset frequency of the load frequency in the monitoring window, reflecting the dynamic deviation between the system load and the power grid frequency.

[0018] Preferably, after obtaining the reactive demand pulsation index and load frequency offset index generated by analyzing the key features, the reactive demand pulsation index and load frequency offset index are input into a pre-trained machine learning model to generate a dynamic change coefficient, and the future dynamic load changes of the power transformer are predicted through the dynamic change coefficient.

[0019] Preferably, the dynamic change coefficient generated when predicting the future dynamic change of the load of the power transformer by the machine learning model is compared and analyzed with a preset dynamic change coefficient reference threshold, and the load change of the power transformer within the monitoring window is divided. The specific division steps are as follows:

[0020] If the dynamic change coefficient is greater than the dynamic change coefficient reference threshold, the load change of the power transformer under the monitoring window is classified as a high-frequency dynamic change type;

[0021] If the dynamic change coefficient is less than or equal to the dynamic change coefficient reference threshold, the load change situation of the power transformer in the monitoring window is classified as a conventional dynamic change type.

[0022] Preferably, the specific steps of generating a reactive power demand pulsation index after analyzing the pulsation of reactive power demand within the monitoring window are as follows:

[0023] Real-time acquisition of reactive power demand within the monitoring window Time series data, represents the reactive power demand at time t;

[0024] Preprocess the reactive power data and convert the preprocessed reactive power time series It will serve as the basic input for subsequent calculations;

[0025] Based on the extracted reactive power demand , construct the reactive power instantaneous pulsation deviation function to quantify the deviation between the time t and the reference state. The expression of the reactive power instantaneous pulsation deviation function is: , where is the reference reactive power level based on historical data, To adjust the coefficient, control the amplitude change of the reference power, The pulsation frequency of reactive power demand reflects the speed of load fluctuation. is the phase angle, indicating the initial phase of the reactive power fluctuation, is the instantaneous pulsation deviation of reactive power at time t;

[0026] The changing trend of the instantaneous pulsation deviation function is combined with the pulsation frequency to generate the reactive power demand pulsation index, which is used to quantitatively describe the comprehensive impact of the pulsation frequency and amplitude of reactive power. The generation expression of the reactive power demand pulsation index is:

[0027] ,

[0028] Where, represents the reactive power demand pulsation index, and Respectively represent the start and end time of the monitoring window, represents the instantaneous pulsation frequency of reactive power demand at time t, is the attenuation coefficient, which suppresses the influence of small pulsation amplitude on the final result.

[0029] Preferably, the specific steps of generating the load frequency deviation index after analyzing the deviation degree and deviation frequency of the load frequency within the monitoring window are as follows:

[0030] Within the set monitoring window, the actual load frequency data of the power transformer and the reference frequency data of the power grid are collected. By comparing the difference between the actual load frequency and the reference frequency, the load frequency offset at each time point is calculated. The calculation expression of the load frequency offset is: , where is the actual load frequency at time t, is the reference frequency of the power grid, is the load frequency offset at time t, reflecting the difference between the actual load and the reference frequency;

[0031] In each monitoring window, the load frequency offset amplitude and offset frequency are calculated to assess the severity of the load fluctuation. The offset amplitude refers to the maximum offset of the load frequency in the monitoring window. The calculation expression is: , where is the load frequency offset amplitude, indicating the maximum absolute value of the frequency offset within the monitoring window. Indicates the maximum value of all load frequency offsets within the monitoring window. The offset frequency is used to measure the load frequency fluctuation frequency, that is, the number of rapid fluctuations of the load frequency within the monitoring window. Each positive or negative change in the frequency offset is defined as a fluctuation. The offset frequency calculation formula is:

[0032] ,

[0033] Where, is the offset frequency of the load frequency, indicating how frequently the load frequency fluctuates within the monitoring window. T is the total time of the monitoring window, N is the number of sampling points in the monitoring window, is the Heaviside step function, which is used to determine whether the load frequency fluctuates. and Represents two consecutive moments, i Represents a time index;

[0034] The load frequency offset index is generated based on the calculated load frequency offset amplitude and offset frequency. It comprehensively reflects the offset degree and fluctuation frequency of the load frequency within the monitoring window. The generation expression of the load frequency offset index is: , where is the load frequency deviation index, which quantifies the degree of deviation and frequency change of the load frequency. and are weight coefficients, which are used to adjust the impact of offset amplitude and offset frequency on the load frequency offset index. and is an exponential coefficient used to amplify the nonlinear effects of the offset amplitude and offset frequency.

[0035] Preferably, for a load change situation classified as a high-frequency dynamic change type, the actual automatic adjustment frequency of the tap changer is adjusted according to the characteristics of the current load change. The specific steps are as follows:

[0036] In order to ensure accurate control, the initial control capability coefficient is introduced to describe the maximum dynamic change level that can be controlled by the initial adjustment frequency. The calculation expression of the initial control capability coefficient is:

[0037] ,

[0038] Where, is the initial adjustment frequency, reflecting the control ability under normal circumstances. is the initial regulation capacity coefficient, which describes the load response capacity of the initial regulation frequency. is the fluctuation of the reference threshold value of the dynamic change coefficient, which is used to describe the deviation degree of the system load. , where Indicates the maximum dynamic change coefficient within the monitoring window, indicating the peak fluctuation of the load during the monitoring window. is the reference threshold value of the dynamic change coefficient;

[0039] According to the deviation between the dynamic change coefficient and the dynamic change coefficient reference threshold, the adjustment frequency boost coefficient is calculated, and the actual automatic adjustment frequency is calculated by the adjustment frequency boost coefficient. The calculation expression of the actual automatic adjustment frequency is: , where is the actual automatic adjustment frequency after adjustment, To adjust the frequency boost coefficient, determine the frequency boost range, adjust the frequency boost coefficient The calculation expression is:

[0040] ,

[0041] Where, is the dynamic change coefficient calculated within the monitoring window, It is the sensitivity coefficient, which is used to control the degree of frequency increase to avoid equipment wear caused by excessive frequency increase.

[0042] Preferably, the automatic adjustment frequency of the tap changer is adjusted in real time according to the calculated actual automatic adjustment frequency, and a frequency feedback coefficient is introduced to monitor the frequency effect after adjustment during actual operation and perform dynamic optimization. The calculation expression of the frequency feedback coefficient is:

[0043] ,

[0044] Where, is the frequency feedback coefficient, which is used to evaluate the control effect of the new frequency. is the actual automatic adjustment frequency at time t, is the dynamic change coefficient at time t;

[0045] The calculated frequency feedback coefficient Reference threshold range with frequency feedback coefficient Compare and analyze the frequency effect after adjustment. If it meets the requirements , it indicates that the actual automatic adjustment frequency after adjustment is within a reasonable range. If it does not meet , it indicates that the actual automatic adjustment frequency after adjustment does not meet the demand, and the actual automatic adjustment frequency should be further optimized to make it within a reasonable range.

[0046] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0047] The present invention uses Internet of Things technology to collect load data in real time, and analyzes and predicts key features through machine learning models to accurately distinguish the dynamic change types of loads. For conventional dynamically changing loads, the initial adjustment frequency is maintained to avoid unnecessary frequent switching, thereby reducing mechanical wear and maintenance costs; and for high-frequency dynamically changing loads, the adjustment frequency of the tap changer is adaptively increased based on the prediction to ensure that the transformer responds quickly to load fluctuations in a short time, avoiding equipment damage and operational failures caused by voltage instability. This not only ensures the efficient operation of power transformers in different load scenarios, but also reduces the safety risks caused by overvoltage or undervoltage, providing more reliable power supply guarantees for highly sensitive scenarios such as key industrial equipment and medical systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0049] Figure 1 The figure is a module diagram of an intelligent monitoring and control system for power transformers according to the present invention. DETAILED DESCRIPTION

[0050] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0051] The present invention provides Figure 1 The intelligent monitoring and control system for power transformers shown includes an initial automatic adjustment module, a data acquisition and transmission module, a data preprocessing and feature extraction module, a feature analysis and prediction module, a load classification module, a conventional load control module, and a high-frequency load adaptive control module;

[0052] Initial automatic adjustment module: In the initial stage, the tap changer is set to the initial automatic adjustment frequency to control the tap position of the transformer winding to ensure that the secondary side output voltage is stable within the predetermined range;

[0053] This initial automatic frequency adjustment is based on historical grid load data and control strategies and is suitable for scenarios with typical load fluctuations. By maintaining voltage stability, it ensures safe and efficient operation of the grid and power-consuming equipment, preventing damage to equipment caused by overvoltage and undervoltage.

[0054] The data acquisition and transmission module collects the load data of the power transformer in real time during operation and transmits the collected data to the central monitoring system through the Internet of Things technology;

[0055] During operation, power transformer load data is collected in real time using sensors and IoT technology. This data primarily includes the following key indicators (including but not limited to): voltage (input and output voltages on the primary and secondary sides), current (load current through the transformer), power factor (the ratio of active power to total power), active and reactive power (a measure of the actual energy delivered by the transformer and reflecting the balance of power demand), frequency (variation in the grid's operating frequency), load factor (the ratio of the transformer's actual load relative to its rated capacity), temperature data (including the temperature of the transformer windings and oil tank), oil level (variation in the transformer's oil level, used to monitor the cooling system's operating status), and harmonic content (power quality issues caused by power harmonics, primarily reflecting the degree of current and voltage distortion). Additional information may also be included, such as winding insulation resistance, grid fluctuations, and external ambient temperature. Using IoT technology, this load data is efficiently transmitted to a central monitoring system, which analyzes, stores, and processes this real-time data to predict transformer load trends and generate early warning and control instructions to ensure efficient and safe power system operation.

[0056] The data preprocessing and feature extraction module preprocesses the collected load change information and extracts key features that reflect the dynamic changes in the load of the power transformer;

[0057] The feature analysis and prediction module analyzes the extracted features within the monitoring window and inputs the analyzed data into a pre-trained machine learning model. The machine learning model uses the input feature data to predict the future dynamic changes in the power transformer's load.

[0058] The key features reflecting the dynamic changes in the load of the power transformer are extracted, wherein the extracted key features include the pulsation of the reactive power demand in the monitoring window and the offset degree and offset frequency of the load frequency in the monitoring window. After acquisition, the pulsation of the reactive power demand in the monitoring window and the offset degree and offset frequency of the load frequency in the monitoring window are analyzed, and a reactive demand pulsation index and a load frequency offset index are generated respectively. The reactive demand pulsation index is used to quantify the pulsation frequency and amplitude changes of the reactive power demand in the monitoring window, reflecting the fluctuation degree and instability of the reactive demand in the power grid. The load frequency offset index is used to quantify the offset degree and offset frequency of the load frequency in the monitoring window, reflecting the dynamic deviation between the system load and the power grid frequency.

[0059] After obtaining the reactive demand pulsation index and load frequency offset index generated by analyzing the key features, the reactive demand pulsation index and load frequency offset index are input into a pre-trained machine learning model to generate a dynamic change coefficient, and the dynamic change coefficient is used to predict the future dynamic load changes of the power transformer.

[0060] A pre-trained machine learning model is a machine learning algorithm model that has been trained and optimized using extensive historical data before actual application to accurately predict dynamic load changes on power transformers. During the training phase, the model utilizes multi-dimensional data collected during power transformer operation, including features such as load fluctuations, voltage, current, reactive power demand, and frequency deviation. Through this training process, the model is able to identify complex underlying patterns and regularities in the power system and extract features closely related to dynamic load changes from this high-dimensional data. These models commonly utilize algorithms such as random forests, support vector machines, neural networks, and LSTM (long short-term memory) networks. The choice of algorithm depends on the nature of the data and the complexity of the problem being solved. During training, the model continuously learns and adjusts its internal parameters to maximize its ability to predict future load changes.

[0061] In practical applications, pre-trained machine learning models are used to infer and predict real-time input feature data, such as key features like the reactive power demand pulsation index and the load frequency deviation index. Because these models have already optimized their predictive capabilities using historical data during the training phase, when real-time data is input, the models can use these features to predict future trends in power transformer load. For example, when the reactive power demand pulsation index rises rapidly, combined with other input features such as the load frequency deviation index, the machine learning model can identify impending high-frequency load fluctuations or sudden load demand in the system. This enables the power control system to proactively adjust the tap changer frequency, maintaining system stability and avoiding equipment failures and power supply instability caused by excessive load frequency fluctuations. In summary, pre-trained machine learning models are intelligent predictive tools based on historical data learning and pattern recognition, effectively improving the control accuracy and responsiveness of power transformers.

[0062] If the reactive power demand fluctuates significantly within the monitoring window, it typically indicates that the power transformer's load is experiencing dynamic, high-frequency fluctuations. This is because reactive power is primarily associated with inductive or capacitive loads (such as motors and capacitor banks) in the power system. The frequent startup and shutdown or rapid changes in these loads can cause significant fluctuations in reactive power demand. High-frequency load fluctuations mean that inductive or capacitive equipment in the system is constantly switching in and out of operation, causing reactive power pulsation. If this pulsation is frequent and significant, it not only reflects the rapid fluctuations in load over a short period of time, but can also lead to voltage instability and power factor degradation. Promptly identifying and regulating these high-frequency reactive demand fluctuations can help avoid frequent transformer tap changers and unstable grid operations.

[0063] The specific steps for generating a reactive power demand pulsation index after analyzing the pulsation of reactive power demand within the monitoring window are as follows:

[0064] Real-time acquisition of reactive power demand within the monitoring window Time series data, represents the reactive power demand at time t;

[0065] Real-time collection of reactive power demand time series data within the monitoring window typically utilizes a combination of smart meters and power monitoring terminals. These devices integrate high-precision sensors capable of collecting parameters such as reactive power, current, and voltage at millisecond or even microsecond levels. These data are then transmitted to the monitoring system in real time via communication protocols such as Modbus, IEC61850, and DL / T645. SCADA systems can also be used to centrally collect and aggregate data on power equipment status and reactive power demand across distributed locations over the network, enabling remote real-time monitoring. For more granular monitoring, power quality analyzers can be deployed, combined with edge computing technology, for local preliminary data processing and storage, reducing network latency. This multi-layered data collection approach ensures high-precision, low-latency monitoring of reactive power demand, providing a stable data foundation for real-time analysis and regulation.

[0066] In order to ensure the accuracy and validity of the data, the reactive power data needs to be preprocessed, including noise removal, outlier detection and time series normalization. The preprocessed reactive power time series It will serve as the basic input for subsequent calculations.

[0067] Based on the extracted reactive power demand , construct the reactive power instantaneous pulsation deviation function to quantify the deviation between the time t and the reference state. The expression of the reactive power instantaneous pulsation deviation function is: , where is the reference reactive power level based on historical data, To adjust the coefficient, control the amplitude change of the reference power, The pulsation frequency of reactive power demand reflects the speed of load fluctuation. is the phase angle, indicating the initial phase of the reactive power fluctuation, is the instantaneous pulsation deviation of reactive power at time t;

[0068] By calculating the instantaneous pulsation deviation of reactive power, the instantaneous fluctuation intensity of reactive power can be captured, reflecting whether the load is in a high-frequency dynamic change state within the monitoring window.

[0069] The changing trend of the instantaneous pulsation deviation function is combined with the pulsation frequency to generate the reactive power demand pulsation index, which is used to quantitatively describe the comprehensive impact of the pulsation frequency and amplitude of reactive power. The generation expression of the reactive power demand pulsation index is:

[0070] ,

[0071] Where, represents the reactive power demand pulsation index, and Respectively represent the start and end time of the monitoring window, represents the instantaneous pulsation frequency of reactive power demand at time t, is the attenuation coefficient, which suppresses the influence of small pulsation amplitude on the final result;

[0072] By multiplying the pulsation deviation function by the instantaneous pulsation frequency and incorporating an attenuation coefficient, we can minimize the impact of small pulsations on the index. The resulting reactive power demand pulsation index accurately quantifies the frequency and amplitude variations of reactive power within the monitoring window, providing an important reference for predicting high-frequency dynamic loads.

[0073] The reactive power demand pulsation index (RPDI) indicates that a larger value, generated by analyzing the pulsation of reactive power demand within the monitoring window, indicates a higher frequency of dynamic load changes on the power transformer. This is because the RPDI quantifies the pulsation frequency and amplitude of reactive power demand within the monitoring window, reflecting dramatic load fluctuations over a short period of time. A high RPDI indicates frequent and large fluctuations in reactive power, suggesting frequent switching or starting and stopping of the transformer's load—a condition typically associated with high-frequency dynamic load changes. Conversely, a low RPDI indicates smaller and more stable fluctuations in reactive power demand, reflecting a more stable transformer load.

[0074] If the load frequency deviates significantly within the monitoring window and the frequency of the deviation is large, it typically indicates that the power transformer's load is experiencing dynamic, high-frequency changes. This is because frequency deviation reflects a frequent disruption of the power balance between the load and the grid in the system. Frequent load switching, the startup of high-power equipment, or transient changes in nonlinear loads can cause rapid fluctuations in system frequency. Frequent load frequency deviations indicate a rapid increase in power demand within a short period of time, requiring the transformer to continuously adjust its output to maintain grid stability. This high-frequency dynamic change not only increases the transformer's regulatory burden but also exacerbates mechanical wear on the equipment and power system instability.

[0075] The specific steps for generating a load frequency deviation index after analyzing the deviation degree and deviation frequency of the load frequency within the monitoring window are as follows:

[0076] Within the set monitoring window, the actual load frequency data of the power transformer and the reference frequency data of the power grid are collected. By comparing the difference between the actual load frequency and the reference frequency, the load frequency offset at each time point is calculated. The calculation expression of the load frequency offset is: , where is the actual load frequency at time t, in Hz, The base frequency of the power grid is usually 50Hz or 60Hz. is the load frequency offset at time t, reflecting the difference between the actual load and the reference frequency.

[0077] In each monitoring window, the load frequency offset amplitude and offset frequency are calculated to assess the severity of the load fluctuation. The offset amplitude refers to the maximum offset of the load frequency in the monitoring window. The calculation expression is: , where is the load frequency offset amplitude, indicating the maximum absolute value of the frequency offset within the monitoring window. Indicates the maximum value of all load frequency offsets within the monitoring window. The offset frequency is used to measure the load frequency fluctuation frequency, that is, the number of rapid fluctuations of the load frequency within the monitoring window. Each positive or negative change in the frequency offset is defined as a fluctuation. The offset frequency calculation formula is:

[0078] ,

[0079] Where, is the offset frequency of the load frequency, indicating how frequently the load frequency fluctuates within the monitoring window. T is the total time of the monitoring window, N is the number of sampling points in the monitoring window, is the Heaviside step function, which is used to determine whether the load frequency fluctuates. and Represents two consecutive moments, i Represents a time index;

[0080] If the load frequency changes positively or negatively between two consecutive moments, then The output is 1, otherwise the output is 0.

[0081] The load frequency offset index is generated based on the calculated load frequency offset amplitude and offset frequency. It comprehensively reflects the offset degree and fluctuation frequency of the load frequency within the monitoring window. The generation expression of the load frequency offset index is: , where is the load frequency deviation index, which quantifies the degree of deviation and frequency change of the load frequency. and are weight coefficients, which are used to adjust the impact of offset amplitude and offset frequency on the load frequency offset index. and is an exponential coefficient used to amplify the nonlinear effects of the offset amplitude and offset frequency.

[0082] The load frequency deviation index (LFDI), generated by analyzing the degree and frequency of load frequency deviation within the monitoring window, indicates that larger values indicate more frequent and drastic load dynamics on the power transformer, while smaller values indicate more stable loads. The LFDI comprehensively quantifies both the magnitude and frequency of frequency deviation. A high LFDI indicates significant and frequent load frequency deviations within the monitoring window, often associated with the connection of nonlinear loads, frequent startup of high-power equipment, or drastic changes in load demand. These high-frequency fluctuations place greater strain on the transformer's control system, increasing the risk of voltage instability and equipment wear. Conversely, a low LFDI indicates smaller and more stable load frequency fluctuations, indicating a relatively stable power transformer operation, which helps extend equipment life and ensure grid reliability.

[0083] The machine learning model is not specifically limited here, and can realize the reactive power demand pulsation index and load frequency deviation index Conduct comprehensive analysis to generate dynamic change coefficients In order to realize the technical solution of the present invention, the present invention provides a specific implementation method; the dynamic change coefficient The resulting calculation formula is:

[0084] ,

[0085] Where, 、 Reactive power demand pulsation index and load frequency deviation index The preset scaling factor of 、 Both are greater than 0.

[0086] It can be seen from the dynamic change coefficient that the greater the performance value of the reactive demand pulsation index generated after analyzing the pulsation of the reactive power demand within the monitoring window, the greater the performance value of the load frequency deviation index generated after analyzing the deviation degree and deviation frequency of the load frequency within the monitoring window. That is, the greater the performance value of the dynamic change coefficient generated when predicting the future dynamic changes in the load of the power transformer through the machine learning model, the higher the frequency of the dynamic changes in the load of the power transformer, and vice versa.

[0087] The load classification module classifies the load changes of power transformers within the monitoring window into two categories: high-frequency dynamic change type and conventional dynamic change type based on the prediction results of the machine learning model;

[0088] The dynamic change coefficient generated when predicting the future dynamic change of the power transformer load through the machine learning model is compared with the pre-set dynamic change coefficient reference threshold value to divide the load change of the power transformer within the monitoring window. The specific division steps are as follows:

[0089] If the dynamic change coefficient is greater than the dynamic change coefficient reference threshold, the load change of the power transformer under the monitoring window is classified as a high-frequency dynamic change type;

[0090] High-frequency dynamic change type: The load fluctuates frequently and violently in a short period of time, with a large amplitude, which is difficult to effectively control through conventional frequency adjustment.

[0091] If the dynamic change coefficient is less than or equal to the dynamic change coefficient reference threshold, the load change of the power transformer in the monitoring window is classified as a conventional dynamic change type;

[0092] Conventional dynamic change type: The load changes relatively smoothly, and the fluctuation amplitude and frequency are within the controllable range. It is suitable to use the initially set adjustment frequency for control.

[0093] The conventional load control module continues to maintain the initial automatic adjustment of the frequency control transformer winding tap position for load changes classified as conventional dynamic changes;

[0094] Because load changes are relatively smooth, the initial adjustment frequency is sufficient to maintain a stable secondary-side output voltage. This approach helps avoid frequent tap-changer operation, extending its service life and reducing maintenance costs, while ensuring stable operation of the power system and meeting users' normal electricity needs.

[0095] The high-frequency load adaptive control module adjusts the actual automatic adjustment frequency of the tap changer based on the characteristics of the current load change when the load changes are classified as high-frequency dynamic changes. Specifically, it increases the adjustment frequency of the tap changer based on the initial adjustment frequency so that the adjusted actual automatic adjustment frequency responds more quickly to the high-frequency dynamic changes of the load.

[0096] For load changes classified as high-frequency dynamic changes, adjust the actual automatic adjustment frequency of the tap changer according to the characteristics of the current load change. The specific steps are as follows:

[0097] In order to ensure accurate control, the initial control capability coefficient is introduced to describe the maximum dynamic change level that can be controlled by the initial adjustment frequency. The calculation expression of the initial control capability coefficient is:

[0098] ,

[0099] Where, is the initial adjustment frequency (number of adjustments per second), reflecting the control capability under normal circumstances. is the initial regulation capacity coefficient, which describes the load response capacity of the initial regulation frequency. is the fluctuation of the reference threshold value of the dynamic change coefficient, which is used to describe the deviation degree of the system load. , where Indicates the maximum dynamic change coefficient within the monitoring window, indicating the peak fluctuation of the load during the monitoring window. is the reference threshold value of the dynamic change coefficient;

[0100] This step is used to analyze the effectiveness of the initial frequency adjustment under the current circumstances and provide a benchmark for the next frequency increase.

[0101] According to the deviation between the dynamic change coefficient and the dynamic change coefficient reference threshold, the adjustment frequency boost coefficient is calculated, and the actual automatic adjustment frequency is calculated by the adjustment frequency boost coefficient. The calculation expression of the actual automatic adjustment frequency is: , where is the actual automatic adjustment frequency after adjustment, To adjust the frequency boost coefficient, determine the frequency boost range, adjust the frequency boost coefficient The calculation expression is:

[0102] ,

[0103] Where, is the dynamic change coefficient calculated within the monitoring window, It is the sensitivity coefficient, which is used to control the degree of frequency increase to avoid excessive frequency increase and cause equipment wear;

[0104] This step combines the current load change with the initial regulation capability by calculating the frequency increase coefficient to ensure that the new frequency increase matches the actual demand.

[0105] According to the calculated actual automatic adjustment frequency, the automatic adjustment frequency of the tap changer is adjusted in real time. At the same time, the frequency feedback coefficient is introduced to monitor the frequency effect after adjustment in actual operation and perform dynamic optimization. The calculation expression of the frequency feedback coefficient is:

[0106] ,

[0107] Where, is the frequency feedback coefficient, which is used to evaluate the control effect of the new frequency. is the actual automatic adjustment frequency at time t, is the dynamic change coefficient at time t;

[0108] The calculated frequency feedback coefficient Reference threshold range with frequency feedback coefficient Compare and analyze the frequency effect after adjustment. If it meets the requirements , it indicates that the actual automatic adjustment frequency after adjustment is within a reasonable range. The actual automatic adjustment frequency after adjustment can ensure that the frequency control of the tap changer is both efficient and will not cause excessive wear. If it does not meet , it indicates that the actual automatic adjustment frequency after adjustment does not meet the demand. The system will further optimize the actual automatic adjustment frequency to make it within a reasonable range.

[0109] By accelerating tap changer operation, timely adjusting the transformer winding tap position, stabilizing the secondary output voltage, and preventing overvoltage or undervoltage, this adaptive control strategy effectively addresses the inability of traditional fixed-frequency regulation to meet the demands of high-frequency load fluctuations, thereby improving the reliability and safety of the power system.

[0110] The present invention uses Internet of Things technology to collect load data in real time, and analyzes and predicts key features through machine learning models to accurately distinguish the dynamic change types of loads. For conventional dynamically changing loads, the initial adjustment frequency is maintained to avoid unnecessary frequent switching, thereby reducing mechanical wear and maintenance costs; and for high-frequency dynamically changing loads, the adjustment frequency of the tap changer is adaptively increased based on the prediction to ensure that the transformer responds quickly to load fluctuations in a short time, avoiding equipment damage and operational failures caused by voltage instability. This intelligent and adaptive control strategy not only ensures the efficient operation of power transformers in different load scenarios, but also reduces the safety risks caused by overvoltage or undervoltage, providing more reliable power supply guarantees for highly sensitive scenarios such as key industrial equipment and medical systems.

[0111] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0112] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0113] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

Claims

1. An intelligent monitoring and control system for power transformers, characterized in that: It includes initial automatic adjustment module, data acquisition and transmission module, data preprocessing and feature extraction module, feature analysis and prediction module, load classification module, conventional load control module and high-frequency load adaptive control module; Initial automatic adjustment module: In the initial stage, the tap changer is set to the initial automatic adjustment frequency to control the tap position of the transformer winding to ensure that the secondary side output voltage is stable within the predetermined range; The data acquisition and transmission module collects the load data of the power transformer in real time during operation and transmits the collected data to the central monitoring system through the Internet of Things technology; The data preprocessing and feature extraction module preprocesses the collected load change information and extracts key features that reflect the dynamic changes in the load of the power transformer; The feature analysis and prediction module analyzes the extracted features within the monitoring window and inputs the analyzed data into a pre-trained machine learning model. The machine learning model uses the input feature data to predict the future dynamic changes in the power transformer's load. The load classification module classifies the load changes of power transformers within the monitoring window into two categories: high-frequency dynamic change type and conventional dynamic change type based on the prediction results of the machine learning model; The conventional load control module continues to maintain the initial automatic adjustment of the frequency control transformer winding tap position for load changes classified as conventional dynamic changes; The high-frequency load adaptive control module adjusts the actual automatic adjustment frequency of the tap changer based on the characteristics of the current load change when the load changes are classified as high-frequency dynamic changes. Specifically, it increases the adjustment frequency of the tap changer based on the initial adjustment frequency so that the adjusted actual automatic adjustment frequency responds more quickly to the high-frequency dynamic changes of the load. Extracting key features reflecting the dynamic changes in the load of the power transformer, wherein the extracted key features include the pulsation of the reactive power demand within the monitoring window and the offset degree and offset frequency of the load frequency within the monitoring window. After analyzing the extracted features, a reactive power demand pulsation index and a load frequency offset index are generated respectively; The specific steps for generating a reactive power demand pulsation index after analyzing the pulsation of reactive power demand within the monitoring window are as follows: The time series data of reactive power demand Q(t) is collected in real time within the monitoring window, where Q(t) represents the reactive power demand at time t. Preprocess the reactive power data and use the preprocessed reactive power time series Q(t) as the basic input for subsequent calculations; Based on the extracted reactive power demand Q(t), the reactive power instantaneous pulsation deviation function is constructed to quantify the deviation between the time t and the reference state. The expression of the reactive power instantaneous pulsation deviation function is: ΔQ(t) = |Q(t)-Q ref (t)·(1+α·cos(ωt+φ))|, where, Q ref (t) is the reference reactive power level based on historical data, α is the adjustment coefficient, which controls the amplitude change of the reference power, ω is the pulsation frequency of the reactive demand, which reflects the speed of load fluctuation, φ is the phase angle, which indicates the initial phase of the reactive power fluctuation, and ΔQ(t) is the instantaneous pulsation deviation of the reactive power at time t; The changing trend of the instantaneous pulsation deviation function is combined with the pulsation frequency to generate the reactive power demand pulsation index, which is used to quantitatively describe the comprehensive impact of the pulsation frequency and amplitude of reactive power. The generation expression of the reactive power demand pulsation index is: Where R pulsation represents the reactive power demand pulsation index, t1 and t2 represent the start and end time of the monitoring window respectively, and f p (t) represents the instantaneous pulsation frequency of reactive power demand at time t, and β is the attenuation coefficient, which suppresses the influence of small pulsation amplitude on the final result.

2. The intelligent monitoring and control system for power transformers according to claim 1, characterized in that: The initial automatic frequency adjustment is based on the historical load data and control strategy of the power grid and is suitable for scenarios with normal load changes.

3. The intelligent monitoring and control system for power transformers according to claim 1, characterized in that: After obtaining the reactive demand pulsation index and load frequency offset index generated by analyzing the key features, the reactive demand pulsation index and load frequency offset index are input into a pre-trained machine learning model to generate a dynamic change coefficient, and the dynamic change coefficient is used to predict the future dynamic load changes of the power transformer.

4. The intelligent monitoring and control system for power transformers according to claim 3, characterized in that: The dynamic change coefficient generated when predicting the future dynamic change of the power transformer load through the machine learning model is compared with the pre-set dynamic change coefficient reference threshold value to divide the load change of the power transformer within the monitoring window. The specific division steps are as follows: If the dynamic change coefficient is greater than the dynamic change coefficient reference threshold, the load change of the power transformer under the monitoring window is classified as a high-frequency dynamic change type; If the dynamic change coefficient is less than or equal to the dynamic change coefficient reference threshold, the load change situation of the power transformer in the monitoring window is classified as a conventional dynamic change type.

5. The intelligent monitoring and control system for power transformers according to claim 1, characterized in that: The specific steps for generating a load frequency deviation index after analyzing the deviation degree and deviation frequency of the load frequency within the monitoring window are as follows: In the set monitoring window, the actual load frequency data of the power transformer and the reference frequency data of the power grid are collected. By comparing the difference between the actual load frequency and the reference frequency, the load frequency offset at each time point is calculated. The calculation expression of the load frequency offset is: Δf(t) = f load (t)-f base , where f load (t) is the actual load frequency at time t, f base is the reference frequency of the power grid, Δf(t) is the load frequency offset at time t, reflecting the difference between the actual load and the reference frequency; In each monitoring window, the load frequency offset amplitude and offset frequency are calculated to evaluate the severity of load fluctuation. The offset amplitude refers to the maximum offset of the load frequency in the monitoring window. The calculation expression is: freq =max(|Δf(t)|), where A freq is the load frequency offset amplitude, indicating the maximum absolute value of the frequency offset within the monitoring window, and max(|Δf(t)|) indicates the maximum value of all load frequency offsets within the monitoring window; The offset frequency is used to measure the fluctuation frequency of the load frequency, that is, the number of rapid fluctuations of the load frequency within the monitoring window. Each positive or negative change in the frequency offset is defined as a fluctuation. The offset frequency calculation formula is: Where, F shift is the offset frequency of the load frequency, indicating the frequency of load frequency fluctuations within the monitoring window. T is the total time of the monitoring window. N is the number of sampling points within the monitoring window. H(x) is the Heaviside step function used to determine whether the load frequency fluctuates. t i and t i+1 Represents two consecutive moments, i represents the time index; The load frequency offset index is generated based on the calculated load frequency offset amplitude and offset frequency. It comprehensively reflects the offset degree and fluctuation frequency of the load frequency within the monitoring window. The generation expression of the load frequency offset index is: Where, I LFD is the load frequency offset index, which quantifies the offset degree and frequency change of the load frequency. α′ and β′ are weight coefficients, which are used to adjust the influence of the offset amplitude and offset frequency on the load frequency offset index, respectively. γ1 and γ2 are exponential coefficients, which are used to amplify the nonlinear influence of the offset amplitude and offset frequency.

6. The intelligent monitoring and control system for power transformers according to claim 4, characterized in that: For load changes classified as high-frequency dynamic changes, adjust the actual automatic adjustment frequency of the tap changer according to the characteristics of the current load change. The specific steps are as follows: In order to ensure accurate control, the initial control capability coefficient is introduced to describe the maximum dynamic change level that can be controlled by the initial adjustment frequency. The calculation expression of the initial control capability coefficient is: Where, f init is the initial adjustment frequency, reflecting the regulation ability under normal circumstances, C init is the initial regulation capacity coefficient, which describes the load response capacity of the initial regulation frequency, ΔC threshold is the fluctuation of the reference threshold value of the dynamic change coefficient, which is used to describe the deviation degree of the system load. ΔC threshold =∣dynamic chan max -C threshold ∣ Where dynamic chan max Indicates the maximum dynamic change coefficient within the monitoring window, indicating the peak fluctuation of the load during the monitoring window, C threshold is the reference threshold value of the dynamic change coefficient; According to the deviation between the dynamic change coefficient and the dynamic change coefficient reference threshold, the adjustment frequency boost coefficient is calculated, and the actual automatic adjustment frequency is calculated by the adjustment frequency boost coefficient. The calculation expression of the actual automatic adjustment frequency is: f new =f init ·K boost , where f new is the actual automatic adjustment frequency after adjustment, K boost To adjust the frequency boost coefficient, determine the frequency increase range, and adjust the frequency boost coefficient K boost The calculation expression is: Where dynamic chan is the dynamic change coefficient calculated within the monitoring window, and γ is the sensitivity coefficient, which is used to control the degree of frequency increase.

7. The intelligent monitoring and control system for power transformers according to claim 6, characterized in that: According to the calculated actual automatic adjustment frequency, the automatic adjustment frequency of the tap changer is adjusted in real time. At the same time, the frequency feedback coefficient is introduced to monitor the frequency effect after adjustment in actual operation and perform dynamic optimization. The calculation expression of the frequency feedback coefficient is: Where K feedback is the frequency feedback coefficient, which is used to evaluate the control effect of the new frequency, f actual (t) is the actual automatic adjustment frequency at time t, dynamic chan (t) is the dynamic change coefficient at time t; The calculated frequency feedback coefficient K feedback Reference threshold range of frequency feedback coefficient [K feedback min , K feedback max ] to compare and analyze the frequency effect after adjustment. If K feedback ∈[K feedback min , K feedback max ], it indicates that the actual automatic adjustment frequency after adjustment is within a reasonable range. If it does not meet K feedback ∈[K feedback min , K feedback max ], it indicates that the actual automatic adjustment frequency after adjustment does not meet the demand, and the actual automatic adjustment frequency should be further optimized to make it within a reasonable range.

Images