Emergency power supply quick access method based on scene recognition

Abstract

The invention discloses an emergency power supply quick access method based on scene recognition, which relates to the technical field of emergency power supplies, and comprises the following steps: in an emergency power supply system, firstly obtaining electrical input parameter information through a preset detection frequency; by introducing a load change evaluation mechanism based on machine learning, the system can identify load fluctuation in real time and dynamically adjust the detection frequency, and the response speed to the rapid change of the load is improved. Compared with a traditional detection method, the load state is intelligently evaluated and accurately divided, it is ensured that the power factor is adjusted in time when the load fluctuates violently, and power loss and waste are reduced. Correction is carried out according to the traditional frequency during stable load fluctuation, the frequency is dynamically adjusted during rapid fluctuation, the electric energy use is optimized, the electric power quality is improved, and harmonic distortion and overload risks are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of emergency power supply, and in particular to a method for quickly accessing an emergency power supply based on scene recognition. Background Art

[0002] The rapid access of emergency power supply based on scene recognition means that through the intelligent scene recognition system, the current power demand and system status can be quickly identified when an emergency occurs in the power system, so as to accurately and quickly access the emergency power supply. This process usually relies on real-time data collection and analysis to identify load changes, frequency fluctuations or other abnormal conditions in the power grid, and switch to the emergency power supply in time to ensure the stability of power supply. The rapid access of emergency power supply requires adjusting the working state of the rectifier system so that the emergency power supply can be seamlessly connected with the power grid or load system through the rectifier. The rectifier plays a key role in this process, providing stable power output for the emergency power supply by converting alternating current (AC) into direct current (DC). The power regulation capability, response speed and load adaptability of the rectifier will directly affect the access speed and efficiency of the emergency power supply, ensuring that the system can restore power supply in the shortest time and prevent the negative impact of power outages.

[0003] The prior art has the following deficiencies:

[0004] During the emergency power supply access process, the rectifier uses active power factor correction (Active PFC) technology to improve power efficiency and reduce power loss. This technology usually relies on regular detection of input current and voltage, and real-time adjustment of the output control signal to ensure that the phase difference between current and voltage is minimized, thereby achieving an ideal power factor. However, when the load of the emergency power supply system changes rapidly, especially during the transition from no-load to full-load or when the load suddenly changes, the adjustment strategy based on regular detection of current and voltage may not respond in time, resulting in the power factor failing to maintain an ideal level.

[0005] This situation will lead to a series of serious consequences. First, when the load changes too quickly, the current waveform may become irregular, generating more harmonics, which not only increase reactive power (ineffective power), but may also cause the total harmonic distortion (THD) in the power grid to increase, further wasting energy. Second, when the output control signal cannot be accurately adjusted, the phase difference between current and voltage increases, the power factor decreases, thereby reducing energy transmission efficiency and increasing power loss.

[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 constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0007] The purpose of the present invention is to provide a method for rapid access to emergency power supply based on scene recognition. By introducing a load change evaluation mechanism based on a machine learning model, the system can identify load fluctuations in real time and dynamically adjust the detection frequency, thereby improving the response speed to rapid load changes. Compared with the traditional periodic detection method, intelligent evaluation can accurately divide the load state, ensure timely adjustment of the power factor when the load fluctuates violently, and reduce power loss and energy waste. For stable load fluctuations, the rectifier continues to correct the power factor at the traditional frequency; in the case of rapid fluctuations, the frequency is dynamically adjusted to reduce current irregularity and harmonic gain, optimize power use and improve power quality, reduce harmonic distortion and equipment overload risks, so as to solve the problems in the above-mentioned background technology.

[0008] In order to achieve the above object, the present invention provides the following technical solution: a method for rapid access to emergency power supply based on scene recognition, comprising the following steps:

[0009] In the emergency power supply system, the electrical input parameter information is first obtained through a preset detection frequency;

[0010] Establish an analysis set with the acquired electrical parameter information, and extract key features reflecting rapid load changes from the analysis set;

[0011] Perform detailed analysis on the extracted key features under the detection window, input the analyzed features into the pre-learned machine learning model, and use the machine learning model to intelligently evaluate the current load status;

[0012] The evaluation results based on the machine learning model divide the load status into two categories: rapid load fluctuation and stable load fluctuation;

[0013] For stable load fluctuations, the rectifier continues to collect electrical parameter information at the preset detection frequency and adjusts the power factor according to the existing power factor correction strategy;

[0014] For rapid load fluctuations, the detection frequency is dynamically adjusted based on the evaluation results of the machine learning model to ensure that the rectifier can correct the power factor in real time and accurately when the load changes drastically.

[0015] Preferably, key features reflecting rapid changes in load are extracted from the analysis set, and the extracted features include the rate at which the current frequency changes over time and the increase in the harmonic distortion in the voltage waveform. Within the detection window, the rate at which the current frequency changes over time and the increase in the harmonic distortion in the voltage waveform are analyzed to generate a current frequency change reference value and a voltage harmonic increase reference value, respectively. The current frequency change reference value reflects the degree of change in the current waveform frequency when the load changes rapidly by quantifying the rate at which the current frequency changes over time; the voltage harmonic increase reference value reflects the impact of rapid load changes on voltage quality by quantifying the increase in the harmonic distortion in the voltage waveform.

[0016] Preferably, after obtaining the current frequency change reference value and the voltage harmonic amplitude reference value generated after analyzing the key features, the current frequency change reference value and the voltage harmonic amplitude reference value are input into a pre-learned machine learning model, and a load change coefficient is generated through the machine learning model, and the current load state is intelligently evaluated through the load change coefficient.

[0017] Preferably, the load variation coefficient generated after analyzing the extracted key features is compared and analyzed with a preset load variation coefficient reference threshold, and the load state of the emergency power supply system is divided. The specific division steps are as follows:

[0018] If the load variation coefficient is greater than or equal to a preset load variation coefficient reference threshold, the load state of the emergency power supply system is classified as rapid load fluctuation;

[0019] If the load variation coefficient is less than a preset load variation coefficient reference threshold, the load state of the emergency power supply system is classified as stable load fluctuation.

[0020] Preferably, the detection frequency is dynamically adjusted based on the evaluation results of the machine learning model to ensure that the rectifier can correct the power factor data in real time and accurately when the load changes drastically. The specific steps are as follows:

[0021] According to the evaluation results of the machine learning model, the load fluctuation type is identified. If it is determined to be a rapid load fluctuation, that is, LVC ≥ LVC ref , the detection frequency is dynamically adjusted according to the preset detection frequency, and the detection frequency is increased. The calculation expression is as follows:

[0022] , where f new is the adjusted detection frequency, f preset is the preset detection frequency, LVC is the load variation coefficient, LVC ref is the load variation coefficient reference threshold;

[0024] When the load fluctuates violently, the detection frequency is increased. During this process, the active power factor correction system will use the updated current, voltage and power factor data to dynamically adjust the output control signal to ensure that the phase difference between current and voltage is minimized. The expression is as follows:

[0025] Power factor correction target = min(|cos(θ current -θ voltage )|)

[0026] , where θ current is the current phase angle, which refers to the phase difference between the current waveform and the reference sine wave, θ voltage is the voltage phase angle, which refers to the phase difference between the voltage waveform and the reference sine wave, cos(θ current -θ voltage ) is the cosine value of the phase difference between current and voltage, which directly reflects the degree of synchronization between current and voltage. The goal of power factor correction is to optimize the phase difference between current and voltage waveforms in the power supply system to ensure that the power factor is maintained at an ideal level;

[0027] Finally, the rectifier will adjust the frequency stably according to the real-time feedback after each detection frequency adjustment, and continue to adjust the detection frequency by periodically evaluating the phase difference change of current and voltage. The expression is as follows:

[0028]

[0029] , where f final is the final adjusted detection frequency, is the dynamic adjustment coefficient, Δφ voltage is the phase change of the voltage waveform, Δφ current is the phase change of the current waveform.

[0030] Preferably, in the detection window, the specific steps of analyzing the rate at which the current frequency changes over time and generating a reference value of the current frequency change are as follows:

[0031] In the detection window, the current signal of the emergency power supply system is collected and the instantaneous frequency is extracted. The analytical form of the signal is obtained through Hilbert transform. The calculation expression is as follows:

[0032]

[0033] , where f current (t) is the instantaneous frequency of the current signal, which indicates the rate of change of the current signal at time point t, θ current (t) is the instantaneous phase of the current signal;

[0034] Calculate the instantaneous frequency f of the current signal current(t) is the rate of change over time, and the calculation expression is as follows:

[0035] Δf current =max(f current (t1)-f current (t0)

[0036] , where Δf current is the instantaneous frequency change rate of the current signal, f current (t1) is the instantaneous frequency change rate of the current signal at the end time t1 of the detection window, f current (t0) is the instantaneous frequency change rate of the current signal at the start time t0 of the detection window, and max represents the maximum value of the frequency change;

[0037] In order to further quantify the instantaneous frequency change of the current signal, the instantaneous frequency change rate Δf of the current signal is calculated by the exponential fitting method. current For high-order feature extraction, the calculation expression is as follows:

[0038]

[0039] , where λ current is the exponential change coefficient of the instantaneous frequency change of the current signal, τ is the time constant, which controls the rate of exponential decay, and max represents the maximum value of the current frequency change rate after exponential processing within the detection window. It is an exponential function, which represents the process of time decay;

[0040] Based on the exponential variation coefficient λ of the instantaneous frequency variation of the extracted current signal current Generate the current frequency change reference value, the calculation expression is as follows:

[0041]

[0042] , where Φ current is the reference value of the current frequency change, and γ is a hyperparameter used to adjust the sensitivity and response speed of the reference value of the current frequency change.

[0043] Preferably, in the detection window, the specific steps of analyzing the harmonic distortion increase in the voltage waveform and generating the voltage harmonic increase reference value are as follows:

[0044] In the detection window, the real-time data of the voltage waveform is obtained from the monitoring system, and the noise is removed by preprocessing. Then, the voltage waveform in the time domain is converted into frequency domain data by Fourier transform in order to analyze the harmonic components. The calculation expression is as follows:

[0045]

[0046] , where V(ω) is the representation of the voltage waveform in the frequency domain, which is obtained by converting the voltage waveform in the time domain through Fourier transform, ω is the frequency variable, which represents the distribution of the voltage waveform in different frequency components, and A n is the amplitude of the nth harmonic, e is the natural base, j is the imaginary unit, ω n is the angular frequency of the nth harmonic, t is the time variable;

[0047] After the voltage waveform is converted into the frequency domain through Fourier transform, the next step is to identify the increase of the harmonic components. The calculation expression is as follows:

[0048]

[0049] , where is the total harmonic increase index, α is the exponential attenuation factor, and the total harmonic increase index is adjusted to reflect the nonlinear characteristics of the power system load change. N is the total number of harmonics, A1 is the fundamental amplitude in the voltage waveform, and β n It is the weight coefficient used to weight the nth harmonic amplification;

[0050] By analyzing the total harmonic increase index, calculating the harmonic increase value in each time period, defining a change coefficient to measure the rate of change of the increase, and capturing the dynamic characteristics of load changes, the calculation expression is as follows:

[0051]

[0052] , where C change (t) is the coefficient of variation, which is used to measure the rate of change of the total harmonic gain index over time. is the total harmonic gain index calculated at time point t, is the total harmonic increase index calculated at time point t-1;

[0053] Comprehensive total harmonic increase index and coefficient of variation C change (t), generate the voltage harmonic increase reference value, the calculation expression is as follows:

[0054]

[0055] , where THD voltage It is the voltage harmonic increase reference value.

[0056] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0057] The present invention introduces a load change assessment mechanism based on a machine learning model, which can more accurately identify the load state and dynamically adjust the detection frequency, especially when the load fluctuates rapidly. The traditional periodic detection method cannot respond quickly when the load changes suddenly, but this method can judge the amplitude of the load fluctuation in real time based on the current frequency change reference value and the voltage harmonic amplitude reference value, thereby adjusting the detection frequency. This not only improves the system's adaptability to load fluctuations, but also ensures that the rectifier can still maintain an ideal power factor when the load changes drastically, reduces power loss and unnecessary energy waste, and ensures the efficient operation of the system under extreme load conditions.

[0058] The present invention accurately divides the load state according to the intelligent evaluation of the load change coefficient, and adopts different power factor correction strategies for stable load fluctuations and rapid load fluctuations, which can ensure that the rectifier continues to perform power factor correction according to the traditional detection frequency when the load changes are not drastic, thereby maintaining the optimization of power energy utilization efficiency. When the load changes drastically, by dynamically adjusting the detection frequency, the rectifier can respond to load changes more promptly, reduce the irregularity of the current waveform and the harmonic gain, thereby avoiding a sharp drop in power factor and reducing harmonic distortion in the power grid. This flexible adjustment strategy not only ensures efficient use of electricity, but also greatly improves the power quality of the power supply system, reducing power loss and equipment overload risks caused by unstable power factor. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced 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.

[0060] Figure 1 The present invention is a method flow chart of a method for rapid access to an emergency power supply based on scene recognition.

[0061] Figure 2-Figure 4 It is a physical picture of the emergency power supply of the present invention. DETAILED DESCRIPTION

[0062] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of 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 the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0063] The present invention provides a method for quickly accessing an emergency power supply based on scene recognition as shown in the figure, comprising the following steps:

[0064] In the emergency power supply system, the electrical input parameter information is first obtained through a preset detection frequency;

[0065] The preset detection frequency refers to the frequency used to regularly collect and monitor electrical input parameters (such as current, voltage, frequency, etc.) in the emergency power supply system at a pre-set time interval based on system requirements and working environment. This frequency determines how often the system collects data in each time period. The preset detection frequency is selected based on factors such as the speed of load fluctuations, the responsiveness of the power supply system, and control accuracy. The purpose is to ensure that the system can capture changes in electrical parameters in a timely manner, especially when the load changes, and can respond quickly. A higher detection frequency helps to track real-time changes in power input parameters more accurately, especially when there are large load fluctuations or sudden load changes, so that the power factor correction strategy can be adjusted quickly to maintain the stable operation of the power supply system.

[0066] Establish an analysis set with the acquired electrical parameter information, and extract key features reflecting rapid load changes from the analysis set;

[0067] The first step to establish an analysis set of the acquired electrical parameter information is to organize and store the regularly collected electrical input data (such as current, voltage, power factor, frequency, etc.) in chronological order. These data should have timestamps so that the trend of each parameter over time can be tracked. The frequency and duration of data collection also need to be set according to system requirements. For example, you can choose to collect data once a second and group the data by hour or minute to ensure that sufficient detailed information is obtained at the critical moment of load change.

[0068] When establishing the analysis set, these electrical parameters also need to be preprocessed to ensure the integrity and accuracy of the data. The preprocessing steps may include removing noise, filling missing data, standardization or normalization, etc. to ensure data consistency. Then, the processed data set is classified and labeled according to the different stages of load change, so that relevant features can be extracted according to the specific load fluctuation type in subsequent analysis. These features will provide support for the next step of load change analysis and the input of the machine learning model.

[0069] Perform detailed analysis on the extracted key features under the detection window, input the analyzed features into the pre-learned machine learning model, and use the machine learning model to intelligently evaluate the current load status;

[0070] Key features reflecting rapid load changes are extracted from the analysis set. The extracted features include the rate at which the current frequency changes over time and the increase in the harmonic distortion in the voltage waveform. Under the detection window, the rate at which the current frequency changes over time and the increase in the harmonic distortion in the voltage waveform are analyzed to generate a current frequency change reference value and a voltage harmonic increase reference value, respectively. The current frequency change reference value reflects the degree of change of the current waveform frequency when the load changes rapidly by quantifying the rate at which the current frequency changes over time; the voltage harmonic increase reference value reflects the impact of rapid load changes on voltage quality by quantifying the increase in the harmonic distortion in the voltage waveform.

[0071] After obtaining the current frequency change reference value and the voltage harmonic increase reference value generated after analyzing the key features, the current frequency change reference value and the voltage harmonic increase reference value are input into the pre-learned machine learning model, and the load change coefficient is generated by the machine learning model, and the current load state is intelligently evaluated by the load change coefficient;

[0072] A pre-learned machine learning model refers to a machine learning algorithm model that is trained and optimized using historical data in an emergency power supply system. The model establishes a mapping relationship between input features and system load states by learning a large amount of input data (such as current frequency change reference values ​​and voltage harmonic amplitude reference values). The model training process usually involves the use of historical load change data, corresponding electrical parameters (such as current, voltage, harmonic distortion, etc.), and the output results of these data under different load conditions. In this way, the machine learning model can automatically identify the complex relationship between load fluctuations and electrical characteristics, thereby having the ability to predict and intelligently evaluate future load state changes. During the training process, the model continuously adjusts its internal parameters to minimize prediction errors, thereby forming a tool that can efficiently evaluate load states under new load change scenarios.

[0073] Once the machine learning model is trained, it can evaluate the current load status in real time by receiving new current frequency change reference values ​​and voltage harmonic gain reference values ​​as input. In this process, the model will derive a load change coefficient through an algorithm based on the learned relationships and patterns, combined with the input feature information. This coefficient can characterize the degree of load change, speed, and current load stability. The generation of the load change coefficient not only depends on basic parameters such as current and voltage, but also takes into account the rate and fluctuation characteristics of these parameters over time, so it can more accurately reflect the actual dynamic changes of the system load. This intelligent evaluation capability enables the system to respond quickly to load fluctuations and automatically adjust strategies such as power factor correction, thereby ensuring the efficient and stable operation of the emergency power supply system.

[0074] Sudden changes in the rate at which the current frequency changes over time usually indicate that the load of the power system is changing rapidly. Rapid changes in load, especially when the load changes from no load to full load, or in situations where the load changes suddenly, can cause the frequency of the current waveform to change significantly. This is because load changes affect the current output characteristics of the power system, which in turn affects the periodicity and frequency of the current. For example, when the load suddenly increases, the power supply must quickly supply more current, and the period of the current may shorten, resulting in an increase in the current frequency; when the load drops sharply, the current output by the power supply decreases, the period lengthens, and the current frequency decreases accordingly. This sudden change in frequency is often not a linear change, but shows obvious acceleration or deceleration characteristics, so the speed of the load change can be detected by the rate of change of the current frequency. When the rate of change of the current frequency changes suddenly, it usually means that the load change is sudden, and the power system needs to quickly adjust its output to cope with this sudden change. Therefore, the sudden change in the current frequency provides a very sensitive indicator that can reflect the violent fluctuations in the load state and is an effective basis for judging whether the power system is experiencing rapid load changes.

[0075] In the detection window, the specific steps for analyzing the rate at which the current frequency changes over time and generating a reference value for the current frequency change are as follows:

[0076] In the detection window, the current signal of the emergency power supply system is collected and the instantaneous frequency is extracted. The analytical form of the signal is obtained through Hilbert transform. The calculation expression is as follows:

[0077]

[0078] , where f current (t) is the instantaneous frequency of the current signal, which indicates the rate of change of the current signal at time point t. The instantaneous frequency is determined by the change of the phase of the current signal over time. current (t) is the instantaneous phase of the current signal. Phase is an important characteristic of the signal, which describes the "position" of the periodic fluctuation of the signal, that is, the progress of the signal in one cycle;

[0079] The purpose of obtaining the analytical form of the signal through Hilbert transform is mainly to extract the instantaneous phase and instantaneous frequency of the signal, so as to analyze the dynamic changes of the signal more accurately.

[0080] Calculate the instantaneous frequency f of the current signal current (t) is the rate of change over time, and the calculation expression is as follows:

[0081] Δf current =max(f current (t1)-f current (t0)

[0082] , where Δf current is the instantaneous frequency change rate of the current signal, f current (t1) is the instantaneous frequency change rate of the current signal at the end time t1 of the detection window, f current (t0) is the instantaneous frequency change rate of the current signal at the start time t0 of the detection window, and max means taking the maximum value of the frequency change. Since the current frequency may fluctuate over time within the detection window, this operation selects the maximum value of the frequency change, that is, the most significant frequency change within this time period;

[0083] In order to further quantify the instantaneous frequency change of the current signal, the instantaneous frequency change rate Δf of the current signal is calculated by the exponential fitting method. current For high-order feature extraction, the calculation expression is as follows:

[0084]

[0085] , where λ current is the exponential change coefficient of the instantaneous frequency change of the current signal, τ is the time constant, which controls the rate of exponential decay, and max represents the maximum value of the current frequency change rate after exponential processing within the detection window. It is an exponential function, which represents the process of time decay;

[0086] The role of the time constant τ in the exponential decay formula is to control the response time or sensitivity of the current frequency change. It determines the weighted decay rate of the current frequency change. Specifically, the time constant v controls how long the frequency change affects the system. A smaller τ value means that the system is more sensitive to rapid changes in current frequency and can respond quickly to sudden changes in the load; a larger τ value means that the system responds more slowly and pays more attention to changes over a longer period of time. When the τ value is small, the system will pay more attention to recent frequency changes because the decay effect of the time constant reduces the impact of frequency changes in the distant time on the final result. By adjusting the time constant τ, the system's sensitivity to load fluctuations can be flexibly controlled to ensure that adjustments can be made in time when the load changes rapidly, thereby optimizing the stability and efficiency of the power supply system.

[0087] The role of the exponential function is to perform weighted attenuation processing on changes in the time series to highlight the impact of recent changes while suppressing the impact of long-term changes. In the detection of load fluctuations, the exponential function is used to calculate the current frequency change rate (such as Δf current) can amplify the frequency changes at the most recent moment, ensuring that the system can respond more sensitively to sudden fluctuations in load changes. Specifically, as time goes by, the value of the exponential function gradually decreases, making the contribution of frequency changes at earlier and more distant time points to the results weaken, thereby focusing on the critical moment of the current load change. This weighted method helps to quickly identify the moment when the load fluctuates violently, especially in the case of sudden or rapid load changes, so that the system can adjust its response strategy more effectively.

[0088] Based on the exponential variation coefficient λ of the instantaneous frequency variation of the extracted current signal current Generate the current frequency change reference value, the calculation expression is as follows:

[0089]

[0090] , where Φ current is the reference value of the current frequency change, γ is a hyperparameter used to adjust the sensitivity and response speed of the reference value of the current frequency change;

[0091] The hyperparameter γ is used to adjust the sensitivity of the current frequency change reference value to load fluctuations. It controls the response speed and sensitivity of the exponential change. Specifically, a larger γ value makes the exponent more sensitive and can respond to sudden changes in the load more quickly; while a smaller γ value makes the response smoother and less sensitive to changes in load fluctuations. By adjusting the hyperparameters, the system can balance the sensitivity and smoothness of the response according to actual needs, thereby optimizing the monitoring and adjustment strategy for load fluctuations.

[0092] Under the detection window, the larger the current frequency change reference value performance value generated after analyzing the rate at which the current frequency changes over time, the greater the load of the power system is experiencing rapid changes. The current frequency change reference value reflects the rate at which the current frequency changes over time. When the load fluctuates violently, the frequency change of the current tends to increase rapidly and significantly. The rapid fluctuation of the load leads to rapid changes in the current cycle, which in turn causes the current frequency to change suddenly, showing a large change index. Therefore, when the performance value of the current frequency change reference value increases, it means that the current frequency has experienced large fluctuations in a short period of time, which directly reflects the rapid change of the load. On the contrary, if the current frequency change reference value is small, it means that the current frequency fluctuates less, the load changes more smoothly, and the system is in a stable change state.

[0093] The exponential growth of harmonic distortion in the voltage waveform usually indicates that the load of the power system is changing rapidly. Rapid fluctuations in the load, especially sudden increases or decreases, will cause significant changes in the shape of the current waveform, which in turn affects the voltage waveform of the power system. When the load changes drastically, especially in the case of nonlinear loads or instantaneous load switching, the current waveform will change from the original smooth sine waveform to an irregular waveform, generating more high-frequency harmonic components. Harmonic distortion is the deviation between the voltage waveform and the ideal sine wave, which is usually caused by mismatched load characteristics or unstable load changes. When the load changes rapidly, the harmonic components of the voltage will increase rapidly, showing an exponential growth, reflecting that the power system cannot adapt to such load changes in a short period of time, resulting in a significant decrease in voltage quality. In particular, after the load changes suddenly, the power system fails to adjust the control strategy in time to stabilize the output, and the harmonic distortion index rises rapidly, indicating that the speed and amplitude of the load change of the power system exceeds the normal coping ability of the system. Therefore, the exponential growth of the voltage harmonic increase is a clear indication of rapid load changes, which means that the power system needs to respond and adjust more quickly to maintain voltage quality and system stability.

[0094] In the detection window, the harmonic distortion increase in the voltage waveform is analyzed and the specific steps for generating the voltage harmonic increase reference value are as follows:

[0095] In the detection window, the real-time data of the voltage waveform is obtained from the monitoring system, and the noise is removed by preprocessing. Then, the voltage waveform in the time domain is converted into frequency domain data by Fourier transform in order to analyze the harmonic components. The calculation expression is as follows:

[0096]

[0097] , where V(ω) is the representation of the voltage waveform in the frequency domain, which is obtained by converting the voltage waveform in the time domain through Fourier transform, ω is the frequency variable, which represents the distribution of the voltage waveform in different frequency components, and A n is the amplitude of the nth harmonic, e is the natural base, j is the imaginary unit, ω n is the angular frequency of the nth harmonic, t is the time variable;

[0098] Convert the voltage waveform from the time domain to the frequency domain in order to analyze the various frequency components in the voltage waveform, especially the harmonic components. Through Fourier transform, the time domain signal is decomposed into a series of sine waves of different frequencies, which are expressed as the amplitude and phase of the fundamental wave and higher harmonics. This conversion helps to reveal the harmonic distortion caused by load fluctuations in the power supply system, and then provide frequency domain data for further harmonic gain analysis. By analyzing the harmonic amplitude A n and its corresponding angular frequency ω n, it can be determined whether the power supply system is affected by rapid changes in load, especially the increase in high-order harmonics, which can reflect the stability of the system and changes in voltage quality when the load fluctuates.

[0099] After the voltage waveform is converted into the frequency domain through Fourier transform, the next step is to identify the increase of the harmonic components. The calculation expression is as follows:

[0100]

[0101] , where is the total harmonic increase index, α is the exponential attenuation factor, and the total harmonic increase index is adjusted to reflect the nonlinear characteristics of the power system load change. N is the total number of harmonics, A1 is the fundamental amplitude in the voltage waveform, and β n It is the weight coefficient used to weight the nth harmonic amplification;

[0102] By analyzing the total harmonic increase index, calculating the harmonic increase value in each time period, defining a change coefficient to measure the rate of change of the increase, and capturing the dynamic characteristics of load changes, the calculation expression is as follows:

[0103]

[0104] , where C change (t) is the coefficient of variation, which is used to measure the rate of change of the total harmonic gain index over time. is the total harmonic gain index calculated at time point t, is the total harmonic increase index calculated at time point t-1;

[0105] Coefficient of variation C change The function of (t) is to measure the rate at which the total harmonic increase index changes over time, thereby reflecting the dynamic characteristics of the load changes in the power system. By comparing the total harmonic increase index at the current moment with the total harmonic increase index at the previous moment, the coefficient of variation can quantify the degree of change in the harmonic increase. If the coefficient of variation is large, it means that the load changes rapidly, and the increase in harmonics in the power system also increases rapidly, which usually indicates that the load has experienced rapid fluctuations or mutations. On the contrary, if the coefficient of variation is small, it means that the load changes relatively smoothly and the harmonic increase changes relatively slowly. This coefficient provides the system with a real-time load fluctuation indicator, helping to identify the severity of load changes in a timely manner and take corresponding adjustment measures.

[0106] Comprehensive total harmonic increase index and coefficient of variation C change (t), generate the voltage harmonic increase reference value, the calculation expression is as follows:

[0107]

[0108] , where THD voltage It is the voltage harmonic increase reference value;

[0109] Under the detection window, the larger the voltage harmonic increase reference value performance value generated after analyzing the harmonic distortion increase in the voltage waveform, it usually means that the load of the power supply system is changing rapidly. Harmonic distortion in the voltage waveform mainly comes from the nonlinear characteristics of the load and load fluctuations. When the load changes rapidly, especially when the load increases or decreases suddenly, the current and voltage waveforms will change dramatically, generating more harmonic components. The larger the harmonic increase index, the more significant the increase in the harmonic components in the voltage waveform, and the system fails to adapt to this rapid change in time, resulting in a decrease in power quality. On the contrary, if the voltage harmonic increase reference value is small, it means that the voltage waveform remains stable and the harmonic distortion increase is small, which usually indicates that the load change is in a relatively stable state and the system can cope with the load change better.

[0110] The machine learning model is not limited here and can achieve the current frequency change reference value Φ current And voltage harmonic increase reference value THD voltage Any machine learning model that performs comprehensive analysis to generate the load variation coefficient LVC can be used. In order to implement the technical solution of the present invention, the present invention provides a specific implementation method;

[0111] The load variation coefficient LVC is generated by the following formula:

[0112]

[0113] , where f1 and f2 are the reference values ​​of current frequency change Φ current And voltage harmonic increase reference value THD voltage The preset proportional coefficient, and f1 and f2 are both greater than 0.

[0114] It can be seen from the calculation expression of the load variation coefficient that, under the detection window, the larger the current frequency variation reference value generated after analyzing the rate of change of the current frequency over time, and the larger the voltage harmonic increase reference value generated after analyzing the harmonic distortion increase in the voltage waveform, the larger the load variation coefficient value generated after analyzing the extracted key features under the detection window, indicating that the more drastic the load change of the emergency power supply system, the lower the power transmission efficiency, the lower the power factor, or the waste of electric energy. The smaller the load variation coefficient value, the smaller the current frequency change and the voltage harmonic increase, the more stable the load fluctuation, the stable state of the emergency power supply system, the higher the efficiency of the power system, the more stable the response, and the lower the power loss.

[0115] In this formula, the preset proportional coefficients f1 and f2 refer to the current frequency change reference value Φ in the process of calculating the load change coefficient LVC. current And voltage harmonic increase reference value THD voltage Constant coefficients set according to the degree of influence on the final result. These proportional coefficients are set in advance according to the system design and optimization goals to ensure that the different characteristics of current frequency and voltage harmonics have appropriate weights when comprehensively evaluating load changes. Specifically, f1 gives the influence of the current frequency change reference value, while f2 reflects the influence of the voltage harmonic increase reference value. By adjusting these proportional coefficients, the contribution of both to the load change coefficient can be flexibly adjusted to achieve the best load assessment effect. In practical applications, these coefficients usually need to be adjusted according to the actual load behavior and the response characteristics of the system to ensure that the calculated load change coefficient can accurately reflect the actual situation of load fluctuations.

[0116] The evaluation results based on the machine learning model divide the load status into two categories: rapid load fluctuation and stable load fluctuation;

[0117] The load variation coefficient generated after analyzing the extracted key features is compared with the preset load variation coefficient reference threshold value to divide the load state of the emergency power supply system. The specific division steps are as follows:

[0118] If the load variation coefficient is greater than or equal to a preset load variation coefficient reference threshold, the load state of the emergency power supply system is classified as rapid load fluctuation;

[0119] If the load variation coefficient is less than a preset load variation coefficient reference threshold, the load state of the emergency power supply system is classified as stable load fluctuation;

[0120] Rapid load fluctuation refers to a dramatic and sudden change in the load of the power system in a relatively short period of time, which is usually manifested as a sharp increase or decrease in the load; stable load fluctuation refers to a slow and continuous change in the load over a longer period of time. This change is usually relatively smooth, with a relatively small fluctuation amplitude and a slow rate of change, so the power system can adapt and respond more easily.

[0121] For stable load fluctuations, the rectifier continues to collect electrical parameter information at the preset detection frequency and adjusts the power factor according to the existing power factor correction strategy;

[0122] For stable load fluctuations, the rectifier continues to collect electrical parameter information at the preset detection frequency and adjusts the power factor according to the existing power factor correction strategy. The main purpose is to ensure that the power system can operate efficiently and stably when the load changes are small. When the load fluctuation is small, the phase difference between current and voltage usually does not fluctuate drastically, so maintaining the current detection frequency is sufficient to monitor the system status in real time and adjust the power factor correction strategy in time. At this time, the rectifier's PFC system can accurately synchronize current and voltage to ensure that the power factor is maintained at an ideal level, avoiding energy waste and power loss caused by a decrease in power factor. This strategy can not only ensure system efficiency, but also avoid unnecessary energy consumption and equipment burden caused by over-adjustment.

[0123] For rapid load fluctuations, the detection frequency is dynamically adjusted based on the evaluation results of the machine learning model to ensure that the rectifier can correct the power factor in real time and accurately when the load changes drastically;

[0124] Dynamically adjust the detection frequency based on the evaluation results of the machine learning model to ensure that the rectifier can accurately correct the power factor data in real time when the load changes drastically. The specific steps are as follows:

[0125] According to the evaluation results of the machine learning model, the load fluctuation type is identified. If it is determined to be a rapid load fluctuation, that is, LVC ≥ LVC ref , the detection frequency is dynamically adjusted according to the preset detection frequency, and the detection frequency is increased. The calculation expression is as follows:

[0126]

[0127] , where f new is the adjusted detection frequency, f preset is the preset detection frequency, LVC is the load variation coefficient, LVC ref is the load variation coefficient reference threshold;

[0128] When the load fluctuates violently, the active power factor correction system of the rectifier updates the electrical parameters more frequently by increasing the detection frequency, so that the power factor can be corrected in real time. In this process, the active power factor correction system will use the updated current, voltage and power factor data to dynamically adjust the output control signal to ensure that the phase difference between current and voltage is minimized. At this time, the phase difference between the current waveform and the voltage waveform can be accurately aligned, and the power factor will be maintained at an ideal level, minimizing power loss and reactive power. The expression is as follows:

[0129] Power factor correction target = min(|cos(θ current -θ voltage )|)

[0130] , where θ current is the current phase angle, which refers to the phase difference between the current waveform and the reference sine wave, θ voltage is the voltage phase angle, which refers to the phase difference between the voltage waveform and the reference sine wave, cos(θ current -θ voltage ) is the cosine value of the phase difference between current and voltage, which directly reflects the degree of synchronization between current and voltage. The goal of power factor correction is to optimize the phase difference between current and voltage waveforms in the power supply system to ensure that the power factor is maintained at an ideal level;

[0131] Finally, after each frequency adjustment, the rectifier will adjust the frequency stably according to the real-time feedback. By periodically evaluating the phase difference changes of current and voltage, the detection frequency will continue to be adjusted to ensure that the response speed and accuracy of the electrical parameters can match the speed of load changes when the load fluctuates greatly. The expression is as follows:

[0132]

[0133] , where f final is the final adjusted detection frequency, is the dynamic adjustment coefficient, Δφ voltage It is the phase change of the voltage waveform, indicating the deviation of the voltage phase angle relative to the reference sine wave, Δφ current It is the phase change of the current waveform, indicating the offset of the current phase angle relative to the reference sine wave;

[0134] In emergency power supply systems, rapid load fluctuations may cause drastic changes in load status, which in turn affects the phase difference between current and voltage and, in turn, the power factor. When the load increases or decreases suddenly, the traditional power factor correction system may not be able to respond in time, resulting in an increase in the phase difference between voltage and current and a decrease in power factor. This will not only reduce the efficiency of power use, but may also cause problems such as power loss, harmonic distortion, and grid instability. To meet this challenge, based on the evaluation results of the machine learning model, the system uses a strategy of increasing the detection frequency to ensure that the rectifier can correct the power factor in a timely and accurate manner when the load changes rapidly.

[0135] Specifically, when the load changes dramatically, the system needs to collect electrical parameters (such as current and voltage) more frequently in order to more accurately analyze the working status of the power supply system. The system's detection frequency can be adjusted dynamically by performing real-time evaluation of load fluctuations through machine learning models. By increasing the detection frequency, the system can capture changes in electrical parameters in a shorter period of time and adjust the power factor correction strategy in a timely manner. In this way, the system can continuously optimize the synchronization between current and voltage during load fluctuations and reduce the phase difference, thereby ensuring that the power factor is close to the ideal value (close to 1), reducing reactive power, and improving energy transmission efficiency.

[0136] Improve the system's response speed and accuracy to load changes and prevent the negative impact of load fluctuations. By adjusting the detection frequency in real time, the system can respond to load changes more flexibly, avoid a drop in power factor, and ensure that the power supply system can operate stably and efficiently even in emergency situations, ultimately reducing power loss, reducing the burden on the power grid, and improving the overall operating efficiency and reliability of the system.

[0137] The model of the emergency power supply of the present invention is AHXT-PRS3000.

[0138] The present invention introduces a load change assessment mechanism based on a machine learning model, which can more accurately identify the load state and dynamically adjust the detection frequency, especially when the load fluctuates rapidly. The traditional periodic detection method cannot respond quickly when the load changes suddenly, but this method can judge the amplitude of the load fluctuation in real time based on the current frequency change reference value and the voltage harmonic amplitude reference value, thereby adjusting the detection frequency. This not only improves the system's adaptability to load fluctuations, but also ensures that the rectifier can still maintain an ideal power factor when the load changes drastically, reduces power loss and unnecessary energy waste, and ensures the efficient operation of the system under extreme load conditions.

[0139] The present invention accurately divides the load state according to the intelligent evaluation of the load change coefficient, and adopts different power factor correction strategies for stable load fluctuations and rapid load fluctuations, so as to ensure that the rectifier continues to perform power factor correction according to the traditional detection frequency when the load changes are not drastic, thereby maintaining the optimization of power energy utilization efficiency. When the load changes drastically, by dynamically adjusting the detection frequency, the rectifier can respond to load changes more promptly, reduce the irregularity of the current waveform and the harmonic increase, thereby avoiding a sharp drop in power factor and reducing harmonic distortion (THD) in the power grid. This flexible adjustment strategy not only ensures efficient use of electricity, but also greatly improves the power quality of the power supply system, reduces power loss and equipment overload risks caused by unstable power factor.

[0140] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for rapid access to emergency power supply based on scene recognition, characterized in that: The following steps are involved: In the emergency power supply system, the electrical input parameter information is first obtained through a preset detection frequency; Establish an analysis set with the acquired electrical parameter information, and extract key features reflecting rapid load changes from the analysis set; Perform detailed analysis on the extracted key features under the detection window, input the analyzed features into the pre-learned machine learning model, and use the machine learning model to intelligently evaluate the current load status; The evaluation results based on the machine learning model divide the load status into two categories: rapid load fluctuation and stable load fluctuation; For stable load fluctuations, the rectifier continues to collect electrical parameter information at the preset detection frequency and adjusts the power factor according to the existing power factor correction strategy; For rapid load fluctuations, the detection frequency is dynamically adjusted based on the evaluation results of the machine learning model to ensure that the rectifier can correct the power factor in real time and accurately when the load changes drastically.

2. According to the method of rapid access to emergency power supply based on scene recognition according to claim 1, it is characterized in that: Key features reflecting rapid load changes are extracted from the analysis set. The extracted features include the rate at which the current frequency changes over time and the increase in the harmonic distortion in the voltage waveform. Under the detection window, the rate at which the current frequency changes over time and the increase in the harmonic distortion in the voltage waveform are analyzed to generate a current frequency change reference value and a voltage harmonic increase reference value, respectively. The current frequency change reference value reflects the degree of change of the current waveform frequency when the load changes rapidly by quantifying the rate at which the current frequency changes over time; the voltage harmonic increase reference value reflects the impact of rapid load changes on voltage quality by quantifying the increase in the harmonic distortion in the voltage waveform.

3. According to the method of rapid access to emergency power supply based on scene recognition according to claim 2, it is characterized in that: After obtaining the current frequency change reference value and the voltage harmonic amplitude reference value generated after analyzing the key features, the current frequency change reference value and the voltage harmonic amplitude reference value are input into a pre-learned machine learning model, and a load change coefficient is generated through the machine learning model. The load change coefficient is used to intelligently evaluate the current load state.

4. According to the method of rapid access to emergency power supply based on scene recognition according to claim 3, it is characterized in that: The load variation coefficient generated after analyzing the extracted key features is compared with the preset load variation coefficient reference threshold value to divide the load state of the emergency power supply system. The specific division steps are as follows: If the load variation coefficient is greater than or equal to a preset load variation coefficient reference threshold, the load state of the emergency power supply system is classified as rapid load fluctuation; If the load variation coefficient is less than a preset load variation coefficient reference threshold, the load state of the emergency power supply system is classified as stable load fluctuation.

5. According to the method of rapid access to emergency power supply based on scene recognition according to claim 4, it is characterized in that: Dynamically adjust the detection frequency based on the evaluation results of the machine learning model to ensure that the rectifier can accurately correct the power factor data in real time when the load changes drastically. The specific steps are as follows: According to the evaluation results of the machine learning model, the load fluctuation type is identified. If it is determined to be a rapid load fluctuation, that is, LVC ≥ LVC ref , the detection frequency is dynamically adjusted according to the preset detection frequency, and the detection frequency is increased. The calculation expression is as follows: , where f new is the adjusted detection frequency, f preset is the preset detection frequency, LVC is the load variation coefficient, LVC ref is the load variation coefficient reference threshold; When the load fluctuates violently, the detection frequency is increased. During this process, the active power factor correction system will use the updated current, voltage and power factor data to dynamically adjust the output control signal to ensure that the phase difference between current and voltage is minimized. The expression is as follows: Power factor correction target = min(|cos(θ current -θ voltage )|), where θ current is the current phase angle, which refers to the phase difference between the current waveform and the reference sine wave, θ voltage is the voltage phase angle, which refers to the phase difference between the voltage waveform and the reference sine wave, cos(θ current -θ voltage ) is the cosine value of the phase difference between current and voltage, which directly reflects the degree of synchronization between current and voltage. The goal of power factor correction is to optimize the phase difference between current and voltage waveforms in the power supply system to ensure that the power factor is maintained at an ideal level; Finally, the rectifier will adjust the frequency stably according to the real-time feedback after each detection frequency adjustment, and continue to adjust the detection frequency by periodically evaluating the phase difference changes of current and voltage. The expression is as follows: , where f final is the final adjusted detection frequency, is the dynamic adjustment coefficient, Δφ voltage is the phase change of the voltage waveform, Δφ current is the phase change of the current waveform.

6. The method for rapid access to emergency power supply based on scene recognition according to claim 2 is characterized in that: In the detection window, the specific steps for analyzing the rate at which the current frequency changes over time and generating a reference value for the current frequency change are as follows: In the detection window, the current signal of the emergency power supply system is collected and the instantaneous frequency is extracted. The analytical form of the signal is obtained through Hilbert transform. The calculation expression is as follows: , where f current (t) is the instantaneous frequency of the current signal, which indicates the rate of change of the current signal at time point t, θ current (t) is the instantaneous phase of the current signal; Calculate the instantaneous frequency f of the current signal current (t) is the rate of change over time, and the calculation expression is as follows: Δf current =max(f current (t1)-f current (t0)), where Δf current is the instantaneous frequency change rate of the current signal, f current (t1) is the instantaneous frequency change rate of the current signal at the end time t1 of the detection window, f current (t0) is the instantaneous frequency change rate of the current signal at the start time t0 of the detection window, and max represents the maximum value of the frequency change; In order to further quantify the instantaneous frequency change of the current signal, the instantaneous frequency change rate Δf of the current signal is calculated by the exponential fitting method. current For high-order feature extraction, the calculation expression is as follows: , where λ current is the exponential change coefficient of the instantaneous frequency change of the current signal, τ is the time constant, which controls the rate of exponential decay, and max represents the maximum value of the current frequency change rate after exponential processing within the detection window. It is an exponential function, which represents the process of time decay; Based on the exponential variation coefficient λ of the instantaneous frequency variation of the extracted current signal current Generate the current frequency change reference value, the calculation expression is as follows: , where Φ current is the reference value of the current frequency change, and γ is a hyperparameter used to adjust the sensitivity and response speed of the reference value of the current frequency change.

7. The method for rapid access to emergency power supply based on scene recognition according to claim 2 is characterized in that: In the detection window, the harmonic distortion increase in the voltage waveform is analyzed and the specific steps for generating the voltage harmonic increase reference value are as follows: In the detection window, the real-time data of the voltage waveform is obtained from the monitoring system, and the noise is removed by preprocessing. Then, the voltage waveform in the time domain is converted into frequency domain data by Fourier transform in order to analyze the harmonic components. The calculation expression is as follows: , where V(ω) is the representation of the voltage waveform in the frequency domain, which is obtained by converting the voltage waveform in the time domain through Fourier transform, ω is the frequency variable, which represents the distribution of the voltage waveform in different frequency components, and A n is the amplitude of the nth harmonic, e is the natural base, j is the imaginary unit, ω n is the angular frequency of the nth harmonic, t is the time variable; After the voltage waveform is converted into the frequency domain through Fourier transform, the next step is to identify the increase of the harmonic components. The calculation expression is as follows: , where is the total harmonic increase index, α is the exponential attenuation factor, and the total harmonic increase index is adjusted to reflect the nonlinear characteristics of the power system load change. N is the total number of harmonics, A1 is the fundamental amplitude in the voltage waveform, and β n It is the weight coefficient used to weight the nth harmonic amplification; By analyzing the total harmonic increase index, calculating the harmonic increase value in each time period, defining a change coefficient to measure the rate of change of the increase, and capturing the dynamic characteristics of load changes, the calculation expression is as follows: , where C change (t) is the coefficient of variation, which is used to measure the rate of change of the total harmonic gain index over time. is the total harmonic gain index calculated at time point t, is the total harmonic increase index calculated at time point t-1; Comprehensive total harmonic increase index and coefficient of variation C change (t), generate the voltage harmonic increase reference value, the calculation expression is as follows: , where THD voltage It is the voltage harmonic increase reference value.

Images