Intelligent monitoring and switching control system and method for backup power supply based on internet of things

Abstract

The application discloses an intelligent monitoring and switching control system and method for a backup power supply based on an Internet of Things, and relates to the technical field of intelligent switching of backup power supplies, and comprises the following steps: setting a best reference switching time threshold and determining an initial safe switching period in combination with historical operation data of backup power supply switching, load characteristics and technical specifications.The application realizes a key leap from "timed switching" to "state-aware driving switching", enables the system to comprehensively perceive the phase synchronization state between the main and backup power supplies before switching, and dynamically predicts and intelligently judges the phase jump trend in combination with a machine learning model, thereby effectively avoiding the impact current, equipment damage and regional trip risk caused by forced switching due to inconsistent phases, and significantly improving the safety, accuracy and intelligent level of the backup power supply switching process.

Description

Technical Field

[0001] This invention relates to the field of intelligent backup power switching technology, and more specifically to an intelligent monitoring and switching control system and method for backup power based on the Internet of Things. Background Technology

[0002] Intelligent monitoring and switching of backup power based on the Internet of Things (IoT) is an intelligent solution that utilizes IoT technologies (such as sensors, communication modules, and cloud platforms) to monitor and automatically control backup power sources (such as UPS, battery packs, and generators) in a power system in real time. This system deploys sensors for voltage, current, and temperature to collect real-time operating status data of both the main and backup power sources. This data is then uploaded to a cloud or local control platform via wireless or wired networks, enabling visualized management of power status, anomaly warnings, and fault diagnosis. When the main power source fails or experiences voltage anomalies, the system can automatically detect and quickly and seamlessly switch to the backup power source, ensuring continuous and stable power supply to critical equipment and systems, and significantly improving the reliability, intelligence, and operational efficiency of the power supply system.

[0003] The existing technology has the following shortcomings:

[0004] In existing intelligent backup power switching technologies, fixed switching time thresholds are typically used to perform the switching operation. However, if a phase transition still exists between the backup power supply and the active power supply just before the switching time is about to expire, forcibly closing the circuit breaker according to the preset time limit is essentially equivalent to instantaneously connecting two power supplies with different voltage phases. This instantaneous operation can trigger an inrush current several times or even tens of times higher than the rated value, far exceeding the withstand capacity of the power equipment, and can easily lead to circuit breaker damage, contactor welding, or cable overheating. More seriously, the huge current will trigger the tripping mechanism through the grounding protection system, directly causing a power outage in the entire power supply area, resulting in serious operational interruptions and equipment damage risks.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide an IoT-based intelligent monitoring and switching control system and method for backup power supplies. This system achieves a key leap from "timed switching" to "state-aware driving switching," enabling the system to fully perceive the phase synchronization status between the primary and backup power supplies before switching. It also combines machine learning models to dynamically predict and intelligently judge phase transition trends, thereby effectively avoiding the risks of inrush current, equipment damage, and regional tripping caused by forced switching due to phase inconsistency. This significantly improves the safety, accuracy, and intelligence of the backup power supply switching process. While ensuring power supply continuity, it provides higher power operation guarantee capabilities for critical load scenarios, thus solving the problems mentioned in the background technology.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an intelligent monitoring and switching control method for backup power supply based on the Internet of Things, comprising the following steps:

[0008] Based on historical operating data, load characteristics, and technical specifications of backup power switching, an optimal baseline switching time threshold is set to define the initial safe switching period.

[0009] During the baseline waiting period for backup power switching, a high-precision synchronous sampling unit is used to perform real-time, high-frequency synchronous sampling of the voltage waveforms of the backup power and the active power, and the real-time phase angle of the main and backup power voltage waveforms is extracted in real time through digital signal processing technology.

[0010] Based on the collected real-time phase data, key indicators that can reflect the phase jump characteristics between the main and backup power supplies are extracted. In-depth feature engineering processing is carried out on the key indicators to construct a set of features that characterize the degree of phase jump.

[0011] The constructed feature set is transformed into feature vectors and input into a pre-trained and put into use machine learning model to accurately judge and dynamically predict the real-time phase jump state and its evolution trend during the backup power switching.

[0012] The risk level of phase transition is assessed in real time based on the output transition prediction results. The safety level of phase synchronization between the current primary and backup power supplies is automatically identified, and the reference switching time threshold is adaptively delayed according to the risk level of phase transition.

[0013] Preferably, the real-time phase angle of the main and backup power supply voltage waveforms is extracted in real time using Fast Fourier Transform. The specific steps are as follows:

[0014] S1. Preprocess the sampled data and perform window function processing to reduce spectral leakage;

[0015] S2. Perform a fast Fourier transform to convert the time-domain signal to the frequency domain, find the amplitude peak near the detected main frequency, and read its phase.

[0016] S3. Using the read phase as a real-time reference point, calculate the phase angle between the main power supply and the backup power supply;

[0017] S4. Continuously repeat S1-S3 and periodically refresh the phase results to achieve real-time monitoring of the phase of the main and backup power supply voltage waveforms.

[0018] Preferably, signal processing technology is used to extract key indicators that can reflect the phase jump characteristics between the main and backup power supplies. The extracted indicators include the cumulative value of the phase angle jump frequency and amplitude per unit time and the skewness of the phase difference change. Within the monitoring window, the extracted indicators are subjected to in-depth feature engineering processing to generate phase jump reference values ​​and phase fluctuation asymmetry reference values, respectively. A set of features characterizing the degree of phase jump is constructed through the phase jump reference values ​​and phase fluctuation asymmetry reference values.

[0019] Preferably, the feature set constructed from the phase change reference value and the phase fluctuation asymmetry reference value is transformed into a feature vector and input into a pre-trained and put into use machine learning model. The jump coefficient is generated based on the machine learning model, and the real-time phase jump state and its evolution trend during the backup power switching are accurately judged and dynamically predicted through the jump coefficient.

[0020] Preferably, the system automatically identifies the safety level of phase synchronization between the primary and backup power supplies and adaptively delays the reference switching time threshold based on the risk level of phase transition. The specific steps are as follows:

[0021] During the real-time switching process of the backup power supply, the machine learning model outputs the current transition coefficient, which is compared with a set transition coefficient reference threshold to assess the phase synchronization risk level between the primary and backup power supplies. A risk level function is introduced for quantification, and the expression of the risk level function is as follows:

[0022]

[0023] ,in: The jump coefficients are the output of the model. The reference threshold for the jump coefficient set for the system. This is the phase transition risk level coefficient, reflecting the intensity ratio of the current transition level relative to the preset safety threshold;

[0024] Based on phase jump risk level coefficient Calculate an adaptive delay factor to adjust the baseline switching time threshold. The expression for the calculation is:

[0025]

[0026] ,in: The adaptive delay factor determines the adjustment factor for the actual switching time. This is the maximum delay gain coefficient, which controls the upper limit of the delay. The jump response sensitivity coefficient controls the steepness of the response of the delay function.

[0027] Preferably, the adaptive delay factor is applied to the original reference switching time threshold. To obtain the adaptive delay switching threshold actually used to control the switching action, the expression is:

[0028]

[0029] ,in: The delay switching time threshold is ultimately used to control the execution. The preset baseline switching time, This is the maximum allowable switching delay time.

[0030] Preferably, within the monitoring window, the specific steps for generating phase change reference values ​​through in-depth feature engineering processing of the accumulated values ​​of phase angle jump frequency and amplitude per unit time are as follows:

[0031] Construct a jump energy function to capture the joint strength of frequency and amplitude;

[0032] The phase jump energy is normalized and a sequential perturbation spread rate is introduced to generate a phase jump reference value.

[0033] Preferably, the specific steps for generating a phase fluctuation asymmetry reference value by performing in-depth feature engineering processing on the degree of skewness of phase difference changes within the monitoring window are as follows:

[0034] Analyze the phase difference curves of the main and backup power supplies within the monitoring window, identify all amplitude extreme points, record the phase difference value corresponding to each extreme point, and form an amplitude extreme point sequence in sequence.

[0035] For all extreme points, distinguish between positive and negative extreme points, and then classify them into sets of positive amplitude extreme points respectively. and negative amplitude extreme value set ;

[0036] For the positive amplitude extreme value set and negative amplitude extreme value set Normalization is performed, and the normalized reference value is used to calculate the relative offset of each extreme point. Then, the phase fluctuation asymmetry reference value is calculated based on the ratio of the relative offsets.

[0037] The IoT-based intelligent monitoring and switching control system for backup power includes a baseline switching strategy configuration module, a real-time phase acquisition and analysis module, a phase jump feature construction module, a phase jump prediction and identification module, and an adaptive switching duration adjustment module.

[0038] The baseline switching strategy configuration module, by combining historical operating data, load characteristics and technical specifications of backup power switching, sets an optimal baseline switching time threshold and clarifies the initial safe switching period.

[0039] The real-time phase acquisition and analysis module uses a high-precision synchronous sampling unit to perform real-time, high-frequency synchronous sampling of the voltage waveforms of the backup power supply and the active power supply during the baseline waiting period of the backup power supply switching execution. It also extracts the real-time phase angle of the main and backup power supply voltage waveforms in real time through digital signal processing technology.

[0040] The phase jump feature construction module extracts key indicators that reflect the phase jump characteristics between the main and backup power supplies based on the collected real-time phase data. It then performs in-depth feature engineering processing on the key indicators to construct a set of features that characterize the degree of phase jump.

[0041] The phase jump prediction and recognition module transforms the constructed feature set into feature vectors, which are then input into a pre-trained and put into use machine learning model to accurately judge and dynamically predict the real-time phase jump state and its evolution trend during backup power switching.

[0042] The adaptive switching time control module assesses the risk level of phase transitions in real time based on the output transition prediction results, automatically identifies the safety level of phase synchronization between the current primary and backup power supplies, and adaptively delays the reference switching time threshold according to the risk level of phase transitions.

[0043] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0044] This invention achieves a key leap from "timed switching" to "state-aware driven switching," enabling the system to fully perceive the phase synchronization status between the primary and backup power supplies before switching. It also combines machine learning models to dynamically predict and intelligently judge phase transition trends, thereby effectively avoiding the risks of inrush current, equipment damage, and regional tripping caused by forced switching due to phase inconsistency. This significantly improves the safety, accuracy, and intelligence of the backup power supply switching process, ensuring power supply continuity while providing higher power operation guarantee capabilities for critical load scenarios. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0046] Figure 1 This is a flowchart of the intelligent monitoring and switching control method for backup power based on the Internet of Things of the present invention.

[0047] Figure 2 This is a schematic diagram of the modules of the Internet of Things-based intelligent monitoring and switching control system for backup power.

[0048] Figure 3 This is a framework diagram of the present invention. Detailed Implementation

[0049] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0050] This invention provides, for example Figure 1 The IoT-based intelligent monitoring and switching control method for backup power supply, as shown, includes the following steps:

[0051] Based on historical operating data, load characteristics, and technical specifications of backup power switching, an optimal baseline switching time threshold is set to define the initial safe switching period.

[0052] Historical operational data refers to a series of key parameters and status information recorded and accumulated by the backup power system during its past operation. This data typically includes the frequency and duration of main power failures, backup power startup time, actual successful switchover time, voltage and frequency trends, phase synchronization status, records of switchover failures, and equipment response delays. Analyzing this data can identify typical time points and risk factors that lead to problems during backup power switchover, thus providing a data basis for rationally setting baseline switchover time thresholds and avoiding incorrect switchovers or delayed power supply caused by setting them too early or too late.

[0053] Load characteristics refer to the operating features of the electrical equipment or systems that a backup power supply needs to support. These characteristics include load type (such as resistive, inductive, and capacitive), load starting current, power fluctuation range, sensitivity to voltage and frequency, and tolerance time for power supply continuity. For example, lighting systems have relatively lenient requirements for switching time, but data center servers, medical equipment, or motor drive systems are extremely sensitive to power supply fluctuations and cannot tolerate even millisecond-level power outages. Therefore, setting a reasonable switching time threshold must fully consider the load type of the currently connected system and its tolerance for switching timing to ensure stable power supply to the system without interference.

[0054] Technical specifications refer to national standards, industry standards, or technical guidance documents provided by equipment manufacturers regarding the configuration, operation, safety, and switching control of backup power systems. Typical specifications include ATS response time ranges, power switching safety thresholds, generator start-up time requirements, phase synchronization standards, voltage and frequency tolerance ranges, and protection mechanism setting requirements. For example, a certain type of UPS system requires switching to be completed within 10ms, while a diesel generator may require 15 seconds to start and stabilize. Therefore, the setting of the baseline switching time should follow relevant technical specifications to ensure that the system design conditions and equipment performance boundaries are met, achieving optimal switching efficiency while ensuring safety.

[0055] The baseline switching time threshold can be determined based on the type of backup power supply (such as a diesel generator, UPS, or energy storage system) and the type of load it supplies. The switching time threshold specifies how long the backup power supply should initially wait before performing a switching action after a main power supply failure or anomaly, ensuring the backup power supply reaches a stable output state and providing a basic time window to ensure the backup power supply reaches a stable frequency, voltage, and phase. Clearly defining the initial "safe switching period" ensures stable operation of the backup power supply and avoids secondary failures caused by blind switching due to the backup power supply starting too quickly or not reaching a steady state.

[0056] During the baseline waiting period for backup power switching, a high-precision synchronous sampling unit is used to perform real-time, high-frequency (2kHz~10kHz recommended) synchronous sampling of the voltage waveforms of the backup power supply and the active power supply. The real-time phase angle (phase information) of the main and backup power supply voltage waveforms is extracted in real time through digital signal processing technology (such as Fast Fourier Transform FFT, phase-sensitive detection or phase-locked loop PLL algorithm).

[0057] The real-time phase angle of the main and backup power supply voltage waveforms is extracted in real time using Fast Fourier Transform (FFT). The specific steps are as follows:

[0058] Preprocessing the sampled data and applying window functions (such as the Hanning window) can reduce spectral leakage.

[0059] Perform a fast Fourier transform to convert the time-domain signal to the frequency domain, find the amplitude peak near the detected main frequency, and read its phase.

[0060] Using this phase as a real-time reference point, the phase angle between the main power supply and the backup power supply is calculated.

[0061] The above process is repeated continuously and the phase results are refreshed periodically to achieve real-time monitoring of the phase of the main and backup power supply voltage waveforms.

[0062] These steps allow for the accurate acquisition of the phase relationship between the primary and backup power supplies, providing crucial data support for assessing phase transitions and implementing synchronous switching.

[0063] Based on the collected real-time phase data, key indicators that can reflect the phase jump characteristics between the main and backup power supplies are extracted. In-depth feature engineering processing is carried out on the key indicators to construct a set of features that characterize the degree of phase jump, providing accurate data support for subsequent model prediction and intelligent judgment.

[0064] Signal processing techniques are used to extract key indicators that reflect the phase jump characteristics between the main and backup power supplies. The extracted indicators include the cumulative value of the phase angle jump frequency and amplitude per unit time and the skewness of the phase difference change. Within the monitoring window, the extracted indicators are subjected to in-depth feature engineering processing to generate phase jump reference values ​​and phase fluctuation asymmetry reference values. A feature set characterizing the degree of phase jump is constructed through the phase jump reference values ​​and phase fluctuation asymmetry reference values.

[0065] During the baseline waiting period for backup power switching, an increase in the frequency and cumulative magnitude of phase angle jumps per unit time usually indicates a significant phase jump phenomenon between the primary and backup power supplies. This suggests frequent and drastic inconsistencies in the voltage waveforms of the two power supplies within a short period, potentially due to factors such as frequency instability, generator output not reaching steady state, control system delays, or external interference. Sustained phase jumps not only indicate that the primary and backup power supplies are not yet synchronized but may also trigger inrush currents and electrical risks during switching. Therefore, monitoring the frequency and magnitude of phase angle jumps per unit time serves as a crucial basis for determining whether a "safe switching window" has been entered.

[0066] Within the monitoring window, the specific steps for generating phase change reference values ​​through in-depth feature engineering processing of the accumulated values ​​of phase angle jump frequency and amplitude per unit time are as follows:

[0067] A jump energy function is constructed to capture the joint strength of frequency and amplitude. The expression for the jump energy function is as follows:

[0068]

[0069] ,in: For the first i The amplitude of the instantaneous change in phase angle between the primary and backup power supplies is calculated from the difference in continuous phase samples. This is the phase jump amplification factor (recommended to be set between 1.5 and 2.5), used to amplify the impact weight of large phase angle jumps. The marker factor for the occurrence of the transition event is defined as follows: If the value exceeds a set small threshold (e.g., 0.01 radians), the value is 1; otherwise, it is 0. N To monitor the total number of phase change events within the monitoring window, This represents the phase transition energy function, used to measure the overall intensity of phase transitions occurring between the primary and backup power supplies within the monitoring window;

[0070] This step involves constructing a nonlinear weighted cumulative jump function. The frequency of the jump (by (determined) and jump amplitude (by) The decision is integrated together, which amplifies the impact of large jumps and effectively avoids the interference of noise or slight disturbances on the overall characteristics.

[0071] The phase jump energy is normalized and a sequential perturbation spread rate is introduced to generate a phase jump reference value. The expression for generating the phase jump reference value is:

[0072]

[0073] ,in: This serves as a reference value for phase abrupt change. and These are the maximum and minimum phase angle values ​​within the monitoring window, used to normalize the overall phase fluctuation amplitude and prevent overall system offset from affecting the judgment results. For very small values ​​(e.g.) To prevent the denominator from being zero, It is the maximum length of a continuous transition segment (i.e., the maximum sequence length of consecutive transition events). The number of transition segments (i.e., how many consecutive transition segments exist);

[0074] This step involves jumping energy. The data is normalized and then multiplied by an expansion factor that considers the "complexity of the phase jump structure." This expansion term measures the "coherence" and "concentration" of the phase jump; if the phase jump exhibits a long-term, continuous, and highly concentrated pattern, the phase jump reference value will increase significantly. This design allows the phase jump reference value to not only reflect the intensity of the jump but also to detect whether it constitutes a "systemic volatility risk."

[0075] As can be seen from the phase jump reference value, within the monitoring window, the larger the performance value of the phase jump reference value generated by in-depth feature engineering processing of the accumulated value of the phase angle jump frequency and amplitude per unit time, the greater the performance value. The phase jump reference value, as a comprehensive quantitative indicator of the phase jump behavior between primary and backup power supplies, is based on the fusion of the jump frequency and jump amplitude per unit time through a nonlinear weighting method, and incorporates extended analysis of jump continuity and disturbance structure. A larger phase jump reference value indicates more frequent and more drastic changes in the phase angle of the primary and backup power supplies within the monitoring window, and may even show a continuous disturbance trend, reflecting a relatively serious phase mismatch state and a high switching risk. Conversely, a smaller phase jump reference value indicates fewer phase changes or smaller amplitude changes, good synchronization between primary and backup power supplies, slight phase jumps, and conditions for safe switching. Therefore, the level of the phase jump reference value can serve as an important basis for intelligently judging the switching timing and the reliability of the synchronization state.

[0076] When the phase difference between the primary and backup power supplies shows a significant skew, especially if the phase difference curves are asymmetrically offset in both directions and the deviation from the center becomes increasingly severe, it usually indicates that the phase synchronization of the two power supplies is deteriorating, reflecting the presence of a phase jump. For example, if the primary power supply phase consistently leads or lags the backup power supply phase, and the skewness continues to increase, it suggests that the primary and backup power supplies have not reached a balanced state and are experiencing a significant phase jump. This increasing skewness can reveal potential instability in the power system.

[0077] Within the monitoring window, the specific steps for generating a phase fluctuation asymmetry reference value through in-depth feature engineering processing of the skewness of phase difference changes are as follows:

[0078] First, analyze the phase difference curves of the primary and backup power supplies within the monitoring window to identify all amplitude extrema (i.e., local maximum and minimum values). These extrema reflect the peak fluctuation state of the phase difference. Record the phase difference value corresponding to each extrema, forming a sequence of amplitude extrema.

[0079] Among these extreme points, we distinguish between positive extremes (i.e., peak values) and negative extremes (i.e., trough values), and then classify them into sets of positive amplitude extremes. and negative amplitude extreme value set ;

[0080] This step aims to identify the main fluctuation amplitude of the phase difference change by extracting the extreme points. The separation of positive and negative extreme points provides a clear reference benchmark for subsequent asymmetry calculations, ensuring that the analysis process focuses more on the actual peak behavior of the jump fluctuations.

[0081] For the positive amplitude extreme value set and negative amplitude extreme value set After normalization, the relative offset of each extreme point is calculated using the normalized reference value. Then, the phase fluctuation asymmetry reference value is calculated based on the ratio of the relative offsets. The expression for calculating the phase fluctuation asymmetry reference value is as follows:

[0082]

[0083] ,in: Indicates the asymmetric reference value for phase fluctuation. This represents the average intensity of the phase difference shifting in the positive direction. This represents the average intensity of the phase difference shifting in the negative direction. This represents a normalized reference value, used to standardize the extreme amplitude of the phase difference so that the feature values ​​extracted under different time periods, different power systems, or different load conditions are comparable and stable.

[0084] The phase fluctuation asymmetry reference value directly characterizes the relative weights of the distribution of positive and negative amplitude extrema. When the phase jump is severe, the distribution of positive and negative amplitude extrema often exhibits obvious asymmetry. By calculating the phase fluctuation asymmetry reference value, the degree of directional shift of the fluctuation can be quantified, avoiding the smoothing effect of traditional statistics and more sensitively capturing the abrupt characteristics of phase jumps.

[0085] As can be seen from the phase fluctuation asymmetry reference value, within the monitoring window, the larger the value of the phase fluctuation asymmetry reference value generated after in-depth feature engineering processing of the phase difference change skewness, the more severe the phase jump between the primary and backup power supplies. The magnitude of the phase fluctuation asymmetry reference value clearly reflects the severity of the phase jump between the primary and backup power supplies. When the phase fluctuation asymmetry reference value is large, it indicates that within the monitoring window, the phase difference fluctuation of the primary and backup power supplies is significantly biased in a specific direction (such as continuously biased in the positive or negative direction), indicating poor synchronization and severe phase jump, and a possible significant risk of synchronization instability. Conversely, when the value of the phase fluctuation asymmetry reference value is close to 1, that is, the positive and negative fluctuations are relatively symmetrical, it indicates that the phase fluctuation of the primary and backup power supplies is small, the synchronization state is relatively stable, and the phase jump phenomenon is slight. Therefore, the magnitude of the phase fluctuation asymmetry reference value can be used to intuitively assess the risk level of phase jump between the primary and backup power supplies.

[0086] The constructed feature set is transformed into feature vectors and input into a pre-trained and put into use machine learning model to accurately judge and dynamically predict the real-time phase jump state and its evolution trend during backup power switching, providing a reliable basis for intelligent switching decisions.

[0087] The feature set constructed from the phase change reference value and the phase fluctuation asymmetry reference value is transformed into a feature vector, which is then input into a pre-trained and put into use machine learning model. The jump coefficient is generated based on the machine learning model, and the real-time phase jump state and its evolution trend during the backup power switching are accurately judged and dynamically predicted through the jump coefficient.

[0088] A pre-trained and deployed machine learning model refers to a high-quality, high-precision, robust, and reliable machine learning model that has undergone rigorous data collection, data cleaning, feature extraction, feature engineering, model training, validation testing, and parameter tuning before being deployed to a backup power switching system for real-time operation. First, "pre-trained" means that before actual application, the machine learning model has undergone systematic and comprehensive supervised or semi-supervised training using historical operating data or simulation data, effectively capturing and identifying various characteristics and trends of phase transitions between primary and backup power supplies. To obtain such an efficient model, a series of rigorous steps are required: for example, the initial collection of a large amount of historical data, including stable phase data during normal operation and data on severe phase transitions under abnormal conditions; secondly, meticulous data cleaning and preprocessing are needed to remove outliers, fill in missing values, and reasonably adjust the data distribution to ensure the overall quality of the training data. Subsequently, the previously constructed data is used... A high-quality feature set composed of phase jump reference values ​​and phase fluctuation asymmetry reference values ​​is used to further optimize the expressiveness of the features through feature engineering methods. This fully reveals the deep correlation between phase jumps and these features, providing a high-quality, high-discrimination data foundation for model input. After feature preparation, a suitable machine learning algorithm (such as Long Short-Term Memory (LSTM) network, gradient boosting decision trees like XGBoost, or random forests in deep learning) is selected for preliminary model structure design and parameter initialization. Subsequently, the feature data is input into the model, and repeated iterative training and parameter tuning (including hyperparameter tuning, cross-validation, and model optimization) are performed to achieve high generalization ability and prediction accuracy.

[0089] The deployment signifies that this machine learning model has moved beyond the laboratory environment and is now formally deployed in the field of intelligent backup power switching systems, entering the real-time operation phase and becoming an indispensable part of the intelligent backup power switching control strategy. This deployment implies that the model must possess extremely high reliability, real-time performance, and robustness, capable of continuously and stably performing real-time data processing and predictive analysis in a real-world production environment, rapidly outputting jump coefficients for accurate judgment and dynamic prediction of real-time phase jump states and trends. Specifically, when the model is deployed, it receives real-time phase data from the primary and backup power supplies obtained through high-frequency, high-precision synchronous sampling, and rapidly calculates feature vectors online based on previously constructed features (such as phase change reference values ​​and phase fluctuation asymmetry reference values). These feature vectors are then input into the model. The pre-trained model immediately outputs jump coefficients (which can be understood as a scoring index quantifying the severity of phase jumps) based on the feature vectors, thereby completing accurate classification, evaluation, and trend prediction of the real-time phase state within milliseconds to seconds, providing real-time guidance for subsequent decisions and delay adjustment mechanisms in the intelligent backup power switching strategy. This real-time predictive capability is a crucial consideration during the deployment phase. The model must withstand the test of real-world, complex operating conditions, such as voltage fluctuations, frequency disturbances, load changes, and ambient temperature variations, rather than being limited to ideal experimental data and controlled conditions. After long-term real-time application, the model must also possess online self-learning and retraining capabilities, automatically or semi-automatically adjusting its weights and parameters based on new real-time operating data to continuously improve its predictive accuracy and adaptability, truly meeting the technical requirements of intelligent real-time dynamic switching of backup power supplies. A pre-trained and deployed machine learning model not only signifies that the model has undergone rigorous and thorough offline training, verification, and optimization, but also emphasizes that the model has been successfully deployed in an actual backup power supply real-time operating system, possessing the ability to stably process field data in real time, quickly and accurately determine the degree of phase jumps and dynamic trends, and generate jump coefficients in real time, truly meeting the stringent requirements of backup power supply intelligent switching control strategies in industrial application scenarios for model reliability, real-time performance, and accuracy.

[0090] The machine learning model is not specifically limited here, but it can achieve the function of converting phase change reference values. and phase fluctuation asymmetric reference value Perform comprehensive analysis to generate jump coefficients Any machine learning model is acceptable. To achieve the technical solution of this invention, this invention provides a specific implementation method; jump coefficient. The generated expression is: In the formula, , These are the phase abrupt change reference values. and phase fluctuation asymmetric reference value The preset proportional coefficient, and , All are greater than 0. The preset proportional coefficient refers to the parameter used in the formula for calculating the jump coefficient to adjust the weight of different indicators on the final result, i.e. and .in, It is the weighting coefficient of the phase change reference value. These are the weighting coefficients for the phase fluctuation asymmetry reference value. By pre-setting the values ​​of these two coefficients, their contribution ratio in the overall jump coefficient can be manually adjusted according to specific scenarios or empirical knowledge. For example, if the system is more concerned with the risks brought about by transient changes, the weighting coefficients can be adjusted accordingly. Set it larger; if the system pays more attention to the anomalies in fluctuation trends, it can be increased. This "preset" refers to a set of constants set before training or deploying a machine learning model, based on actual needs, historical data statistics, or engineering experience. These constants are used to keep the model structure simple and controllable, while providing a certain degree of flexibility and interpretability.

[0091] As can be seen from the jump coefficient, within the monitoring window, the larger the performance value of the phase change reference value generated by in-depth feature engineering processing of the accumulated value of the phase angle jump frequency and amplitude per unit time, the larger the performance value of the phase fluctuation asymmetry reference value generated after in-depth feature engineering processing of the skewness of the phase difference change, that is, the larger the performance value of the jump coefficient generated when the machine learning model dynamically predicts the real-time phase jump state during the backup power switching, the more serious the phase jump between the main and backup power supplies, and vice versa.

[0092] Based on the output transition prediction results, the risk level of phase transition is assessed in real time, the safety level of phase synchronization between the current main and backup power supplies is automatically identified, and the reference switching time threshold is adaptively delayed according to the risk level of phase transition to ensure that the final switching is completed when the phase is stable.

[0093] The system automatically identifies the safety level of phase synchronization between the primary and backup power supplies and adaptively delays the reference switching time threshold based on the risk level of phase transition. The specific steps are as follows:

[0094] During the real-time switching process of the backup power supply, the machine learning model outputs the current transition coefficient, which is compared with a set transition coefficient reference threshold to assess the phase synchronization risk level between the primary and backup power supplies. A risk level function is introduced for quantification, and the expression of the risk level function is as follows:

[0095]

[0096] ,in: This represents the phase transition coefficient output by the model; a larger value indicates a more severe phase transition. The reference threshold for the jump coefficient set for the system is usually determined based on historical data experience or model tuning. This is the phase transition risk level coefficient, reflecting the intensity ratio of the current transition level relative to the preset safety threshold;

[0097] This step is used to standardize the phase jump level and convert it into a universal phase jump risk level coefficient. This facilitates adaptive adjustment and switching strategies in different scenarios.

[0098] Based on phase jump risk level coefficient An adaptive delay factor is calculated to adjust the baseline switching time threshold. The adaptive delay factor is calculated based on an exponentially smoothed delay function, and the expression for the calculation is as follows:

[0099]

[0100] ,in: The adaptive delay factor determines the adjustment factor for the actual switching time. This is the maximum delay gain coefficient, which controls the upper limit of the delay. The jump response sensitivity coefficient controls the steepness of the response of the delay function;

[0101] This step controls the delay amplitude through exponential smoothing, enabling the system to maintain high efficiency when the risk is low and automatically slow down the switching pace when the change is severe, thereby improving robustness and system stability.

[0102] Apply the adaptive delay factor to the original reference switching time threshold. To obtain the adaptive delay switching threshold actually used to control the switching action, the expression is:

[0103]

[0104] ,in: The delay switching time threshold is ultimately used to control the execution. Set the system's preset baseline switching time (e.g., 10 seconds). This is the maximum allowable switching delay time, used to prevent excessive delays from affecting power supply.

[0105] A maximum switching time threshold is set. If phase synchronization cannot be achieved within the maximum delay threshold, an alarm system is automatically triggered. Alarm information is rapidly pushed to maintenance personnel, managers, and engineers through various channels such as the human-machine interface, monitoring platform, SMS alarms, or voice prompts. The alarm details the severity of the phase jump and recommended handling measures (such as manual switching, emergency disconnection of backup power, and generator inspection) to ensure timely manual intervention and prevent further escalation of serious incidents, thus ensuring the safe and stable operation of the system. A critical safety fallback mechanism is provided to ensure rapid manual intervention to minimize system operational risks should the automatic control fail to handle the phase jump risk.

[0106] The aforementioned IoT-based intelligent monitoring and switching control method for backup power achieves a key leap from "timed switching" to "state-aware driving switching." This allows the system to fully perceive the phase synchronization status between the primary and backup power supplies before switching, and uses machine learning models to dynamically predict and intelligently judge phase transition trends. This effectively avoids the risks of inrush current, equipment damage, and regional tripping caused by forced switching due to phase inconsistency. This method significantly improves the safety, accuracy, and intelligence of the backup power switching process, ensuring power continuity while providing higher power operation assurance capabilities for critical load scenarios such as hospitals, data centers, and industrial production lines.

[0107] This invention provides, for example Figure 2 The IoT-based intelligent monitoring and switching control system for backup power includes a baseline switching strategy configuration module, a real-time phase acquisition and analysis module, a phase jump feature construction module, a phase jump prediction and identification module, and an adaptive switching duration adjustment module.

[0108] The baseline switching strategy configuration module, by combining historical operating data, load characteristics and technical specifications of backup power switching, sets an optimal baseline switching time threshold and clarifies the initial safe switching period.

[0109] The real-time phase acquisition and analysis module uses a high-precision synchronous sampling unit to perform real-time, high-frequency synchronous sampling of the voltage waveforms of the backup power supply and the active power supply during the baseline waiting period of the backup power supply switching execution. It also extracts the real-time phase angle of the main and backup power supply voltage waveforms in real time through digital signal processing technology.

[0110] The phase jump feature construction module extracts key indicators that reflect the phase jump characteristics between the main and backup power supplies based on the collected real-time phase data. It then performs in-depth feature engineering processing on the key indicators to construct a set of features that characterize the degree of phase jump.

[0111] The phase jump prediction and recognition module transforms the constructed feature set into feature vectors, which are then input into a pre-trained and put into use machine learning model to accurately judge and dynamically predict the real-time phase jump state and its evolution trend during backup power switching.

[0112] The adaptive switching time control module assesses the risk level of phase transitions in real time based on the output transition prediction results, automatically identifies the safety level of phase synchronization between the current primary and backup power supplies, and adaptively delays the reference switching time threshold according to the risk level of phase transitions.

[0113] The IoT-based intelligent monitoring and switching control method for backup power provided in this invention is implemented through the aforementioned IoT-based intelligent monitoring and switching control system for backup power. For details of the specific methods and processes of the IoT-based intelligent monitoring and switching control system for backup power, please refer to the embodiments of the aforementioned IoT-based intelligent monitoring and switching control method for backup power, which will not be repeated here.

[0114] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0115] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, 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 foregoing 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.

[0116] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for intelligent monitoring and switching control of backup power supply based on Internet of Things, characterized in that, The method comprises the following steps: Set a best reference switching time threshold based on historical operation data, load characteristics and technical specifications of the backup power supply switching, and determine an initial safe switching period; In the reference waiting period of the backup power supply switching, use a high-precision synchronous sampling unit to perform real-time and high-frequency synchronous sampling on the voltage waveforms of the backup power supply and the active power supply, and use a digital signal processing technology to extract real-time phase angles of the voltage waveforms of the active power supply and the backup power supply in real time; Based on the collected real-time phase data, key indicators reflecting the phase jump characteristics between the active power supply and the backup power supply are extracted, and a feature set representing the phase jump degree is constructed through in-depth feature engineering processing on the key indicators; The constructed feature set is converted into a feature vector, which is input into a machine learning model that has been pre-trained and put into use, to accurately judge and dynamically predict the real-time phase jump state and its evolution trend during the backup power supply switching; Based on the output jump prediction result, the risk degree of phase jump is evaluated in real time, the safety level of phase synchronization between the active power supply and the backup power supply is automatically identified, and the reference switching time threshold is adaptively delayed according to the risk degree of phase jump; The safety level of phase synchronization between the active power supply and the backup power supply is automatically identified, and the reference switching time threshold is adaptively delayed according to the risk degree of phase jump, and the specific steps are as follows: In the real-time process of the backup power supply switching, the jump coefficient at the current time is output by the machine learning model, which is compared with the set jump coefficient reference threshold to evaluate the risk level of phase synchronization between the active power supply and the backup power supply, and a risk level function is introduced for quantification, and the expression of the risk level function is: ， wherein: is a jump coefficient outputted by the model, is a jump coefficient reference threshold set by the system, is a phase jump risk level coefficient, reflecting the intensity proportion of the current jump degree relative to the preset safety threshold; Phase jump risk level coefficient An adaptive delay factor is calculated for adjusting the reference switching time threshold, the expression of which is: ， wherein: is an adaptive delay factor, determining the adjustment multiple of the actual switching time, is a maximum delay gain coefficient, controlling the upper limit of the delay, is a jump response sensitivity coefficient, controlling the response steepness of the delay function; In the monitoring window, the accumulated values of the jump frequency and amplitude of the phase angle per unit time are subjected to in-depth feature engineering processing to generate the specific steps of the phase mutation reference value as follows: A jump energy function is constructed to capture the joint strength of the frequency and amplitude; The jump energy is normalized and a sequential disturbance expansion rate is introduced to generate the phase mutation reference value; In the monitoring window, the skewness of the phase difference change is subjected to in-depth feature engineering processing to generate the specific steps of the phase fluctuation asymmetry reference value as follows: In the monitoring window, the phase difference curve of the active power supply and the backup power supply is analyzed, all amplitude extreme points are identified, the phase difference value corresponding to each extreme point is recorded, and an amplitude extreme point sequence is formed in sequence; For all extreme points, distinguish between positive and negative extreme, and then classify them into the set of positive amplitude extreme values and the set of negative amplitude extreme values, respectively and negative amplitude extreme values ​ to a set of positive amplitude extrema and a set of negative amplitude extrema The normalized reference value is used to calculate the relative offset of each extrema point, and then the phase fluctuation asymmetry reference value is calculated according to the ratio of the relative offset.

2. The IoT-based intelligent monitoring and switching control method of backup power supply according to claim 1, characterized in that, The real-time phase angle of the voltage waveform of the active power supply and the backup power supply is extracted in real time through fast Fourier transform, and the specific steps are as follows: S1, pre-process and window function process the sampling data to reduce spectral leakage; S2, perform fast Fourier transform to convert the time domain signal to the frequency domain, find the amplitude peak value near the detected main frequency, and read the phase; S3, take the read phase as a real-time reference point to calculate the phase angles of the active power supply and the backup power supply; S4, continuously repeat S1-S3 and periodically refresh the phase results to realize real-time monitoring of the phase of the voltage waveforms of the active power supply and the backup power supply. 3.The IoT-based intelligent monitoring and switching control method of backup power supply according to claim 1, characterized in that, The signal processing technology is used to extract the key indicators reflecting the phase jump characteristics between the main and standby power sources. The extracted indicators include the accumulated value of the jump frequency and amplitude of the phase angle per unit time and the skew degree of the phase difference change. In the monitoring window, the extracted indicators are subjected to in-depth feature engineering processing to generate a phase mutation reference value and a phase fluctuation asymmetry reference value. A feature set representing the phase jump degree is constructed by the phase mutation reference value and the phase fluctuation asymmetry reference value. 4.The IoT-based intelligent monitoring and switching control method of backup power supply according to claim 3, characterized in that, The feature set constructed by the phase mutation reference value and the phase fluctuation asymmetry reference value is converted into a feature vector and input into a machine learning model that has been pre-trained and put into use. A jump coefficient is generated based on the machine learning model to accurately judge and dynamically predict the real-time phase jump state and its evolution trend during the standby power source switching. 5.The IoT-based intelligent monitoring and switching control method of backup power supply according to claim 1, characterized in that, Applying an adaptive delay factor to an original reference handover time threshold , obtaining an adaptive delay handover threshold actually used to control the handover action, the obtained expression being: ， wherein: is a final for control execution delay switching time threshold, is a preset reference switching time, is a maximum allowed switching delay time.

6. The Internet of Things-based intelligent monitoring and switching control system for backup power supply, which is used for implementing the Internet of Things-based intelligent monitoring and switching control method for backup power supply according to any one of claims 1-5, characterized in that, The system comprises a benchmark switching strategy configuration module, a real-time phase acquisition and analysis module, a phase jump feature construction module, a phase jump prediction and identification module, and an adaptive switching time length regulation module. The benchmark switching strategy configuration module sets a best benchmark switching time threshold in combination with the historical operation data, load characteristics, and technical specifications of the standby power source switching, and determines an initial safe switching period. The real-time phase acquisition and analysis module uses a high-precision synchronous sampling unit to perform real-time and high-frequency synchronous sampling of the voltage waveforms of the standby power source and the active power source in the benchmark waiting period of the standby power source switching, and extracts the real-time phase angle of the voltage waveforms of the main and standby power sources in real time through digital signal processing technology. The phase jump feature construction module extracts key indicators reflecting the phase jump characteristics between the main and standby power sources from the collected real-time phase data, and subjects the key indicators to in-depth feature engineering processing to construct a feature set representing the phase jump degree. The phase jump prediction and identification module converts the constructed feature set into a feature vector and inputs it into a machine learning model that has been pre-trained and put into use to accurately judge and dynamically predict the real-time phase jump state and its evolution trend during the standby power source switching. The adaptive switching time length regulation module evaluates the risk degree of the phase jump in real time based on the output jump prediction result, automatically identifies the safety level of the phase synchronization between the current main and standby power sources, and adaptively delays the benchmark switching time threshold according to the risk degree of the phase jump.

Images