DC bus power failure warning method and system combined with battery detection

Abstract

The application provides a direct-current bus power failure early warning method and system combined with battery detection, relates to the technical field of power failure early warning, and comprises the following steps: performing multivariate time sequence feature monitoring on a main direct-current bus, and performing direct-current bus side risk prediction according to a monitoring result; if the first power failure early warning is a non-empty set, after switching a backup direct-current bus to connect a battery for power supply, battery side state data is traced back according to the first power failure early warning and a first early warning time scale; the credibility of the first power failure early warning is verified based on battery historical state data; the verification demand is matched according to the first risk authenticity; device side state data is traced back according to the first early warning time scale; device side operation consistency is quantified according to N device historical state data; after quantifying the power failure early warning level according to the second risk authenticity, the operation and maintenance of the first power failure early warning are transmitted out by using the power failure early warning level. Through the application, the technical effect of improving fault detection precision can be achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power loss warning, and particularly relates to a direct current bus power loss warning method and system combined with battery detection. BACKGROUND

[0002] Direct current bus power supply systems are widely used in critical fields such as data centers, rail transit, communication base stations and industrial automation. The core function of the direct current bus power supply system is to transmit the energy of the battery to multiple power-consuming devices through the bus to ensure stable power supply of the system. However, with the increase of power demand and the complication of devices, the existing technology still has many deficiencies in power loss detection, warning mechanism, power supply switching and long-term operation stability, which makes it difficult to guarantee the reliability of the system and may even cause serious power interruption accidents.

[0003] At present, the direct current bus power supply system mainly relies on fixed threshold alarm and simple voltage and current detection methods to judge the system state. However, this method has great limitations. First, the fixed threshold alarm strategy is difficult to adapt to changes in different working conditions. For example, in the case of large load fluctuation, short-term voltage drop may not affect the normal operation of the system, but the fixed threshold strategy may misreport faults, causing maintenance personnel to frequently check invalid alarms and reducing work efficiency. At the same time, if the fixed threshold is set too wide, some potential power loss risks may not be identified in time, resulting in remediation after power anomalies occur, which cannot achieve early warning.

[0004] In summary, the existing technology has the technical problem that the fixed threshold alarm strategy is difficult to adapt to changes in different working conditions, resulting in misreporting or missing faults, which further affects the fault warning accuracy and maintenance efficiency of the direct current bus power supply system. SUMMARY

[0005] The purpose of the present application is to provide a direct current bus power loss warning method and system combined with battery detection, to solve the technical problem in the prior art that the fixed threshold alarm strategy is difficult to adapt to changes in different working conditions, resulting in misreporting or missing faults, which further affects the fault warning accuracy and maintenance efficiency of the direct current bus power supply system.

[0006] In view of the above problems, the present application provides a direct current bus power loss warning method and system combined with battery detection.

[0007] In a first aspect, the application provides a direct current bus power failure early warning method combined with storage battery detection, which is realized by a direct current bus power failure early warning system combined with storage battery detection, and includes: performing multi-element time sequence feature monitoring on a main direct current bus, performing direct current bus side risk prediction according to a monitoring result, and outputting a first power failure early warning; if the first power failure early warning is a non-empty set, after switching a backup direct current bus to connect a storage battery for power supply, performing battery side state data backtracking according to the first power failure early warning matched with a first early warning time scale to obtain battery historical state data of the storage battery; performing a credibility check of the first power failure early warning based on the battery historical state data to output a first risk authenticity; performing a check demand matching according to the first risk authenticity to obtain N electrical equipment; performing device side state data backtracking according to the first early warning time scale to obtain N device historical state data of the N electrical equipment; performing device side operation consistency quantification according to the N device historical state data to obtain a second risk authenticity; after quantifying a power failure early warning level according to the second risk authenticity, performing operation and maintenance transmission of the first power failure early warning by using the power failure early warning level.

[0008] In a second aspect, the application further provides a direct current bus power failure early warning system combined with storage battery detection, which is used to execute the direct current bus power failure early warning method combined with storage battery detection as described in the first aspect, and includes: a risk prediction module configured to perform multi-element time sequence feature monitoring on a main direct current bus, perform direct current bus side risk prediction according to a monitoring result, and output a first power failure early warning; a battery state backtracking module configured to, if the first power failure early warning is a non-empty set, perform battery side state data backtracking according to the first power failure early warning matched with a first early warning time scale after switching a backup direct current bus to connect a storage battery for power supply to obtain battery historical state data of the storage battery; a credibility check module configured to perform a credibility check of the first power failure early warning based on the battery historical state data to output a first risk authenticity; a check demand matching module configured to perform a check demand matching according to the first risk authenticity to obtain N electrical equipment; a device state backtracking module configured to perform device side state data backtracking according to the first early warning time scale to obtain N device historical state data of the N electrical equipment; an operation consistency quantification module configured to perform device side operation consistency quantification according to the N device historical state data to obtain a second risk authenticity; and an operation and maintenance transmission module configured to, after quantifying a power failure early warning level according to the second risk authenticity, perform operation and maintenance transmission of the first power failure early warning by using the power failure early warning level.

[0009] The technical solution provided in this application has at least the following technical effects or advantages: by achieving the technical goals of power failure warning and fault diagnosis based on dynamic thresholds, multi-dimensional data fusion and intelligent prediction, the technical effects of improving fault detection accuracy, reducing false alarms and missed alarms, optimizing power supply switching strategies, and improving system stability and operation and maintenance efficiency are achieved.

[0010] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, which can be implemented in accordance with the contents of the description, and to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically listed below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in this application or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely exemplary, and a person of ordinary skill in the art can obtain other drawings based on the provided drawings without creative work.

[0012] Figure 1 This is a flow chart of the DC bus power failure warning method combined with battery detection in this application;

[0013] Figure 2 This is a structural diagram of the DC bus power failure warning system combined with battery detection in this application.

[0014] Explanation of the accompanying symbols: risk prediction module 11, battery status backtracking module 12, credibility verification module 13, verification requirement matching module 14, equipment status backtracking module 15, operation consistency quantification module 16, operation and maintenance output module 17. DETAILED DESCRIPTION

[0015] This application provides a DC bus power failure warning method and system combined with battery detection, resolving the technical problem in the prior art where fixed threshold alarm strategies struggle to adapt to varying operating conditions, leading to false or missed fault reports, further impacting the accuracy of fault warnings and the operational efficiency of DC bus power supply systems. This application achieves the technical goals of power failure warning and fault diagnosis based on dynamic thresholds, multi-dimensional data fusion, and intelligent prediction, achieving the technical effects of improving fault detection accuracy, reducing false and missed reports, optimizing power supply switching strategies, and enhancing system stability and operational efficiency.

[0016] Below, the technical solutions in the present application will be described clearly and completely with reference to the drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application. In addition, it should be noted that, for the convenience of description, only parts related to the present application are shown in the drawings, not all.

[0017] Embodiment one, please refer to the attached Figure 1 The present application provides a direct current bus power failure early warning method combined with storage battery detection, applied to a direct current bus power failure early warning system combined with storage battery detection, and specifically includes the following steps:

[0018] S1: multi-element time sequence feature monitoring is performed on the main direct current bus, and according to the monitoring result, a direct current bus side risk prediction is performed, and a first power failure early warning is output.

[0019] Further, S1 includes: S11: setting a power failure prediction scale according to a power failure switching response delay; S12: performing multi-element time sequence feature monitoring on the main direct current bus with K times the power failure prediction scale as a time constraint, to obtain a bus voltage fluctuation rate, a current harmonic distortion rate and a temperature gradient change rate; S13: performing a direct current bus side risk prediction by multi-modal time sequence coupling analysis of the bus voltage fluctuation rate, the current harmonic distortion rate and the temperature gradient change rate, and outputting the first power failure early warning.

[0020] Further, S12 includes: S121: performing voltage monitoring at a preset first sampling frequency by a voltage sensor connected in parallel between the positive and negative poles of the direct current bus, to obtain a voltage instantaneous value sequence; S122: performing current monitoring at a preset second sampling frequency by a Hall current sensor connected in series to the direct current bus, to obtain a current waveform sampling sequence; S123: performing temperature time-varying monitoring at a preset third sampling frequency by a PT100 sensing array arranged along the direct current bus, to obtain a temperature distribution matrix; S124: performing numerical change feature extraction on the voltage instantaneous value sequence and the current waveform sampling sequence respectively, to obtain the bus voltage fluctuation rate and the current harmonic distortion rate; S125: performing spatio-temporal gradient analysis on the temperature distribution matrix, to obtain the temperature gradient change rate.

[0021] Further, S123 further includes: the voltage instantaneous value sequence, the current waveform sampling sequence and the temperature distribution matrix are time sequence aligned, and meet K times the power failure prediction scale.

[0022] Further, S13 includes: S131: using the unique identifiers of multiple electrical devices and batteries as search conditions, performing network data collection, and obtaining sample time series power failure records; S132: using the power failure node as the sampling starting point, using K+1 times the power failure prediction scale as the sampling scale, performing reverse data collection of the sample time series power failure records, and obtaining multiple positive sample multivariate time series features, wherein the positive sample multivariate time series features include positive sample voltage fluctuation rate, positive sample harmonic distortion rate, and positive sample gradient change rate; S133: removing the multiple positive sample multivariate time series features from the sample time series power failure records. After obtaining the meta-time series features, randomly sample the sample time series power failure records with K+1 times the power failure prediction scale as the sampling scale to obtain multiple negative sample multi-time series features; S134: use the multiple positive sample multi-time series features and the multiple negative sample multi-time series features as training data to optimize the parameters of the power failure prediction model, wherein the power failure prediction model is constructed based on the LSTM model; S135: input the bus voltage fluctuation rate, current harmonic distortion rate and temperature gradient change rate into the power failure prediction model to predict the risk on the DC bus side, and output the first power failure warning.

[0023] Specifically, setting the power outage prediction scale refers to determining an appropriate timeframe within which a possible DC bus power outage can be effectively predicted. A power outage occurs when the DC bus fails to deliver power, which can be caused by a battery failure, a faulty electrical device, or a problem with the bus itself. The switching response refers to the process of initiating a backup power source or taking other measures after a power outage is detected, while the response delay refers to the time interval between the detection of a power outage and the completion of the switchover. When setting the power outage prediction scale, the length of the response delay needs to be comprehensively considered to ensure that the prediction time is sufficiently advanced to allow sufficient time for switching or other emergency measures to mitigate the impact of the power outage.

[0024] By connecting a voltage sensor in parallel between the positive and negative poles of the DC bus, voltage monitoring is performed at a preset first sampling frequency to obtain a sequence of instantaneous voltage values. Parallel refers to an electrical connection method, indicating that the two terminals of the voltage sensor are connected to the positive and negative poles of the DC bus, respectively, to enable real-time voltage measurement. The DC bus is a conductor system used to transmit DC power and is used to power multiple electrical devices. A voltage sensor is a device used to measure voltage and can be a resistor divider, a capacitor sensor, or a photoelectric isolation sensor. The preset first sampling frequency means that voltage data is regularly collected within a set fixed time interval. For example, if the sampling frequency is one hundred times per second, the sensor will record the voltage value every ten milliseconds. Voltage monitoring refers to the continuous observation of the voltage on the bus in order to analyze its changing trends. The sequence of instantaneous voltage values ​​represents a series of voltage values ​​measured at different time points and can be used for subsequent analysis, such as determining whether there are abnormal fluctuations in the bus.

[0025] The current waveform sampling sequence is obtained by performing current monitoring at a preset second sampling frequency through a Hall current sensor connected in series to the DC bus. Series connection means that the current sensor is installed in the path of the line so that all current flowing through the bus will pass through the sensor, thereby accurately measuring the current. The Hall current sensor is a current detection device based on the Hall effect. It can measure the current size in a non-contact manner and has high response speed and safety. The preset second sampling frequency means that current data is collected regularly according to the interval set by the system. For example, fifty times per second means that the current value is recorded every twenty milliseconds. Current monitoring refers to the real-time observation of current changes in the bus to analyze the load situation or detect anomalies. The current waveform sampling sequence represents a series of current values ​​recorded at different time points, which can be used to analyze the harmonic components, instantaneous changes and other characteristics of the current, so as to determine whether the bus is disturbed or the load suddenly changes.

[0026] In addition to voltage and current, temperature is also a significant factor affecting busbar operation. Therefore, a PT100 sensor array deployed along the DC bus performs time-varying temperature monitoring at a preset third sampling frequency, generating a temperature distribution matrix. Deploying sensors along the DC busbar means installing them at regular intervals to acquire temperature data at various locations. PT100 sensors are high-precision platinum resistance temperature sensors whose resistance varies with temperature, enabling precise temperature measurement. A sensor array consists of multiple sensors arranged in a regular pattern, forming a system capable of acquiring spatial temperature information. The preset third sampling frequency collects temperature data at fixed intervals, for example, 20 times per second, meaning a temperature value is recorded every 50 milliseconds. Time-varying temperature monitoring continuously observes busbar temperature changes at different time points to detect abnormal temperature rise or localized overheating. The temperature distribution matrix represents a collection of temperature values ​​measured at different locations and time points and can be used to analyze spatial temperature variations.

[0027] After collecting voltage and current data, numerical variation feature extraction is performed on the instantaneous voltage value sequence and the current waveform sampling sequence, respectively, to obtain the bus voltage fluctuation rate and current harmonic distortion rate. Numerical variation feature extraction involves extracting key features from the raw data that describe system state changes for analysis. The bus voltage fluctuation rate indicates the degree of fluctuation of the bus voltage relative to its stable value and is calculated from the range of variation in the instantaneous voltage value. The current harmonic distortion rate refers to the proportion of higher-order harmonic components contained in the current waveform. Higher-order harmonics can cause problems such as equipment heating and increased power loss.

[0028] Furthermore, a spatiotemporal gradient analysis is required on the temperature distribution matrix to determine the temperature gradient rate of change. This analysis considers both spatial and temporal temperature variations to determine if there are abnormal heat sources or localized overheating. For example, if a section of a busbar heats up by five degrees Celsius in five seconds, while other areas remain largely unchanged, this area may be overloaded or experiencing poor contact. The temperature gradient rate of change indicates the rate at which the temperature changes at different locations or time points.

[0029] Time-aligned voltage instantaneous value sequences, current waveform sampling sequences, and temperature distribution matrices mean that these three types of data must be processed synchronously according to the same time base. Time series refers to the order in which data is arranged over time, while alignment means that data from different sources must match in time.

[0030] Because the instantaneous voltage value sequence, current waveform sampling sequence, and temperature distribution matrix are needed to predict power outages, they must meet a K-fold power outage prediction scale. K is an integer greater than 0. For example, if the backup power supply switchover time is five seconds, the power outage prediction scale may require at least five seconds to provide a timely warning before the voltage drops. K-fold means that this time range can be proportionally expanded. If K is 2, the prediction scale will need to be ten seconds, providing a longer time window to observe data trends and improve prediction accuracy.

[0031] Therefore, to meet the K-fold power outage prediction scale, the voltage instantaneous value sequence, current waveform sampling sequence, and temperature distribution matrix must cover a sufficiently long time range. For example, if the system needs to predict a power outage ten seconds in the future, and the voltage is sampled 100 times per second, the current is sampled 50 times per second, and the temperature is sampled 20 times per second, then at least 1,000, 100, and 200 data points must be stored, respectively, to ensure that the data volume is sufficient to support the prediction model for trend analysis. If K increases from two to three, meaning the power outage prediction scale increases from ten seconds to fifteen seconds, then 1,500 data points must be stored for voltage, 750 for current, and 300 for temperature, with the data volume increasing proportionally.

[0032] The unique identifiers of the plurality of electrical equipment and the battery are used as retrieval conditions for network data collection to obtain sample time sequence power failure records. The plurality of electrical equipment refers to all electrical equipment powered by the battery. The unique identifier refers to a unique code used to distinguish different equipment, such as a serial number or MAC address for each electrical equipment and battery. The retrieval condition refers to the filtering criteria used when querying data, meaning that the data of specific equipment is filtered using the unique identifier. Network data collection means collecting data scattered on different devices into a platform or database through network connection. Sample time sequence power failure records refer to power failure data stored in chronological order, including the time of power failure, device status, and related electrical parameters. For example, if a certain busbar loses power at 10:30 on a certain day, the system will record the changes in voltage, current, and temperature at that time and store the data for a certain period of time for subsequent analysis.

[0033] The power failure node is used as the starting point for sampling, and the K+1 times power failure prediction scale is used as the sampling scale for reverse data collection of sample time sequence power failure records to obtain a plurality of positive sample multivariate time sequence features. The power failure node refers to the time point at which the busbar actually loses power, for example, if the busbar loses power at 10:30, then this time point is the power failure node. The sampling starting point indicates the starting time of data sampling, which is the power failure node. The K+1 times power failure prediction scale indicates the length of the sampling time window, for example, if the power failure prediction scale is ten seconds, then K+1 times is twenty seconds, meaning that twenty seconds of data will be backtracked. Reverse data collection means collecting historical data from the time of power failure to analyze the state changes before power failure. Positive samples refer to samples that actually lose power, and multivariate time sequence features refer to multiple different categories of time series data.

[0034] Among them, the positive sample multivariate time sequence features include positive sample voltage fluctuation rate, positive sample harmonic distortion rate, and positive sample gradient change rate. Voltage fluctuation rate refers to the degree of change of voltage over time. Harmonic distortion rate refers to the proportion of high-order harmonics in current. Gradient change rate represents the rate of change of temperature over time and space.

[0035] The plurality of positive sample multivariate time sequence features are removed from the sample time sequence outage record. Removing positive sample multivariate time sequence features means removing time periods that have actually experienced outages from the sample data set to avoid data confusion. Random sampling of the sample time sequence outage record is performed with a sampling scale of K+1 times the outage prediction scale to obtain a plurality of negative sample multivariate time sequence features. Random sampling means randomly selecting time periods that have not experienced outages from the remaining data for comparative analysis. For example, if the positive sample corresponds to data twenty seconds before an outage occurs, the negative sample can be any twenty seconds of data in normal operation. Negative sample multivariate time sequence features represent the voltage fluctuation rate, harmonic distortion rate, and temperature gradient change rate of these data that have not experienced outages. Table 1 is the data record of the plurality of devices that recently obtained positive and negative sample multivariate time sequence features.

[0036] Table 1: Data record of the plurality of devices that recently obtained positive and negative sample multivariate time sequence features

[0037]

[0038] The plurality of positive sample multivariate time sequence features and the plurality of negative sample multivariate time sequence features are used as training data for parameter tuning and optimization of the outage prediction model, wherein the outage prediction model is based on an LSTM model. Training data refers to sample data used for machine learning model learning, including features and corresponding categories. Parameter tuning and optimization refers to adjusting the hyperparameters of the model to make the prediction more accurate, such as adjusting the number of layers, the number of neurons, or the learning rate of LSTM, etc. LSTM model is a long short-term memory network suitable for processing time series data, which can identify long-term trends and short-term fluctuations to effectively predict outages. For example, if the past twenty seconds of voltage, current, and temperature data are input, LSTM can predict whether an outage will occur in the next five seconds.

[0039] The bus voltage fluctuation rate, current harmonic distortion rate, and temperature gradient change rate are input into the outage prediction model for risk prediction on the DC bus side, and the first outage early warning is output. Risk prediction means predicting whether an outage is likely to occur in the future based on current data. Bus voltage fluctuation rate, current harmonic distortion rate, and temperature gradient change rate as input features, their change trend determines the prediction result. The first outage early warning represents the preliminary alarm issued by the system before the actual occurrence of an outage, which can be used to remind maintenance personnel to check the equipment, or trigger an automatic protection mechanism, such as switching to a backup power source in advance.

[0040] S2: If the first outage early warning is a non-empty set, after switching to a backup DC bus connected to a battery for power supply, the battery historical state data of the battery is obtained by backtracking the battery side state data according to the first outage early warning matching the first early warning time scale.

[0041] Further, S2 further comprises: S21: collecting a first use duration of the main DC bus if the first power failure warning is an empty set; S22: determining whether the first use duration meets a preset switching time scale; and S23: switching the backup DC bus to connect the battery power supply if the first use duration meets the preset switching time scale.

[0042] Specifically, if the first power failure warning is a non-empty set, the backup DC bus is switched to connect the battery power supply. A non-empty set means that potential power failure risks have been detected, rather than blank or invalid data. When the first power failure warning exists, it means that the bus is about to lose power or has already entered a power failure state, so measures need to be taken. Switching the backup DC bus means that after the main bus loses power, it is automatically or manually switched to the backup DC bus to ensure uninterrupted power supply. Connecting the battery power supply means that the backup DC bus obtains energy from the battery to provide continuous power to the load. Then, according to the time range of the first power failure warning, the data of a certain time interval is traced back. The first warning time scale represents the time window set for power failure warning. For example, if the first warning time scale is set to thirty seconds, the battery operation data within thirty seconds will be traced back after detecting the power failure risk. Battery-side state data backtracking refers to extracting key operating parameters of the battery from historical data, such as voltage, current, temperature, and internal resistance changes, to analyze the battery state before power failure. Collecting and organizing the battery operation data within a certain period of time before power failure, the battery historical state data of the battery is obtained. The battery historical state data includes the voltage change trend, current output characteristics, temperature fluctuation, and internal resistance growth of the battery.

[0043] If the first power failure warning is an empty set, the first use duration of the main DC bus is collected, which means that if no power failure warning is generated after detection, i.e., the first power failure warning is an empty set, no emergency measures will be taken immediately, and the operation of the main DC bus will continue to be monitored. The first use duration refers to the continuous running time of the main DC bus since the last switching or starting. The purpose of collecting the first use duration is to determine whether the bus has been running for a long time, so as to decide whether switching is needed.

[0044] Determining whether the first use duration meets the preset switching time scale means that it is necessary to check whether the use duration of the main DC bus has reached the preset time threshold. The preset switching time scale is used to control the maximum continuous running time of the main DC bus to avoid equipment aging or failure due to long-term load operation.

[0045] If the first use duration meets the preset switching time scale, the standby DC bus is connected to the battery power supply, indicating that when the continuous running time of the main DC bus reaches the set threshold, the switching operation is automatically performed, the standby DC bus is connected to the battery, and power is provided for the load, preventing unpredictable failures of the main DC bus due to long-time operation, while also balancing the running load of the bus and prolonging the overall service life of the equipment.

[0046] S3: performing credibility verification of the first power failure early warning based on the battery historical state data, and outputting a first risk authenticity.

[0047] Further, S3 includes: S31: interactively obtaining a plurality of sample power failure state data; S32: obtaining a plurality of sample power failure probabilities by state backtracking on the plurality of sample power failure state data; S33: storing the plurality of sample power failure state data and the plurality of sample power failure probabilities by using a knowledge graph, and completing construction of a power failure state library; S34: inputting the battery historical state data into the power failure state library for state similarity matching, obtaining H sample power failure state data and H sample power failure probabilities falling within a state similarity threshold; and S35: performing weighted calculation on the H sample power failure probabilities according to H state similarity parameters of the battery historical state data and the H sample power failure state data, and outputting the first risk authenticity.

[0048] Specifically, the interactive obtaining of the plurality of sample power failure state data means that the power failure state information of different samples is obtained through various ways. For example, the power failure state is determined by real-time monitoring of the voltage, current, temperature and other parameters of the battery and the DC bus, and historical data or manually entered data can also be used for correction. The plurality of sample power failure state data refers to the power failure information under different time, different equipment or different operating conditions, which can help establish a more comprehensive power failure mode.

[0049] The plurality of sample power failure probabilities are obtained by state backtracking on the plurality of sample power failure state data, indicating that the collected power failure data is analyzed by backtracking to calculate the probability of power failure. State backtracking means tracking the historical operating data of each sample to find out the key factors leading to power failure. For example, the voltage of one battery dropped by five volts in the last five minutes before power failure, while the voltage of another battery dropped by two volts in the same time, but both of them occurred power failure. By comparing these data, the power failure probability corresponding to different voltage drop amplitudes can be calculated. Assuming that among the one hundred samples, sixty samples lose power when the voltage drops more than five volts, then the power failure probability under this condition can be calculated as 60%, while in the case of voltage drop of two volts, the power failure probability may be only 20%.

[0050] The construction of the power failure state library means that the power failure data and the power failure probability are stored in a database that can be associated and queried. The knowledge graph is a storage method that uses nodes and relationships to build a data network, allowing different power failure state data to be linked. For example, the knowledge graph can associate a certain power failure state with its corresponding voltage fluctuations, current changes, and temperature abnormalities, making it more efficient to query the power failure probability under certain conditions. The power failure state library is a database containing a large amount of power failure state information, and all stored data can be used for subsequent matching and prediction.

[0051] The battery historical state data is input into the power failure state library for state similarity matching, obtaining H sample power failure state data and H sample power failure probability within the state similarity threshold. This means that the battery operating state data is used to find similar historical samples in the power failure state library, and the power failure state and corresponding power failure probability of these samples are extracted. State similarity matching is a mathematical calculation-based method, such as calculating the Euclidean distance or cosine similarity between the battery historical state data and the samples in the database to determine their similarity. The state similarity threshold represents the minimum standard for matching, for example, if the threshold is set to 90%, the system will only select samples that are at least 90% similar to the current state.

[0052] According to the H state similarity parameters of the battery historical state data and the H sample power failure state data, the H sample power failure probability is weighted and calculated, and the first risk authenticity is output, which represents the real power failure risk of the current battery state according to the matched sample data. Weighted calculation means that the power failure probabilities of different samples are combined according to the weights of their similarities to calculate the real power failure risk of the current battery state. The first risk authenticity represents the final output risk score, which can be used as an important basis for determining whether emergency measures need to be taken.

[0053] S4: According to the first risk authenticity, the verification demand matching is performed, and N electric equipment is obtained.

[0054] Specifically, according to the first risk authenticity, the verification demand matching is performed to determine which electric equipment may be affected and match the corresponding demand. Verification demand matching refers to automatically selecting electric equipment that may be affected according to the size of the risk authenticity and matching their work demands. For example, if some electric equipment has a very high requirement for power supply continuity, such as medical equipment or server rooms, when the first risk authenticity exceeds 70%, it is necessary to take measures in advance, such as switching to a backup power source or reducing the power consumption of non-critical loads, to ensure the normal operation of critical equipment. Then the number of electric equipment affected by the current power failure risk is determined, where N represents the number of specific equipment.

[0055] S5: performing device-side state data backtracking according to the first early warning time scale to obtain N device historical state data of the N power-using devices.

[0056] Specifically, performing device-side state data backtracking according to the first early warning time scale means that after receiving the first power loss early warning, the historical running states of the N power-using devices need to be backtracked according to the first early warning time scale to analyze their running conditions before the early warning. Device-side state data backtracking refers to querying and analyzing the historical data of each power-using device, including voltage, current, temperature and load condition, to determine whether the power loss has an impact on the device. Obtaining N device historical state data of the N power-using devices means that after completing data backtracking, the historical state information of all affected devices is successfully extracted, and a set of data records is established for each device. The N power-using devices refer to all devices affected by the power loss early warning, and the N device historical state data contain their respective historical running information.

[0057] S6: performing device-side running consistency quantification according to the N device historical state data to obtain a second risk authenticity.

[0058] Further, S6 includes: S61: adopting linear interpolation to unify the sampling rates of the N device historical state data to obtain N calibrated historical state data; S62: adopting a sliding time window to perform transient feature extraction on the N device historical state data to obtain N voltage sag rate sequences and N current interruption node sequences; S63: constructing a correlation coefficient matrix according to the N voltage sag rate sequences; S64: calculating a synchronous power loss rate according to the N current interruption node sequences; S65: adopting the correlation coefficient matrix and the synchronous power loss rate to perform fusion quantification of device-side running consistency to obtain the second risk authenticity.

[0059] Further, S65 includes: S651: traversing the matrix non-diagonal line elements of the correlation coefficient matrix to obtain a first element quantity satisfying a preset linear correlation; S652: calculating a device pair number proportion of the first element quantity in the matrix elements; S653: when the device pair number proportion satisfies a preset first percentage and the synchronous power loss rate satisfies a preset second percentage, weighting and fusing the device pair number proportion and the synchronous power loss rate as the second risk authenticity output.

[0060] Specifically, adopting linear interpolation to unify the sampling rates of the N device historical state data means that after obtaining the historical state data of the N devices, the data of different devices need to be processed to keep the sampling rates consistent, and N calibrated historical state data is obtained. Linear interpolation is a common numerical processing method used to fill in the gaps between data or adjust the time steps of data.

[0061] A sliding time window is used to extract transient features from the historical status data of N devices. This means that after aligning the sampling rates, a sliding time window method is used to extract the transient features of the devices. A sliding time window is a technique used to analyze time series data. It can calculate data changes over continuous time periods, thereby discovering short-term fluctuation characteristics. Transient feature extraction refers to extracting key parameters that reflect the operating status of the device, such as voltage sag rate and current interruption nodes, from the fluctuating data within these short time windows. Obtaining N voltage sag rate sequences and N current interruption node sequences means calculating the voltage change rate and current interruption status of each device. The voltage sag rate sequence refers to the voltage drop rate of each device within different time windows. The current interruption node sequence records the time points when the current is interrupted during the operation of each device.

[0062] A correlation coefficient matrix is ​​constructed based on N voltage sag rate sequences. This represents the calculation of correlations between the voltage sag rates of N devices, forming a matrix. The correlation coefficient matrix is ​​a mathematical tool used to measure the degree of correlation between multiple variables and to determine whether voltage sags across different devices exhibit similar trends. For example, if the voltage sag rates of two devices maintain the same downward trend over multiple time windows, their correlation coefficients are close to one, indicating that they may be affected by the same external factor, such as DC bus voltage instability. However, if the voltage sag rates of two devices exhibit significantly different trends, the correlation coefficients are close to zero, indicating that their voltage sags may be caused by independent factors.

[0063] The synchronous power outage rate is calculated based on a sequence of N current interruption nodes. This means using the current interruption time points of N devices to calculate whether these devices experienced current interruptions at similar times, and using this to estimate the synchronous power outage rate. The synchronous power outage rate is an indicator that measures whether the power outage of devices is caused by the same event. For example, if eight out of ten devices experience current interruptions at similar times within a certain time period, the synchronous power outage rate is 80%, indicating that these devices likely experienced synchronous power outages due to a bus voltage drop or bus power outage. However, if only two of the ten devices have the same current interruption time, the synchronous power outage rate is lower, indicating that the power outage may be caused by an internal device fault rather than a complete bus power outage.

[0064] Traversing the off-diagonal elements of the correlation coefficient matrix means sequentially accessing all elements except the diagonal elements in the correlation coefficient matrix to obtain the correlation data between devices. The correlation coefficient matrix is ​​used to describe the relationship between the voltage sag rates of multiple devices, where the diagonal elements of the matrix are one, representing the correlation of the device itself, while the off-diagonal elements represent the correlation between different devices. For example, if there are four devices, the correlation coefficient matrix is ​​a square matrix with four rows and four columns, where the elements in the first row and second column represent the correlation coefficient between device one and device two, and the elements in the second row and third column represent the correlation coefficient between device two and device three. Traversing the off-diagonal elements of the matrix means that the correlation between these pairs of devices needs to be checked one by one, and the first element quantity that meets the preset linear correlation requirements is screened out.

[0065] Calculating the ratio of the first element to the number of device pairs in the matrix indicates the ratio of the number of device pairs that meet the preset correlation requirements to the total number of device pairs in the matrix. The number of device pairs refers to the number of pairwise combinations between different devices.

[0066] When the device pair ratio meets the first preset percentage and the synchronized power outage rate meets the second preset percentage, it's time to check whether the device pair ratio has reached the first threshold and whether the synchronized power outage rate has reached the second threshold. The first percentage measures the overall correlation of voltage dips between devices, while the second percentage measures whether devices experience synchronized power outages.

[0067] The weighted fusion of the device pair ratio and the simultaneous outage rate is used as the output of the second risk authenticity. This means that the device pair ratio and the simultaneous outage rate are weighted together to generate the final second risk authenticity. Weighted fusion means that different indicators can have different influence weights. For example, if the device pair ratio is more important than the simultaneous outage rate, it can be given a higher weight.

[0068] S7: After quantifying the power failure warning level according to the second risk reality, the power failure warning level is used to perform operation and maintenance output of the first power failure warning.

[0069] Specifically, after quantifying the power outage warning level based on the second risk authenticity, the power outage risk is graded based on the calculated second risk authenticity. The second risk authenticity reflects the credibility of the DC bus power supply anomaly, ranging from zero to one, with higher values ​​indicating a greater power outage risk.

[0070] The first power failure early warning operation and maintenance transmission indicates that the power failure early warning information is sent to the operation and maintenance personnel or the automatic operation and maintenance system according to the calculated early warning level. The power failure early warning level determines the mode and emergency degree of information transmission. For example, low risk may only record logs in the background system, medium risk may trigger SMS or email notification, and high risk may need to trigger an alarm sound and light signal immediately and automatically execute an emergency processing scheme.

[0071] In summary, the DC bus power failure early warning method provided in the application has the following technical effects: by achieving the technical target of power failure early warning and fault diagnosis based on dynamic threshold, multi-dimensional data fusion and intelligent prediction, the technical effects of improving fault detection accuracy, reducing false positives and false negatives, optimizing power supply switching strategy, and improving system stability and operation and maintenance efficiency are achieved.

[0072] Embodiment two, based on the same inventive concept as the DC bus power failure early warning method combined with battery detection in the foregoing embodiments, the application also provides a DC bus power failure early warning system combined with battery detection, please refer to the attached Figure 2 , comprising: a risk prediction module 11 for monitoring the main DC bus through multi-element time sequence characteristics, and performing DC bus side risk prediction according to the monitoring result, and outputting a first power failure early warning; a battery state backtracking module 12 for, if the first power failure early warning is a non-empty set, then after switching the standby DC bus to connect the battery power supply, performing battery side state data backtracking according to the first power failure early warning matched with the first early warning time scale, to obtain battery historical state data of the battery; a credibility verification module 13 for performing credibility verification of the first power failure early warning based on the battery historical state data, and outputting a first risk authenticity; a verification demand matching module 14 for performing verification demand matching according to the first risk authenticity, to obtain N electrical equipment; a device state backtracking module 15 for performing device side state data backtracking according to the first early warning time scale, to obtain N device historical state data of the N electrical equipment; a running consistency quantification module 16 for performing device side running consistency quantification according to the N device historical state data, to obtain a second risk authenticity; and an operation and maintenance transmission module 17 for, after quantifying the power failure early warning level according to the second risk authenticity, performing operation and maintenance transmission of the first power failure early warning by using the power failure early warning level.

[0073] Furthermore, the DC bus power failure warning system combined with battery detection is further used to: set a power failure prediction scale based on a power failure switching response delay; perform multivariate time series feature monitoring on the main DC bus using K times the power failure prediction scale as a time constraint to obtain a bus voltage fluctuation rate, a current harmonic distortion rate, and a temperature gradient change rate; analyze the bus voltage fluctuation rate, the current harmonic distortion rate, and the temperature gradient change rate through multimodal time series coupling to prejudge the risk on the DC bus side and output the first power failure warning.

[0074] Furthermore, the DC bus power failure warning system combined with battery detection is also used to: use the unique identifiers of multiple electrical devices and batteries as retrieval conditions to perform network data acquisition to obtain sample time-series power failure records; use the power failure node as the sampling starting point and K+1 times the power failure prediction scale as the sampling scale to perform reverse data acquisition of the sample time-series power failure records to obtain multiple positive sample multivariate time series features, wherein the positive sample multivariate time series features include the positive sample voltage fluctuation rate, the positive sample harmonic distortion rate and the positive sample gradient change rate; remove the positive sample voltage fluctuation rate, the positive sample harmonic distortion rate and the positive sample gradient change rate from the sample time-series power failure records. After obtaining the multiple positive sample multivariate time series features, random sampling of the sample time series power failure records is performed with K+1 times the power failure prediction scale as the sampling scale to obtain multiple negative sample multivariate time series features; the multiple positive sample multivariate time series features and the multiple negative sample multivariate time series features are used as training data to optimize the parameters of the power failure prediction model, wherein the power failure prediction model is constructed based on the LSTM model; the bus voltage fluctuation rate, current harmonic distortion rate and temperature gradient change rate are input into the power failure prediction model to predict the risk on the DC bus side, and the first power failure warning is output.

[0075] Furthermore, the DC bus power failure warning system combined with battery detection is also used to: perform voltage monitoring at a preset first sampling frequency through a voltage sensor connected in parallel between the positive and negative poles of the DC bus to obtain a voltage instantaneous value sequence; perform current monitoring at a preset second sampling frequency through a Hall current sensor connected in series on the DC bus to obtain a current waveform sampling sequence; perform temperature time-varying monitoring at a preset third sampling frequency through a PT100 sensor array deployed along the DC bus to obtain a temperature distribution matrix; perform numerical change feature extraction on the voltage instantaneous value sequence and the current waveform sampling sequence respectively to obtain the bus voltage fluctuation rate and the current harmonic distortion rate; and perform spatiotemporal gradient analysis on the temperature distribution matrix to obtain the temperature gradient change rate.

[0076] Furthermore, the DC bus power failure warning system combined with battery detection is also used for: the time-aligned voltage instantaneous value sequence, current waveform sampling sequence and temperature distribution matrix meet K times the power failure prediction scale.

[0077] Furthermore, the DC bus power failure warning system combined with battery detection is also used to: interactively obtain multiple sample power failure status data; obtain multiple sample power failure probabilities by performing state backtracing on the multiple sample power failure status data; use a knowledge graph to associate and store the multiple sample power failure status data and the multiple sample power failure probabilities to complete the construction of a power failure status library; input the battery historical status data into the power failure status library for state similarity matching to obtain H sample power failure status data and H sample power failure probabilities that fall within the state similarity threshold; perform weighted calculation on the H sample power failure probabilities based on the H state similarity parameters of the battery historical status data and the H sample power failure status data, and output the first risk authenticity.

[0078] Furthermore, the DC bus power failure warning system combined with battery detection is also used to: use linear interpolation to unify the sampling rates of the N device historical status data to obtain N calibrated historical status data; use a sliding time window to extract transient features of the N device historical status data to obtain N voltage sag rate sequences and N current interruption node sequences; construct a correlation coefficient matrix based on the N voltage sag rate sequences; calculate the synchronous power outage rate based on the N current interruption node sequences; use the correlation coefficient matrix and the synchronous power outage rate to perform fusion quantification of the device side operation consistency to obtain the second risk authenticity.

[0079] Furthermore, the DC bus power failure warning system combined with battery detection is also used to: if the first power failure warning is an empty set, collect the first usage time of the main DC bus; determine whether the first usage time meets the preset switching time scale; if the first usage time meets the preset switching time scale, switch the backup DC bus to connect to the battery for power supply.

[0080] Furthermore, the DC bus power failure warning system combined with battery detection is also used to: traverse the non-diagonal elements of the correlation coefficient matrix to obtain the first element quantity that satisfies the preset linear correlation; calculate the proportion of the first element quantity in the device logarithm of the matrix elements; when the device logarithm proportion meets the preset first percentage and the synchronous power outage rate meets the preset second percentage, weightedly fuse the device logarithm proportion and the synchronous power outage rate as the second risk authenticity output.

[0081] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The DC bus power failure warning method combined with battery detection and the specific examples in the aforementioned embodiment 1 are also applicable to the DC bus power failure warning system combined with battery detection in this embodiment. Through the aforementioned detailed description of the DC bus power failure warning method combined with battery detection, those skilled in the art can clearly understand the DC bus power failure warning system combined with battery detection in this embodiment. Therefore, for the sake of brevity of the specification, it will not be described in detail here.

[0082] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

[0083] Obviously, those skilled in the art may make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application is intended to include these modifications and variations.

Claims

1. A DC bus power failure warning method combined with battery detection is characterized in that: include: By monitoring the multi-element time series characteristics of the main DC bus, and predicting the risks on the DC bus side based on the monitoring results, the first power failure warning is output; If the first power failure warning is a non-empty set, after switching the backup DC bus to connect to the battery for power supply, the battery side status data is backtracked according to the first power failure warning matching the first warning time scale to obtain the battery historical status data of the battery; performing a credibility check on the first power failure warning based on the battery historical status data, and outputting a first risk authenticity; Verify demand matching based on the first risk authenticity to obtain N electrical equipment; Backtracking device-side status data according to the first warning time scale to obtain N device historical status data of the N electrical devices; quantifying the device-side operation consistency based on the N device historical status data to obtain a second risk authenticity; After the power failure warning level is quantified according to the second risk reality, the power failure warning level is used to perform operation and maintenance output of the first power failure warning.

2. The DC bus power failure warning method combined with battery detection according to claim 1 is characterized in that: By monitoring the multi-element time-series characteristics of the main DC bus, and predicting the risks on the DC bus side based on the monitoring results, the first power failure warning is output, including: Setting the power failure prediction scale according to the power failure switching response delay; Using K times the power failure prediction scale as a time constraint, the main DC bus is monitored for multivariate time series characteristics to obtain bus voltage fluctuation rate, current harmonic distortion rate, and temperature gradient change rate; The bus voltage fluctuation rate, current harmonic distortion rate and temperature gradient change rate are analyzed through multi-modal time series coupling to predict the risk on the DC bus side and output the first power failure warning.

3. The DC bus power failure warning method combined with battery detection according to claim 2, characterized in that: Analyzing the bus voltage fluctuation rate, current harmonic distortion rate, and temperature gradient change rate through multi-modal time series coupling to pre-judge the risk on the DC bus side and output the first power failure warning includes: Using the unique identifiers of multiple electrical devices and batteries as search conditions, network data collection is performed to obtain sample time-series power outage records; Taking the power failure node as the sampling starting point and K+1 times the power failure prediction scale as the sampling scale, reverse data collection of the sample time series power failure record is performed to obtain multiple positive sample multivariate time series features, wherein the positive sample multivariate time series features include a positive sample voltage fluctuation rate, a positive sample harmonic distortion rate, and a positive sample gradient change rate; After removing the multiple positive sample multivariate time series features from the sample time series power failure record, randomly sampling the sample time series power failure record with K+1 times the power failure prediction scale as the sampling scale to obtain multiple negative sample multivariate time series features; Using the multiple positive sample multivariate time series features and the multiple negative sample multivariate time series features as training data, optimizing the parameters of a power failure prediction model, wherein the power failure prediction model is constructed based on an LSTM model; The bus voltage fluctuation rate, current harmonic distortion rate and temperature gradient change rate are input into the power failure prediction model to pre-judge the risk on the DC bus side, and the first power failure warning is output.

4. The DC bus power failure warning method combined with battery detection according to claim 2, characterized in that: Using K times the power failure prediction scale as a time constraint, the main DC bus is monitored for multivariate time series characteristics to obtain bus voltage fluctuation rate, current harmonic distortion rate, and temperature gradient change rate, including: A voltage sensor connected in parallel between the positive and negative poles of the DC bus performs voltage monitoring at a preset first sampling frequency to obtain a voltage instantaneous value sequence; A current waveform sampling sequence is obtained by performing current monitoring at a preset second sampling frequency through a Hall current sensor connected in series to the DC bus; A temperature distribution matrix is ​​obtained by performing time-varying temperature monitoring at a preset third sampling frequency using a PT100 sensor array deployed along the DC bus; Extracting the numerical variation characteristics of the voltage instantaneous value sequence and the current waveform sampling sequence respectively to obtain the bus voltage fluctuation rate and the current harmonic distortion rate; Performing a spatiotemporal gradient analysis on the temperature distribution matrix to obtain the temperature gradient change rate.

5. The DC bus power failure warning method combined with battery detection according to claim 4 is characterized in that: The time-aligned voltage instantaneous value sequence, current waveform sampling sequence and temperature distribution matrix meet K times the power failure prediction scale.

6. The DC bus power failure warning method combined with battery detection according to claim 1, characterized in that: Performing a credibility check on the first power failure warning based on the battery historical status data and outputting a first risk authenticity includes: Interactively obtain multiple sample power-off status data; Obtaining multiple sample power failure probabilities by performing state backtracking on the multiple sample power failure state data; The knowledge graph is used to associate and store the multiple sample power failure state data and the multiple sample power failure probabilities to complete the construction of the power failure state library; Inputting the battery historical state data into the power failure state library for state similarity matching to obtain H sample power failure state data and H sample power failure probabilities that fall within a state similarity threshold; According to H state similarity parameters between the battery historical state data and H sample power-off state data, a weighted calculation is performed on the H sample power-off probabilities to output the first risk authenticity.

7. The DC bus power failure warning method combined with battery detection according to claim 1, characterized in that: Quantifying the device-side operation consistency based on the N device historical status data to obtain a second risk authenticity includes: Unifying the sampling rates of the N historical status data of the devices by using linear interpolation to obtain N calibration historical status data; Using a sliding time window to extract transient features of the N device historical status data, to obtain N voltage sag rate sequences and N current interruption node sequences; Constructing a correlation coefficient matrix according to the N voltage sag rate sequences; Calculating a synchronous power outage rate according to the N current interruption node sequences; The correlation coefficient matrix and the synchronous power failure rate are used to perform fusion quantification of the equipment side operation consistency to obtain the second risk authenticity.

8. The DC bus power failure warning method combined with battery detection according to claim 1, characterized in that: Also includes: If the first power failure warning is an empty set, collecting a first usage time of the main DC bus; Determining whether the first usage duration meets a preset switching time scale; If the first usage time meets the preset switching time scale, the backup DC bus is switched to connect to the battery for power supply.

9. The DC bus power failure warning method combined with battery detection according to claim 7, characterized in that: The correlation coefficient matrix and the synchronous power failure rate are used to perform fusion quantification of the device-side operation consistency to obtain the second risk authenticity, including: Traversing the matrix off-diagonal elements of the correlation coefficient matrix to obtain a first element quantity that satisfies a preset linear correlation; Calculate the proportion of the first element quantity in the device logarithm of the matrix elements; When the proportion of the number of device pairs meets a preset first percentage and the synchronous power outage rate meets a preset second percentage, the proportion of the number of device pairs and the synchronous power outage rate are weighted and fused to output as the second risk authenticity.

10. The DC bus power failure warning system combined with battery detection is characterized by: The steps for implementing the DC bus power failure warning method combined with battery detection as described in any one of claims 1 to 9 include: The risk prediction module is used to monitor the multi-element time series characteristics of the main DC bus, predict the risk of the DC bus side based on the monitoring results, and output the first power failure warning; a battery status backtracking module configured to, if the first power failure warning is a non-empty set, then after switching the backup DC bus to connect to the battery for power supply, perform battery-side status data backtracking based on the first power failure warning and matching the first warning time scale to obtain battery historical status data of the battery; a credibility verification module, configured to perform a credibility verification on the first power failure warning based on the battery historical status data, and output a first risk authenticity; A verification demand matching module, configured to perform verification demand matching according to the first risk authenticity to obtain N electrical devices; an equipment status backtracking module, configured to backtrack equipment-side status data according to the first warning time scale to obtain N equipment historical status data of the N electrical devices; an operation consistency quantification module, configured to quantify the operation consistency of the device side according to the N device historical status data to obtain a second risk authenticity; The operation and maintenance output module is used to, after quantifying the power failure warning level according to the second risk reality, use the power failure warning level to perform operation and maintenance output of the first power failure warning.

Images