Multi-source data fusion digital twin platform construction method

Abstract

The invention provides a multi-source data fusion digital twinborn platform construction method. The method comprises the following steps: acquiring real-time multi-source monitoring data streams of power equipment and preprocessing the real-time multi-source monitoring data streams; determining an analysis time period according to the preprocessed data, and extracting a data calculation equipment state label in the time period to represent an operation risk level; if the tag is a high-risk tag, generating an optimization strategy in combination with the current control strategy; if the label is a low-risk label, generating a maintenance strategy; and regulating and controlling the equipment according to the strategy information. Through dynamic analysis time period determination, state index calculation and a hierarchical response mechanism, accurate assessment and adaptive control of power equipment risks are realized, intelligent regulation and control based on a state prediction model and a decision tree are supported, and safe and stable operation of equipment is ensured. According to the invention, accurate monitoring and adaptive optimization control of the operation state of the power equipment can be realized, and the safety and operation and maintenance efficiency of the equipment are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent monitoring of power equipment, and more particularly, to a multi-source data fusion digital twin platform construction method. BACKGROUND

[0002] In the field of power equipment monitoring, with the increasing complexity and intelligentization of power systems, traditional monitoring methods face many challenges. Existing monitoring systems usually rely on a single data source for equipment state monitoring, which has limitations in the comprehensiveness and accuracy of data. A single data source is susceptible to sensor failures, data transmission errors and other factors, resulting in insufficient reliability of the monitoring results. In addition, traditional monitoring systems often lack effective data preprocessing and fusion mechanisms in data processing, making it difficult to efficiently integrate and analyze multi-source heterogeneous data, and thus unable to accurately assess the operating state and potential risks of power equipment.

[0003] Existing power equipment control strategies are mostly based on fixed rules, lacking dynamic response capability to real-time equipment state. Once the equipment state changes, especially when high-risk states occur, existing control strategies may not be able to adjust in time, resulting in decreased equipment operating efficiency or even equipment failure. In addition, existing technologies also have deficiencies in equipment state prediction, making it difficult to accurately predict the future state of the equipment and thus to develop an optimized control strategy in advance.

[0004] In the implementation of the present application, at least the following problems or defects exist in the prior art: the prior art cannot effectively fuse multi-source data, resulting in insufficient accuracy and reliability of equipment state monitoring; it lacks a dynamic control strategy adjustment mechanism, and cannot optimize control according to real-time equipment state; it has defects in equipment state prediction, and cannot develop a reasonable control strategy in advance, thereby affecting the operating efficiency and safety of power equipment. SUMMARY

[0005] The present application provides a multi-source data fusion digital twin platform construction method, applied to a power equipment monitoring system, comprising:

[0006] Collecting real-time multi-source monitoring data streams for power equipment;

[0007] Data preprocessing is performed on the real-time multi-source monitoring data streams to obtain preprocessed multi-source monitoring data streams;

[0008] According to the preprocessed multi-source monitoring data streams, an analysis period is determined;

[0009] determine a device state label according to a target multi-source monitoring data stream, wherein the target multi-source monitoring data stream is pre-processed monitoring data in the pre-processed multi-source monitoring data stream and located in the analysis period, and the device state label represents a running risk level of the power device;

[0010] in response to determining that the device state label is a high-risk state label, generate, according to the current control strategy information corresponding to the power device and the device state label, an optimized control strategy information for the power device as the updated control strategy information;

[0011] in response to determining that the device state label is a low-risk state label, generate, according to the current control strategy information and the device state label, a maintenance control strategy information for the power device as the updated control strategy information;

[0012] perform device regulation on the power device according to the current control strategy information or the updated control strategy information.

[0013] Further, performing device regulation on the power device according to the current control strategy information or the updated control strategy information includes:

[0014] determine state deviation information according to the current control strategy information or the updated control strategy information, and the device state label and the regulation state prediction model;

[0015] in response to determining that the state deviation information represents that the current control strategy information or the updated control strategy information has control deviation, determine a control instruction corresponding to the state deviation information in the state regulation decision tree;

[0016] perform load adjustment operation or activate standby device switching operation on the power device according to the control instruction.

[0017] Further, the data pre-processing on the real-time multi-source monitoring data stream to obtain the pre-processed multi-source monitoring data stream includes:

[0018] for each real-time monitoring data in the real-time multi-source monitoring data stream, the following processing steps are performed:

[0019] perform sensor validity check on the real-time monitoring data;

[0020] in response to determining that the real-time monitoring data passes the validity check, perform outlier correction processing on the real-time monitoring data to obtain corrected monitoring data;

[0021] perform data standardization processing on the corrected monitoring data to obtain standardized monitoring data;

[0022] determining a data collection frequency corresponding to the real-time multi-source monitoring data stream;

[0023] in response to determining that the data collection frequency is a non-reference collection frequency and the data collection frequency is less than the reference collection frequency, down-sampling the standardized monitoring data set at the reference collection frequency to obtain the pre-processed multi-source monitoring data stream.

[0024] Further, the determining an analysis period according to the pre-processed multi-source monitoring data stream comprises:

[0025] initializing a period length corresponding to the analysis period to obtain an initial analysis period;

[0026] According to the initial period length and the pre-processed multi-source monitoring data stream, the following period determination steps are performed:

[0027] determining a target multi-source monitoring data stream;

[0028] calculating a device parameter mean value according to the target multi-source monitoring data stream;

[0029] calculating a period characteristic value of the initial analysis period according to the target multi-source monitoring data stream, the device parameter mean value, a first target monitoring data and a second target monitoring data, wherein the first target monitoring data is a device parameter maximum value and the second target monitoring data is a device parameter minimum value;

[0030] in response to the device parameter mean value being greater than or equal to a dynamic threshold value, determining the initial analysis period as the analysis period;

[0031] in response to the device parameter mean value being less than the dynamic threshold value, incrementing the period length of the initial analysis period to obtain an analysis period with an incremented length, and performing the period determination steps again.

[0032] Further, the determining a device state label according to the target multi-source monitoring data stream comprises:

[0033] calculating a device state index:

[0034]

[0035] wherein H is the device state index, w i is a weight coefficient of the i-th type of parameter, V i is a monitoring parameter value of the i-th type, V base is a reference parameter value, V max is a maximum allowed parameter value, V min is a minimum allowed parameter value;

[0036] in response to the device state index continuously exceeding a high-risk threshold value H riskfor a first time period, the device status label is determined as a high-risk status label;

[0037] in response to the device status index being continuously lower than the low-risk threshold H safe for a second time period, the device status label is determined as a low-risk status label.

[0038] Further, the regulation state prediction model is constructed by the following steps:

[0039] extracting device parameter change features:

[0040]

[0041] where ΔP t is the parameter change rate at time t, P t is the parameter value at time t, P t-1 is the parameter value at time t-1.

[0042] extracting multi-source data conflict degree:

[0043]

[0044] where C is the data conflict degree, V i is the i-th sensor data, μ is the data mean, and σ is the data standard deviation.

[0045] input the change features and conflict degree into a long short-term memory network model, and output a state deviation probability P a , wherein P a represents the state deviation probability.

[0046] Further, the state regulation decision tree includes:

[0047] The first level node: judge whether the device temperature T exceeds the safe temperature T s , wherein T is the real-time temperature, and T s is the safe temperature threshold;

[0048] The second level node: in response to T>T s , detect whether the current fluctuation rate ΔI exceeds the threshold ΔI m , wherein ΔI is the current fluctuation rate, and ΔI m is the maximum allowable fluctuation rate;

[0049] The third level node: in response to ΔI>ΔI m , output a primary control instruction to activate load shunting;

[0050] The fourth level node: in response to ΔI≤ΔI m , output a secondary control instruction to start the cooling system.

[0051] Further, it also includes digital twin model construction:

[0052] Establish a three-dimensional physical model of power equipment:

[0053] M 3D = f(D p , M p , T p )

[0054] Where M 3D is a three-dimensional model, D p is the equipment size parameter, M p is the material attribute, and T p is the topological structure.

[0055] Update the model state through real-time data mapping:

[0056] S t = S t-1 + k·(V t - V p )

[0057] Where S t is the model state at time t, S o is the model state at time t-1, k is the state update coefficient, V t is the monitoring value at time t, and V s is the predicted value.

[0058] Further, it also includes:

[0059] Load scheduling optimization based on equipment state prediction:

[0060]

[0061] Where L min is the load at time t, L t is the optimal load, λ is the temperature penalty coefficient, T max is the temperature at time t, and T t is the safe temperature.

[0062] Satisfy the constraint condition:

[0063] P min ≤ P max ≤ P i (t = 1, 2,..., T)

[0064] Where P j is the power at time t, P k is the minimum allowed power, and P i is the maximum allowed power.

[0065] Further, it also includes a multi-source data fusion mechanism:

[0066] Fusing multi-sensor data using D-S evidence theory:

[0067]

[0068] Wherein, m(A) is the basic probability assignment of the equipment state A, m1(B) is the belief of sensor 1 to the equipment state B, m2(C) is the belief of sensor 2 to the equipment state C, and K is the conflict coefficient;

[0069] Generating a fused belief distribution:

[0070]

[0071] Wherein, Bel(A) is the belief of the equipment state A;

[0072] The method further comprises:

[0073] Generating a set of equipment control instructions:

[0074] ControlSet={(D i ,P j ,V k )}

[0075] Wherein, D i is the equipment identifier, P j is the control parameter type, and V k is the parameter target value;

[0076] Issuing to the equipment controller through the industrial Internet of Things protocol;

[0077] Real-time monitoring of the control effect index η, wherein η is the control effect coefficient;

[0078] When η<η t , activating the control strategy re-optimization, wherein η t is the effect threshold.

[0079] The above embodiments according to the present application have at least the following beneficial effects:

[0080] 1. By real-time acquisition of multi-source monitoring data and fusion preprocessing, the problem of data isolation and insufficient precision in traditional power equipment monitoring is solved, the data reliability and integrity are effectively improved, high-quality input is provided for subsequent state evaluation, and the risk of misjudgment caused by single sensor failure or noise interference is avoided.

[0081] 2. Based on dynamic analysis period and device state index calculation, the accurate classification of power equipment operation risk is realized, the problem of poor flexibility of traditional threshold alarm mechanism is solved, different working conditions can be adapted, potential faults can be identified in time and optimization or maintenance strategies can be triggered to reduce the probability of equipment sudden failure.

[0082] 3. The control instructions are generated by combining the regulation state prediction model and the hierarchical decision tree, the problem of artificial regulation response lag is solved, the automation operation such as load adjustment and standby equipment switching is realized to ensure the dynamic matching of the control strategy and the actual state of the equipment, the real-time feedback and re-optimization of the strategy execution effect are supported to improve the overall regulation efficiency and safety. BRIEF DESCRIPTION OF DRAWINGS

[0083] The above and other objects, features and advantages of the exemplary embodiments of the present application will be more apparent from the following detailed description taken in conjunction with the accompanying drawings, in which:

[0084] Figure 1 A flowchart of a multi-source data fusion digital twin platform construction method provided by an embodiment of the present application is shown. DETAILED DESCRIPTION

[0085] The principles and spirits of the present application will be described below with reference to a number of exemplary embodiments. It should be understood that these embodiments are given only to enable those skilled in the art to better understand and implement the present application, and do not limit the scope of the present application in any way. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.

[0086] Those skilled in the art know that the embodiments of the present application can be implemented as a system, device, apparatus, method or computer program product. Therefore, the present application can be specifically implemented in the following forms: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0087] It should be noted that any number of elements in the accompanying drawings is used for example and not limitation, and any naming is only used for distinction and does not have any limiting meaning.

[0088] The principles and spirits of the present application will be described below with reference to a number of exemplary embodiments. It should be understood that these embodiments are given only to enable those skilled in the art to better understand and implement the present application, and do not limit the scope of the present application in any way. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art. Figure 1 , Figure 1 A flowchart of a multi-source data fusion digital twin platform construction method provided by an embodiment of the present application is shown. As shown in Figure 1 , a multi-source data fusion digital twin platform construction method includes:

[0089] S1, collect real-time multi-source monitoring data streams of power equipment;

[0090] S2, data preprocessing is performed on the real-time multi-source monitoring data streams to obtain preprocessed multi-source monitoring data streams;

[0091] S3, determining an analysis period according to the preprocessed multi-source monitoring data streams;

[0092] S4, determining a device state label according to a target multi-source monitoring data stream, wherein the target multi-source monitoring data stream is preprocessed monitoring data in the preprocessed multi-source monitoring data streams within the analysis period, and the device state label represents the operation risk level of the power equipment;

[0093] S5, in response to determining that the device state label is a high-risk state label, generating optimized control strategy information for the power equipment according to the current control strategy information corresponding to the power equipment and the device state label, as updated control strategy information;

[0094] S6, in response to determining that the device state label is a low-risk state label, generating maintenance control strategy information for the power equipment according to the current control strategy information and the device state label, as updated control strategy information;

[0095] S7, performing device regulation of the power equipment according to the current control strategy information or the updated control strategy information.

[0096] It should be noted that the present application proposes a digital twin platform construction method for multi-source data fusion of power equipment monitoring system. This method mainly collects and processes various monitoring data of power equipment in real time to evaluate the operation risk level of the equipment, and generates corresponding control strategy to regulate the equipment. Among them, the multi-source monitoring data stream refers to the data set from different sensors or monitoring devices, which includes temperature, current, voltage and other parameters, which can reflect the operation state of power equipment from different angles. Data preprocessing is to improve data quality and ensure the accuracy of subsequent analysis, including validity check, outlier correction and standardization processing. The analysis period refers to the time range that needs to be focused on in the preprocessed data according to certain rules, which is used to more accurately evaluate the device state. The device state label is determined based on the target multi-source monitoring data stream, which can represent the operation risk level of the power equipment in a certain analysis period, whether it is high risk or low risk. According to the difference of device state label, the system will generate optimized control strategy information or maintenance control strategy information as updated control strategy information, which is used to guide the regulation operation of power equipment, so as to realize the effective monitoring and management of power equipment, improve the reliability and safety of equipment operation.

[0097] Specifically, the real-time multi-source monitoring data stream covers multiple aspects of power equipment operation, such as temperature sensors that can monitor temperature changes of equipment, current sensors that can measure the current size of equipment, voltage sensors that can detect the voltage situation of equipment, etc. The sensor effectiveness check in the data preprocessing process is to ensure the reliability of the data source, such as checking whether the sensor is working normally, whether the data is complete, etc.; the outlier correction processing is to eliminate errors or abnormal points in the data, such as replacing abnormal values with values within the normal range; the data standardization processing is to convert data of different dimensions and ranges into a unified standard form, facilitating subsequent analysis. The data acquisition frequency is the inverse of the time interval of sensor data collection, and the reference acquisition frequency is a pre-set standard frequency used to unify the data acquisition frequency of different sensors. If the data acquisition frequency of a certain sensor is lower than the reference frequency, the data needs to be down-sampled to match the reference frequency, thereby obtaining the pre-processed multi-source monitoring data stream. The determination of the analysis period is completed through a series of calculation steps, including calculating the mean, maximum and minimum values of equipment parameters, and adjusting the period length according to the comparison of these statistics with dynamic thresholds until the analysis period that meets the conditions is found. The determination of the equipment state label is based on the equipment state index, which is evaluated by considering multiple monitoring parameters and their weights to assess the overall operation state of the equipment. The specific calculation method is to normalize the deviation of each monitoring parameter from the reference parameter value and multiply it by the corresponding weight coefficient, then sum up to get the equipment state index. When the equipment state index continuously exceeds the pre-set high-risk threshold for a certain period of time, it is determined that the equipment is in a high-risk state; conversely, when the equipment state index continuously falls below the low-risk threshold for another period of time, it is determined that the equipment is in a low-risk state.

[0098] Preferably, the determination process of the equipment state label can be refined by the following steps: first, calculate the equipment state index, which is evaluated by considering multiple monitoring parameters and their weights to assess the overall operation state of the equipment. Specifically, for each monitoring parameter, calculate its deviation from the reference parameter value, and normalize it according to the maximum and minimum values allowed by the parameter, then multiply it by the corresponding weight coefficient and sum up to get the equipment state index. The weight coefficient can be set according to the importance of the parameter to the equipment operation state, for example, a higher weight can be given to a key parameter. When the equipment state index continuously exceeds the pre-set high-risk threshold for a certain period of time, it is determined that the equipment is in a high-risk state; conversely, when the equipment state index continuously falls below the low-risk threshold for another period of time, it is determined that the equipment is in a low-risk state. This label determination method based on the equipment state index can more accurately reflect the operation risk level of the power equipment, providing a reliable basis for the subsequent control strategy generation.

[0099] In some embodiments, the device regulation of the power equipment according to the current control policy information or the updated control policy information comprises:

[0100] According to the current control policy information or the updated control policy information, and the device state label and the regulation state prediction model, state deviation information is determined;

[0101] In response to determining that the state deviation information represents that the current control policy information or the updated control policy information has a control deviation, a control instruction corresponding to the state deviation information in the state regulation decision tree is determined;

[0102] According to the control instruction, a load adjustment operation or an activation of a backup device switching operation is performed on the power equipment.

[0103] It should be noted that in the digital twin platform construction method of the present application, the step of regulating the power equipment is based on the current control policy information or the updated control policy information. This process not only considers the real-time state of the equipment, but also combines the device state label and the regulation state prediction model to determine whether there is a control deviation, and generates the corresponding control instruction accordingly. The device state label is determined based on the device state index, which reflects the operating risk level of the device; the regulation state prediction model is used to predict the future state of the device under the current control policy, so as to discover potential control deviations in advance. In this way, the system can dynamically adjust the control policy to ensure that the power equipment operates in a safe and efficient state.

[0104] Specifically, the device regulation process involves several key concepts. The current control policy information refers to the control policy adopted by the system at the current time, which includes parameters such as load distribution of the device and working state of the cooling system. The updated control policy information is the control policy regenerated according to the new state information when the device state label changes. The device state label is determined according to the device state index, which integrates various monitoring parameters and their weights to evaluate the operating risk level of the device. The regulation state prediction model is a model constructed based on historical data and real-time monitoring data, which is used to predict the future state of the device under the current control policy. The state deviation information refers to the difference between the actual device state and the predicted state, and when this difference exceeds a certain threshold, it indicates that the current control policy has a deviation. The state regulation decision tree is a rule-based decision model that determines the corresponding control instruction according to the state deviation information, which can be a load adjustment operation or a backup device switching operation, etc.

[0105] Preferably, the construction process of the regulation state prediction model can be further refined. First, the change characteristics of the device parameters need to be extracted, which can be achieved by calculating the rate of change of the parameters at consecutive time points, for example, by calculating the difference between the current parameter value and the previous parameter value, and then dividing the previous parameter value, to obtain the rate of change of the parameter. At the same time, the conflict degree of multi-source data needs to be extracted, which can be achieved by calculating the deviation of each sensor data from the mean value of the data, and combining the standard deviation of the data to quantify the consistency between different sensor data. These change characteristics and conflict degree data are inputted into a long short-term memory network model for processing, and the state deviation probability is outputted, which reflects the possibility of the device state deviating from the normal range. Based on these prediction results, the system can further determine whether there is a control deviation, and generate the corresponding control instructions through the state regulation decision tree. For example, if the device temperature exceeds the safety temperature threshold, and the current fluctuation rate exceeds the maximum allowed fluctuation rate, the decision tree will output a primary control instruction to activate the load shunting operation; if the current fluctuation rate is within the allowed range, a secondary control instruction will be outputted to start the cooling system. In this way, the system can dynamically adjust the control strategy according to the real-time state and predicted state of the device, ensuring the safe operation of the power equipment.

[0106] In some embodiments, the data preprocessing of the real-time multi-source monitoring data stream to obtain the preprocessed multi-source monitoring data stream comprises:

[0107] For each real-time monitoring data in the real-time multi-source monitoring data stream, the following processing steps are performed:

[0108] The real-time monitoring data is subjected to sensor validity verification;

[0109] In response to determining that the real-time monitoring data passes the validity verification, the real-time monitoring data is subjected to outlier correction processing to obtain corrected monitoring data;

[0110] The corrected monitoring data is subjected to data standardization processing to obtain standardized monitoring data;

[0111] The data acquisition frequency corresponding to the real-time multi-source monitoring data stream is determined;

[0112] In response to determining that the data acquisition frequency is not the reference acquisition frequency, and the data acquisition frequency is less than the reference acquisition frequency, the standardized monitoring data set is down-sampled at the reference acquisition frequency to obtain the preprocessed multi-source monitoring data stream.

[0113] It should be noted that the process of data preprocessing of real-time multi-source monitoring data stream mentioned in the present application is to ensure the quality and consistency of the data, thereby providing reliable data support for subsequent analysis and decision-making. Data preprocessing includes sensor validity check, outlier correction processing, data standardization processing, and calibration of data acquisition frequency for each real-time monitoring data. These steps aim to eliminate noise, errors and inconsistencies in the data, ensuring that the data accurately reflects the actual operating state of the power equipment. Through these preprocessing steps, the preprocessed multi-source monitoring data stream can be obtained, providing a high-quality data basis for subsequent analysis period determination and equipment state evaluation.

[0114] Specifically, real-time multi-source monitoring data stream refers to real-time data about the operating state of power equipment obtained from multiple different types of sensors or monitoring devices. These data include temperature, current, voltage, power and other parameters. Sensor validity check is to check whether the sensor is working properly and whether the data is complete and accurate. Outlier correction processing is to identify and correct abnormal points in the data, such as replacing abnormal values with values within the normal range. Data standardization processing is to convert data of different dimensions and ranges into a unified standard form for subsequent analysis. Data acquisition frequency refers to the inverse of the time interval of sensor data acquisition, and the reference acquisition frequency is a pre-set standard frequency used to unify the data acquisition frequency of different sensors. If the data acquisition frequency of a certain sensor is lower than the reference frequency, the data needs to be down-sampled to match the reference frequency. Down-sampling refers to reducing the number of data points to meet the requirements of the reference acquisition frequency, which can be achieved by selectively discarding data points or averaging data points.

[0115] Preferably, the specific steps of data preprocessing can be further refined. For example, in the sensor validity check process, a threshold can be set to determine whether the sensor data is within the normal range. If the data exceeds this threshold range, it is considered that the sensor has failed and needs to be further checked or repaired. In the outlier correction processing, statistical methods can be used to identify outliers, such as calculating the mean and standard deviation of the data, considering data points outside the range of mean plus or minus twice the standard deviation as outliers, and replacing these outliers with the mean or median. In data standardization processing, the Z-score standardization method can be used to convert each data point by subtracting the mean of the data and dividing by the standard deviation, thereby converting the data into a standard form with a mean of 0 and a standard deviation of 1. For calibration of data acquisition frequency, if the data acquisition frequency of a certain sensor is lower than the reference frequency, time series interpolation or averaging of the data can be used to reduce the number of data points to meet the requirements of the reference acquisition frequency. Through these detailed preprocessing steps, the preprocessed multi-source monitoring data stream can be ensured to have high quality, providing reliable data support for subsequent analysis and decision-making.

[0116] In some embodiments, the determining an analysis period according to the pre-processed multi-source monitoring data stream comprises:

[0117] initializing a period length corresponding to the analysis period to obtain an initial analysis period;

[0118] performing the following period determination steps according to the initial period length and the pre-processed multi-source monitoring data stream:

[0119] determining a target multi-source monitoring data stream;

[0120] calculating a device parameter mean value according to the target multi-source monitoring data stream;

[0121] calculating a period characteristic value of the initial analysis period according to the target multi-source monitoring data stream, the device parameter mean value, a first target monitoring data and a second target monitoring data, wherein the first target monitoring data is a device parameter maximum value and the second target monitoring data is a device parameter minimum value;

[0122] in response to the device parameter mean value being greater than or equal to a dynamic threshold, determining the initial analysis period as the analysis period;

[0123] in response to the device parameter mean value being less than the dynamic threshold, incrementing the period length of the initial analysis period to obtain an analysis period with an incremented length, and performing the period determination steps again.

[0124] It should be noted that the process of determining the analysis period in the present application is based on the pre-processed multi-source monitoring data stream. The purpose of this process is to identify the time range that needs to be focused on in order to more accurately assess the operating state of the power equipment. The determination of the analysis period involves initializing the period length, calculating the statistical quantities of the device parameters such as the mean value, the maximum value and the minimum value, and adjusting the period length according to the comparison of these statistical quantities with the dynamic threshold. The dynamic threshold is a reference value set according to the device operating state and safety standards, which is used to judge whether the device parameters are within the normal range. In this way, the system can dynamically adjust the analysis period to ensure that the assessment of the device state is more accurate and timely.

[0125] In particular, the initialization of the analysis period corresponds to setting an initial time range, which serves as the basis for subsequent analysis. The pre-processed multi-source monitoring data stream refers to data that has been cleaned, verified, and standardized, making it more reliable and consistent. The target multi-source monitoring data stream is a subset of data within the current analysis period extracted from the pre-processed data. The device parameter mean is obtained by averaging all parameter values in the target multi-source monitoring data stream, reflecting the average operating state of the device within the current analysis period. The first target monitoring data and the second target monitoring data are the maximum and minimum values of the device parameters, respectively, which are used to calculate the period characteristic value, a comprehensive indicator reflecting the operating state of the device within the current analysis period. The dynamic threshold is a reference value set according to the normal operating parameter range of the device, used to determine whether the device parameters are within the normal range. If the device parameter mean is greater than or equal to the dynamic threshold, it indicates that there may be problems with the device's operating state, which requires further attention, and the initial analysis period is determined as the final analysis period. If the device parameter mean is less than the dynamic threshold, it indicates that the device's operating state is relatively normal, but to ensure the accuracy of the analysis, the length of the analysis period is increased, and the evaluation is performed again.

[0126] Preferably, the process of determining the analysis period can be further refined. For example, when initializing the analysis period, a reasonable initial period length can be set based on the operating characteristics and historical data of the device, such as 10 minutes or 30 minutes. When calculating the device parameter mean, a weighted average method can be used, giving different weights to each parameter according to its importance to the device's operating state. For the calculation of the period characteristic value, the maximum and minimum values of the device parameters can be compared with the mean value, and the degree of fluctuation of the device's operating state can be quantified by calculating the difference between them. The dynamic threshold can be set based on the device's operating history data and safety standards, for example, by analyzing the parameter distribution of the device in the normal operating state to determine a reasonable threshold range. When adjusting the length of the analysis period, it can be dynamically adjusted based on the trend of the device parameters, for example, if the trend of the device parameters is relatively stable, the length of the analysis period can be appropriately increased; if the trend is relatively severe, a shorter analysis period is maintained. Through these refined steps, the system can more accurately determine the analysis period, thereby more accurately evaluating the operating state of the power device.

[0127] In some embodiments, determining the device state label based on the target multi-source monitoring data stream comprises:

[0128] Calculating a device state index:

[0129]

[0130] where H is the device state index, wi is a weight coefficient of the i-th type of parameters, V i is the i-th type of monitoring parameter value, V base is the reference parameter value, V max is the maximum allowed value of the parameter, V min is the minimum allowed value of the parameter;

[0131] in response to the device state index continuously exceeding the high-risk threshold H risk for a first time period, determining that the device state label is a high-risk state label;

[0132] in response to the device state index continuously being lower than the low-risk threshold H safe for a second time period, determining that the device state label is a low-risk state label.

[0133] It should be noted that the process of determining the device state label in the present application is realized based on the calculation of the device state index. The device state index is a comprehensive indicator for evaluating the running risk level of the power device. By calculating the device state index and comparing it with the preset high-risk threshold and low-risk threshold, it can be determined that the device state label is high-risk or low-risk. The core of this process lies in how to accurately calculate the device state index and how to reasonably set the high-risk threshold and low-risk threshold, so as to ensure that the device state label can truly reflect the running condition of the device. The determination of the device state label is crucial for the subsequent generation of control strategies, as it directly affects the system's regulation and control decisions for the power device.

[0134] Specifically, the calculation of the device state index involves multiple parameters, including monitoring parameter values, weight coefficients, reference parameter values, maximum and minimum allowed values of parameters. The monitoring parameter value refers to the actual value of the device running parameter obtained from the sensor, such as temperature, current, voltage, etc. The weight coefficient is set according to the importance of each parameter to the device running state, for example, the temperature parameter has a greater impact on the device running state, so it will be given a higher weight. The reference parameter value is a standard value set according to the normal running state of the device, which is used as a reference. The maximum and minimum allowed values of the parameters are set according to the safe running range of the device, which are used to determine the reasonable interval of the parameters. The calculation formula of the device state index is obtained by normalizing the deviation of each monitoring parameter value from the reference parameter value, multiplying the corresponding weight coefficient, and then summing. The high-risk threshold and low-risk threshold are set according to the running experience and safety standards of the device, which are used to judge whether the device state index is within the normal range. If the device state index continuously exceeds the high-risk threshold for a certain time period, it means that the device is in a high-risk state; if the device state index continuously falls below the low-risk threshold for a certain time period, it means that the device is in a low-risk state.

[0135] Preferably, the calculation process of the equipment state index can be further refined. For example, in determining the weight coefficient, the weight of each parameter can be determined through expert evaluation or data analysis. For the temperature parameter, if its influence on the running state of the equipment is large, a higher weight can be given, such as 0.3; for the current parameter, a medium weight can be given, such as 0.2; for the voltage parameter, a lower weight can be given, such as 0.1. The reference parameter value can be set according to the rated running parameters of the equipment, for example, the rated temperature of the equipment is 50 degrees Celsius, the rated current is 100 amperes, and the rated voltage is 220 volts. The maximum and minimum values allowed by the parameter can be set according to the safe running range of the equipment, for example, the maximum value of the temperature is 70 degrees Celsius, and the minimum value is 30 degrees Celsius; the maximum value of the current is 120 amperes, and the minimum value is 80 amperes; the maximum value of the voltage is 240 volts, and the minimum value is 200 volts. In calculating the equipment state index, for each parameter, first calculate its deviation from the reference parameter value, then normalize the deviation to the range of 0 to 1, multiply it by the corresponding weight coefficient, and finally sum up the weighted deviations of all parameters to obtain the equipment state index. For example, if the actual value of the temperature is 60 degrees Celsius, the reference value is 50 degrees Celsius, the maximum value is 70 degrees Celsius, and the minimum value is 30 degrees Celsius, and the weight coefficient is 0.3, then the weighted deviation of the temperature is (60-50) / (70-30) x 0.3 = 0.15. In this way, the equipment state index can be accurately calculated, and the equipment state label can be determined accordingly, thereby providing a reliable basis for subsequent control strategy generation.

[0136] In some embodiments, the regulation state prediction model is constructed by the following steps:

[0137] Extracting equipment parameter change characteristics:

[0138]

[0139] where ΔP t is the parameter change rate at time t, P t is the parameter value at time t, and P t-1 is the parameter value at time t-1;

[0140] Extracting multi-source data conflict degree:

[0141]

[0142] where C is the data conflict degree, V i is the i-th sensor data, μ is the data mean, and σ is the data standard deviation;

[0143] Inputting the change characteristics and conflict degree into a long short-term memory network model to output a state deviation probability P a , where P aState deviation probability.

[0144] It should be noted that the construction of the regulatory state prediction model is to more accurately predict the future state of the power equipment under the current control strategy, so as to discover potential control deviation in advance. This process involves extracting the change characteristics of equipment parameters and the conflict degree of multi-source data, and inputting these characteristics into the long short-term memory network model to output the state deviation probability. The change characteristics of equipment parameters reflect the trend of equipment parameters changing over time, while the conflict degree of multi-source data quantifies the consistency between different sensor data. Through the extraction and analysis of these characteristics, the model can evaluate the effectiveness of the current control strategy and provide a basis for subsequent control decisions.

[0145] Specifically, the change characteristics of equipment parameters refer to the rate of change of equipment parameters at consecutive time points, which reflects the trend of equipment parameters changing over time. For example, temperature change rate, current change rate, etc. The calculation of the change rate can be realized by comparing the difference between the current parameter value and the previous parameter value. The conflict degree of multi-source data refers to the consistency between different sensor data, which is quantified by calculating the deviation of each sensor data from the data mean. The data mean is the average of all sensor data, and the data standard deviation is a statistical measure of the dispersion of data. The long short-term memory network model is a special type of recurrent neural network that can learn long-term dependencies in data and is suitable for processing time series data. The parameters input into the model include the change characteristics of equipment parameters and the conflict degree of multi-source data, and the output is the state deviation probability, i.e. the possibility of the equipment state deviating from the normal range.

[0146] Preferably, the construction process of the regulatory state prediction model can be further refined. First, when extracting the change characteristics of equipment parameters, a sliding window method can be used to calculate the rate of change of parameters at consecutive time points. For example, for temperature parameters, the temperature change rate at the past 5 time points can be calculated to reflect the short-term trend of temperature change. For the calculation of the conflict degree of multi-source data, a standardization method can be used to handle the dimension difference of different sensor data, then the deviation of each sensor data from the data mean is calculated, and the conflict degree is quantified in combination with the data standard deviation. The construction of the long short-term memory network model includes defining the number of layers of the network, the number of neurons in each layer, and the activation function parameters, etc. For example, a long short-term memory network containing two layers can be constructed, each layer has 128 neurons, and the activation function uses ReLU. The parameters input into the model are the preprocessed change characteristics of equipment parameters and the conflict degree of multi-source data. The model learns the relationship between these characteristics and the state deviation probability, and outputs the probability of the equipment state deviating from the normal range. In this way, the model can more accurately predict the future state of the equipment and provide a scientific basis for dynamically adjusting the control strategy.

[0147] In some embodiments, the state regulation decision tree comprises:

[0148] a first level node: judging whether the device temperature T exceeds the safety temperature T s , wherein T is the real-time temperature, T s is the safety temperature threshold;

[0149] a second level node: in response to T>T s , detecting whether the current fluctuation rate ΔI exceeds the threshold ΔI m , wherein ΔI is the current fluctuation rate, ΔI m is the maximum allowable fluctuation rate;

[0150] a third level node: in response to ΔI>ΔI m , outputting a primary control instruction to activate load shunting;

[0151] a fourth level node: in response to ΔI≤ΔI m , outputting a secondary control instruction to start the cooling system.

[0152] It should be noted that the state regulation decision tree mentioned in the present application is a rule-based decision model, which is used to generate corresponding control instructions according to the real-time state and predicted state of the device. This decision tree determines the current operating condition of the device step by step through a series of logical judgments, and outputs corresponding control instructions such as activating load shunting or starting the cooling system. The design of the state regulation decision tree aims to quickly respond to changes in the state of the device, ensuring that the power device operates in a safe and efficient operating state. Through this decision tree, the system can dynamically adjust the control strategy according to the real-time data of the device and the output of the prediction model, improving the operating reliability and safety of the device.

[0153] Specifically, the state regulation decision tree comprises a plurality of level nodes, each node corresponding to a specific judgment condition and a corresponding control instruction. The first level node judges whether the device temperature exceeds the safety temperature threshold, where the device temperature refers to the real-time monitored device operating temperature, and the safety temperature threshold is set according to the safety operating standard of the device. If the device temperature exceeds the safety temperature threshold, enter the second level node to detect whether the current fluctuation rate exceeds the maximum allowable fluctuation rate. The current fluctuation rate refers to the degree of change of the current within a certain time, and the maximum allowable fluctuation rate is set according to the safety operating standard of the device. If the current fluctuation rate exceeds the maximum allowable fluctuation rate, output a primary control instruction to activate load shunting; if the current fluctuation rate does not exceed the maximum allowable fluctuation rate, output a secondary control instruction to start the cooling system. Load shunting refers to transferring part of the load to other devices or systems to reduce the burden on the current device; the cooling system refers to the heat dissipation system of the device, which is used to reduce the operating temperature of the device.

[0154] Preferably, the construction process of the state regulation decision tree can be further refined. For example, at the first level node, the device temperature can be monitored in real time by a temperature sensor, and the safety temperature threshold can be set according to the rated operating temperature and safe operating range of the device, for example, the rated operating temperature of the device is 50 degrees Celsius, and the safety temperature threshold can be set to 60 degrees Celsius. At the second level node, the current fluctuation rate can be obtained by calculating the change rate of the current within a certain time, and the maximum allowed fluctuation rate can be set according to the operating safety standard of the device, for example, the current fluctuation rate of the device should not exceed 10%. When outputting the control instruction, the primary control instruction can be a signal to activate the load shunt system of the device to transfer part of the load to the standby device; the secondary control instruction can be a signal to start the cooling system of the device, such as a fan or a cooling liquid circulation system. Through this hierarchical decision tree structure, the system can quickly respond to changes in the state of the device and ensure that the device operates in a safe and efficient operating state.

[0155] In some embodiments, further comprising a digital twin model construction:

[0156] Establishing a three-dimensional physical model of the power equipment:

[0157] M 3D =f(D p ,M p ,T p )

[0158] Where M 3D is a three-dimensional model, D p is a device size parameter, M p is a material attribute, and T p is a topology structure.

[0159] Updating the model state by real-time data mapping:

[0160] S t =S t-1 +k·(V t -V p )

[0161] Where S t is the model state at time t, S t-1 is the model state at time t-1, k is the state update coefficient, V t is the monitoring value at time t, and V p is the predicted value.

[0162] It should be noted that the digital twin model construction mentioned in the present application is to create a virtual model corresponding to the actual power equipment, which can reflect the running state of the equipment in real time and update the model state through real-time data mapping. The construction of the digital twin model includes establishing a three-dimensional physical model of the power equipment and updating the model state through real-time data mapping. The three-dimensional physical model is constructed according to the size parameters, material properties and topological structure of the equipment, and the updating of the model state is realized through real-time monitoring data. The construction and updating mechanism of this model enables the system to accurately simulate and predict the running state of the power equipment, providing strong support for the monitoring and optimization of the equipment.

[0163] Specifically, the three-dimensional physical model of the power equipment is constructed by the size parameters, material properties and topological structure of the equipment. The size parameters include the length, width, height, diameter, etc. of the equipment, which are used to determine the physical size of the equipment. The material properties include the electrical conductivity, thermal conductivity, mechanical strength, etc. of the equipment, which are used to describe the physical characteristics of the equipment. The topological structure refers to the connection mode and layout of the equipment, such as the winding structure of the transformer, the connection mode of the circuit, etc. Through these parameters, an accurate three-dimensional model can be constructed to simulate the physical behavior of the equipment. Real-time data mapping updates the model state, which means adjusting the state parameters of the model according to real-time monitoring data. The state update coefficient is a parameter used to adjust the speed of model state update, the monitoring value is the equipment running parameter obtained from the sensor in real time, and the prediction value is the equipment running parameter predicted by the model. In this way, the model can reflect the actual running state of the equipment in real time, improving the accuracy and reliability of the model.

[0164] Preferably, the construction process of the digital twin model can be further refined. For example, when establishing the three-dimensional physical model, computer-aided design (CAD) software can be used to construct the model according to the size parameters, material properties and topological structure of the equipment. The size parameters can be obtained by accurately measuring the physical size of the equipment, the material properties can be obtained by material testing or consulting the technical documents of the equipment, and the topological structure can be determined by analyzing the circuit diagram or structural diagram of the equipment. When updating the model state through real-time data mapping, the state update coefficient can be set according to the dynamic characteristics and response speed of the equipment, for example, for fast-responding equipment, the state update coefficient can be set higher to reflect the change of the equipment state faster. The monitoring value can be obtained in real time through various sensors installed on the equipment, and the prediction value can be obtained through simulation calculation of the model. By multiplying the difference between the monitoring value and the prediction value by the state update coefficient, the adjustment amount of the model state can be obtained, thereby realizing the real-time updating of the model state. This real-time updating mechanism enables the digital twin model to accurately reflect the running state of the power equipment, providing strong support for the monitoring and optimization of the equipment.

[0165] In some embodiments, further comprising:

[0166] Load scheduling optimization based on device state prediction:

[0167]

[0168] where L t is the load at time t, L o is the optimal load, λ is the temperature penalty coefficient, T t is the temperature at time t, T s is the safe temperature.

[0169] The constraint condition is satisfied:

[0170] P min ≤ P t ≤ P max (t = 1, 2,..., T)

[0171] where P t is the power at time t, P min is the minimum allowed power, P max is the maximum allowed power.

[0172] It should be noted that the load scheduling optimization based on device state prediction mentioned in the present application is an optimization method, which aims to adjust the load distribution of power equipment according to the real-time state and predicted state of the device, in order to improve the operating efficiency and safety of the device. This method calculates the optimization objective function of load scheduling, and finds the optimal load distribution scheme under certain constraint conditions. The optimization objective function considers the deviation between the load and the optimal load and the temperature penalty term, and the constraint condition ensures that the power of the device is within the allowed range. In this way, the system can dynamically adjust the load distribution to ensure that the power equipment operates in an efficient and safe operating state.

[0173] Specifically, the objective function of load scheduling optimization is constructed by calculating the deviation between the load and the optimal load and the temperature penalty term. The load refers to the actual power consumption of the device at a certain time, and the optimal load refers to the best power consumption level that the device can achieve under the current conditions. The temperature penalty coefficient is a parameter used to adjust the influence of temperature on load distribution, which reflects the degree of constraint of device temperature on load distribution. The real-time temperature of the device is obtained by temperature sensor, and the safe temperature is set according to the safety operation standard of the device. The purpose of the optimization objective function is to minimize the deviation between the load and the optimal load, while considering the influence of temperature on device operation. The constraint condition ensures that the power of the device is between the minimum allowed power and the maximum allowed power, which are set according to the safety operation standard and performance requirements of the device. Through the optimization objective function and the constraint condition, the system can find the optimal load distribution scheme to improve the operating efficiency and safety of the device.

[0174] Preferably, the load scheduling optimization process can be further refined. For example, the optimization objective function can be constructed through the following steps: First, calculate the deviation between the actual load and the optimal load of each device at a certain moment; then calculate the difference between the real-time temperature and the safe temperature of the device, and multiply it by a temperature penalty coefficient to obtain the temperature penalty term. Add the load deviation and the temperature penalty term to obtain the value of the optimization objective function. The minimization of the optimization objective function can be achieved through mathematical optimization algorithms, such as gradient descent or genetic algorithms. During the optimization process, it is necessary to ensure that the power of the devices meets the constraints, i.e., the power is between the minimum allowable power and the maximum allowable power. For example, the minimum allowable power can be set to 50% of the rated power of the device, and the maximum allowable power can be set to 120% of the rated power of the device. In this way, the system can dynamically adjust the load distribution to ensure that the power equipment operates in an efficient and safe state, while avoiding damage to the equipment due to overload or excessive temperature.

[0175] In some embodiments, a multi-source data fusion mechanism is also included:

[0176] Using DS evidence theory to fuse multi-sensor data:

[0177]

[0178] Where m(A) is the basic probability assignment of device state A, m1(B) is the confidence of sensor 1 in device state B, m2(C) is the confidence of sensor 2 in device state C, and K is the conflict coefficient.

[0179] Generate fusion credibility distribution:

[0180]

[0181] Where Bel(A) represents the confidence level of device state A;

[0182] The method further includes:

[0183] Generate device control instruction set:

[0184] ControlSet = {(D i ,P j V k )}

[0185] Among them, D i For device identifier, P j For control parameter type, V k The target value for the parameter;

[0186] The data is transmitted to the device controller via the Industrial Internet of Things (IIoT) protocol.

[0187] Real-time monitoring control effect index η, wherein η is the control effect coefficient;

[0188] When η < η t , activate control strategy re-optimization, wherein η t is the effect threshold.

[0189] It should be noted that the multi-source data fusion mechanism mentioned in the present application is to fuse multi-sensor data by using D-S evidence theory, so as to generate a credibility distribution of the device state, and accordingly generate a device control instruction set. This method can effectively handle the conflicts and uncertainties between multi-source data, and improve the accuracy of device state evaluation. D-S evidence theory is a mathematical tool for handling uncertainty and conflicting information, which can more accurately evaluate the device state by calculating basic probability distribution and credibility distribution. The generated device control instruction set is used to guide the regulation and control operation of the device, and the effectiveness of the control strategy is evaluated by real-time monitoring of the control effect index. When the control effect does not meet the expectation, the system will activate the re-optimization process of the control strategy to ensure the stability and reliability of the device operation.

[0190] Specifically, the process of D-S evidence theory fusing multi-sensor data includes calculating the basic probability distribution and credibility distribution of the device state. The basic probability distribution is a measure of uncertainty of the device state, which reflects the confidence of different sensors on the device state. For example, the confidence of sensor 1 on device state A is m1(A), and the confidence of sensor 2 on device state B is m2(B). The conflict coefficient K is used to quantify the degree of conflict between different sensor data, which is obtained by calculating the difference between sensor data. The credibility distribution is a comprehensive evaluation of the device state, which calculates the credibility of the device state by combining the data of multiple sensors. The device control instruction set is a set composed of device identifier, control parameter type and parameter target value, which is used to guide the regulation and control operation of the device. The control effect index is a parameter used to evaluate the effectiveness of the control strategy, which reflects the actual operation effect of the device under the control strategy. The effect threshold is a preset reference value used to judge whether the control effect meets the expectation.

[0191] Preferably, the implementation process of the multi-source data fusion mechanism can be further refined. For example, when calculating the basic probability assignment, the initial belief of each sensor on the device state can be set according to the historical data and reliability of the sensor. For different types of sensors, the belief value can be adjusted according to the physical quantity and accuracy of the measurement. The calculation of the conflict coefficient K can be achieved by comparing the differences between different sensor data, for example, by calculating the deviation of the sensor data from the data mean, and combining the data standard deviation to quantify the conflict degree. The generation of the belief distribution is achieved by combining the basic probability assignment and the conflict coefficient to calculate the comprehensive belief of the device state. The generation of the device control instruction set can be determined according to the belief distribution of the device state, for example, if the belief of the device state is high, the control instruction to maintain the current state can be generated; if the belief is low, the instruction to optimize the control strategy can be generated. The control effect index can be calculated by monitoring the actual operating parameters of the device, for example, by comparing the deviation between the actual load of the device and the target load to evaluate the control effect. When the control effect index is lower than the effect threshold, the system will activate the re-optimization process of the control strategy to adjust the control parameters, ensuring the stability and reliability of the device operation.

[0192] The above-mentioned various embodiments of the present application have the following beneficial effects:

[0193] 1. By collecting multi-source monitoring data in real time and fusing pre-processing, the problem of data isolation and insufficient precision in traditional power equipment monitoring is solved, effectively improving the reliability and integrity of the data, providing high-quality input for subsequent state evaluation, and avoiding the risk of misjudgment caused by the failure of a single sensor or noise interference.

[0194] 2. Based on dynamic analysis period and device state index calculation, the precise classification of power equipment operation risk is realized, solving the problem of poor flexibility of traditional threshold alarm mechanism, which can adapt to different working conditions, identify potential faults in time and trigger optimization or maintenance strategy, reducing the probability of sudden equipment failure.

[0195] 3. Combined with the control state prediction model and hierarchical decision tree to generate control instructions, the problem of manual control response lag is solved, through automatic operation such as load adjustment and standby device switching, to ensure the dynamic matching of control strategy and actual state of the device, while supporting real-time feedback and re-optimization of strategy execution effect, improving the overall control efficiency and safety.

[0196] Further, the storage medium of the embodiments of the present application stores program instructions capable of realizing all the methods described above, wherein the program instructions can be stored in the storage medium in the form of a software product, and include a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor execute all or part of the steps of the methods described in the embodiments of the present application. And the aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media capable of storing program codes, or a computer, a server, a mobile phone, a tablet, and other terminal devices.

[0197] The above description is merely some preferred embodiments of the present application and a description of the principles of the technology used. Those skilled in the art should understand that the scope of the application involved in the embodiments of the present application is not limited to the technical solutions formed by the specific combinations of the technical features described above, and should also cover other technical solutions formed by any combinations of the above technical features or equivalent features without departing from the above inventive concept. For example, the technical solutions formed by mutually replacing the above features and the technical features disclosed in the embodiments of the present application (but not limited to) having similar functions.

Claims

1. A method for constructing a digital twin platform through multi-source data fusion, applied to a power equipment monitoring system, comprising: Collect real-time multi-source monitoring data streams for power equipment; The real-time multi-source monitoring data stream is preprocessed to obtain a preprocessed multi-source monitoring data stream; The analysis period is determined based on the preprocessed multi-source monitoring data stream; Based on the target multi-source monitoring data stream, determine the equipment status label, wherein the target multi-source monitoring data stream is the preprocessed monitoring data located within the analysis period in the preprocessed multi-source monitoring data stream, and the equipment status label characterizes the operational risk level of the power equipment; In response to determining that the device status label is a high-risk status label, an optimized control strategy information for the power device is generated based on the current control strategy information corresponding to the power device and the device status label, and this is used as the updated control strategy information. In response to determining that the device status label is a low-risk status label, maintenance control strategy information for the power equipment is generated based on the current control strategy information and the device status label, as the updated control strategy information; Based on the current control strategy information or the updated control strategy information, the power equipment is regulated.

2. The method according to claim 1, characterized in that, Based on the current control strategy information or the updated control strategy information, equipment regulation of power equipment includes: Based on the current control strategy information or the updated control strategy information, as well as the device status label and the control status prediction model, determine the status deviation information; In response to determining that the state deviation information indicates a control deviation in the current control strategy information or the updated control strategy information, a control instruction corresponding to the state deviation information is determined in the state control decision tree; According to the control command, the power equipment is subjected to load adjustment operation or the standby equipment switching operation is activated.

3. The method according to claim 2, characterized in that, The step of preprocessing the real-time multi-source monitoring data stream to obtain a preprocessed multi-source monitoring data stream includes: For each real-time monitoring data in the real-time multi-source monitoring data stream, perform the following processing steps: The sensor validity is verified by analyzing the real-time monitoring data. In response to determining that the real-time monitoring data has passed the validity check, outlier correction processing is performed on the real-time monitoring data to obtain corrected monitoring data; The corrected monitoring data is then subjected to data standardization processing to obtain standardized monitoring data; Determine the data acquisition frequency corresponding to the real-time multi-source monitoring data stream; In response to determining that the data acquisition frequency is not a reference acquisition frequency and that the data acquisition frequency is less than the reference acquisition frequency, the standardized monitoring dataset is downsampled at the reference acquisition frequency to obtain the preprocessed multi-source monitoring data stream.

4. The method according to claim 3, characterized in that, The step of determining the analysis period based on the preprocessed multi-source monitoring data stream includes: Initialize the time period length corresponding to the analysis time period to obtain the initial analysis time period; Based on the initial time period length and the preprocessed multi-source monitoring data stream, the following time period determination steps are performed: Determine the target multi-source monitoring data stream; Calculate the average values ​​of equipment parameters based on the target multi-source monitoring data stream; Based on the target multi-source monitoring data stream, the average value of equipment parameters, the monitoring data of the first target and the monitoring data of the second target, the time period characteristic value of the initial analysis period is calculated, where the monitoring data of the first target is the maximum value of the equipment parameters and the monitoring data of the second target is the minimum value of the equipment parameters; In response to the average value of the device parameters being greater than or equal to the dynamic threshold, the initial analysis period is determined as the analysis period; In response to the average value of the device parameters being less than the dynamic threshold, the length of the initial analysis period is incremented to obtain the analysis period with the increased length, and the period determination step is executed again.

5. The method according to claim 4, characterized in that, The step of determining the device status label based on the target multi-source monitoring data stream includes: Calculate the equipment condition index: Where H is the equipment status index, w i V represents the weight coefficients of the i-th type of parameter. i Let V be the value of the i-th type of monitoring parameter. base V is the baseline parameter value. max V is the maximum allowed value for the parameter. min Minimum allowed value for the parameter; In response to the device condition index continuously exceeding the high-risk threshold H risk Upon reaching the first time period, the device status label is determined to be a high-risk status label; In response to the device condition index continuously falling below the low-risk threshold H safe Upon reaching the second time period, the device status label was determined to be a low-risk status label.

6. The method according to claim 5, characterized in that, The regulation state prediction model is constructed through the following steps: Extracting characteristics of changes in equipment parameters: Where, ΔP t Let P be the rate of change of the parameter at time t. t Let P be the parameter value at time t. t-1 The parameter value at time t-1; Extracting conflict levels from multi-source data: Where C represents the data conflict degree, and V i Let μ be the data from the i-th sensor, and σ be the data mean and standard deviation. The aforementioned change characteristics and conflict degree are input into the Long Short-Term Memory network model, which outputs the state deviation probability P. a , where P a This represents the probability of state deviation.

7. The method according to claim 2, characterized in that, The state control decision tree includes: First-level node: Determine if the equipment temperature T exceeds the safe temperature T. s Where T is the real-time temperature, T s The safe temperature threshold; Second-level node: Responding to T>T s Detect whether the current fluctuation rate ΔI exceeds the threshold ΔI m Where ΔI is the current fluctuation rate, ΔI m Maximum permissible volatility; Third-level node: Response to ΔI>ΔI m Output a first-level control command to activate load splitting; Fourth-level node: Response to ΔI≤ΔI m It outputs a secondary control command to start the cooling system.

8. The method of claim 7, further comprising the construction of a digital twin model: Establish a three-dimensional physical model of the power equipment: M 3D =f(D p ,M p ,T p ) in, M 3D For a three-dimensional model, D p For equipment size parameters, M p For material properties, T p For topological structure; Update the model state through real-time data mapping: S t =S t-1 +k·(V t -V p ) Among them, S t Let S be the model state at time t. t-1 Let V be the model state at time t-1, k be the state update coefficient, and V be the model state at time t-1. t V is the monitoring value at time t. p These are predicted values.

9. The method according to claim 8, characterized in that, Also includes: Load scheduling optimization based on device state prediction: Among them, L t Let L be the load at time t. o For the optimal load, λ is the temperature penalty coefficient, T t Let T be the temperature at time t. s For safe temperature; The constraints are satisfied: P min ≤P t ≤P max (t=1,2,...,T) Among them, P t Let P be the power at time t. min For the minimum allowable power, P max This represents the maximum permissible power.

10. The method according to claim 1, characterized in that, It also includes multi-source data fusion mechanisms: Using DS evidence theory to fuse multi-sensor data: Where m(A) is the basic probability assignment of device state A, m1(B) is the confidence of sensor 1 in device state B, m2(C) is the confidence of sensor 2 in device state C, and K is the conflict coefficient. Generate fusion credibility distribution: Where Bel(A) represents the confidence level of device state A; The method further includes: Generate device control instruction set: ControlSet={(D i ,P j ,V k )} Among them, D i For device identifier, P j For control parameter type, V k The target value for the parameter; The data is transmitted to the device controller via the Industrial Internet of Things (IIoT) protocol. Real-time monitoring of the control effectiveness index η, where η is the control effectiveness coefficient; When η < η t At that time, the activation control strategy is re-optimized, where η t This represents the effect threshold.

Images