A method for monitoring the business operation status based on enterprise power data

Abstract

This application belongs to the field of data analysis technology and discloses a method for monitoring the business status of enterprises based on enterprise power data. By performing feature decomposition and modal component analysis on the time series data of the total power consumption, and combining with the reference power feature set of the main production equipment, it realizes the accurate identification and analysis of the working status of each main production equipment of the enterprise. This method overcomes the limitation of traditional methods that only rely on the total power consumption, can more accurately reflect the actual production situation of the enterprise, is not easily manipulated by humans, and thus provides a more reliable technical means for monitoring the business status of the enterprise. It realizes the real-time monitoring and early warning of the business status of the enterprise, can improve the informatization level, reduce the workload of loan management personnel, ensure the authenticity and reliability of data, and improve the real-time performance and accuracy of data analysis.

Description

Technical Field

[0001] This application relates to the technical field of data analysis. Specifically, it relates to a method for monitoring the business status of enterprises based on enterprise power data. Background Art

[0002] Traditional post-loan management methods of financial institutions usually conduct post-loan management by collecting information such as the financial data, bank statements, electricity bills, tax bills, and legal lawsuits of enterprises, as well as on-site inspections by post-loan management personnel to enterprises. However, this method has many problems: the digital degree of information collection is low, and data collation takes a long time; data information is easily falsified; the collected data often can only reflect the business conditions of enterprises over a long period in the past, cannot accurately reflect the current business status of enterprises in a timely manner, and even less can predict the business risks of enterprises in advance, resulting in financial institutions possibly missing the best opportunity to dispose of risks.

[0003] Currently, the means of post-loan management of financial institutions are limited, inefficient, ineffective, and the degree of informatization and digitization is relatively low. Therefore, there is an urgent need to find new methods to improve the informatization level, reduce the workload of loan management personnel, ensure the authenticity and reliability of data, and improve the real-time and accuracy of data analysis.

[0004] The business status of an enterprise is closely related to the order saturation, and the order saturation directly affects the production and manufacturing activities of the enterprise. The more orders there are, the busier the production is, and the more frequently the production equipment is started. Since the operation of production equipment consumes electric energy, by monitoring the production power consumption data of the enterprise, the order situation of the enterprise can be grasped, and then it can be understood whether the business status of the enterprise is good.

[0005] Currently, the commonly used methods for judging the business status of enterprises based on power consumption data usually use the change and trend of power consumption of enterprises over a past period of time to judge the business conditions of enterprises. However, this method has the following problems:

[0006] The amount of information of pure power consumption is small, and the data dimension is too single. An enterprise may increase its power consumption by turning on power-consuming equipment that has nothing to do with production and operation, disguising itself as having sufficient orders and a full production load, so as to deceive financial institutions.

[0007] There are many power-consuming equipment in an enterprise, but not all power-consuming equipment has a high correlation with the business status of the enterprise. The existing methods for judging the business status of enterprises analyze using the power consumption information of all power-consuming equipment, resulting in an unsatisfactory accuracy of the judgment results.

[0008] In addition, the existing methods cannot monitor the business status of enterprises in real time, and it is difficult to discover potential business risks in a timely manner. At the same time, they lack the ability to predict the future business status of enterprises and cannot provide forward-looking risk management suggestions for financial institutions.

[0009] In view of the above problems, the existing technology urgently needs to be improved. Summary of the Invention

[0010] The purpose of this application is to provide a method for monitoring the business operation status of enterprises based on enterprise power data, which can improve the informatization level, reduce the workload of loan management personnel, ensure the authenticity and reliability of data, and improve the real-time and accuracy of data analysis.

[0011] This application provides a method for monitoring the business operation status of enterprises based on enterprise power data, including the steps of:

[0012] A1. Obtain the total power consumption time series data of the enterprise, and perform feature decomposition on the total power consumption time series data according to a preset time cycle length to obtain multiple modal components corresponding to each time cycle;

[0013] A2. Obtain the reference power feature set of each main production equipment of the enterprise; the reference power feature set includes reference time domain features and reference frequency domain features;

[0014] A3. Determine the effectiveness of each modal component according to the feature entropy of the modal component, and perform noise filtering processing on the effective modal components to obtain filtered modal components;

[0015] A4. Extract the time domain features and frequency domain features of each filtered modal component to form an actual power feature set corresponding to each filtered modal component, and use it to match with the reference power feature set to identify the corresponding relationship between each filtered modal component and each main production equipment;

[0016] A5. Obtain the working state data of each main production equipment according to the matching recognition results corresponding to each time cycle;

[0017] A6. Judge the current business operation status of the enterprise according to the working state data.

[0018] By performing feature decomposition and modal component analysis on the total power consumption time series data, and combining with the reference power feature set of the main production equipment, this method realizes the accurate identification and analysis of the working states of each main production equipment of the enterprise. This method overcomes the limitation of traditional methods that only rely on the total power consumption, can more accurately reflect the actual production situation of the enterprise, is not easily manipulated by humans, and thus provides a more reliable technical means for monitoring the business operation status of the enterprise. It realizes the real-time monitoring and early warning of the business operation status of the enterprise, can improve the informatization level, reduce the workload of loan management personnel, ensure the authenticity and reliability of data, and improve the real-time and accuracy of data analysis.

[0019] Preferably, step A1 includes:

[0020] A101. Obtain the total power consumption time series data of the enterprise for a preset duration;

[0021] A102. Divide the total power consumption time series data according to the time period length to obtain multiple total power consumption time series data segments;

[0022] A103. Based on the modal decomposition algorithm, perform feature decomposition on each total power consumption time series data segment to obtain multiple modal components corresponding to each time period.

[0023] Preferably, in step A103, the variational mode decomposition algorithm is used to perform feature decomposition on each total power consumption time series data segment.

[0024] The variational mode decomposition algorithm is an advanced signal processing technology that can effectively decompose complex time series data into multiple modal components. In solving the problem of feature decomposition of the enterprise's total power consumption time series data, the variational mode decomposition algorithm has the following advantages: it can adaptively determine the number of decomposed modes without manual setting; it has strong robustness to noise and can better extract useful signal features; compared with other modal decomposition methods, such as empirical mode decomposition (EMD), the variational mode decomposition algorithm can better handle the mode mixing problem.

[0025] Preferably, the reference time domain features include at least one of reference mean, reference standard deviation, reference kurtosis, reference mean square value, reference root mean square value, reference peak factor, reference impulse factor, and reference waveform factor;

[0026] The reference frequency domain features include at least one of reference center frequency, reference bandwidth, reference power spectral density, reference frequency variance, and reference center of gravity frequency;

[0027] The time domain features include at least one of mean, standard deviation, kurtosis, mean square value, root mean square value, peak factor, impulse factor, and waveform factor;

[0028] The frequency domain features include at least one of center frequency, bandwidth, power spectral density, frequency variance, and center of gravity frequency.

[0029] The combined use of these features realizes a comprehensive description of the characteristics of power data. Through standardized feature definitions, a reliable feature basis is provided for enterprise operation status monitoring. The time domain features reflect the statistical distribution characteristics of power data, and the frequency domain features reflect the frequency composition characteristics of power data. The combined use of the two types of features can comprehensively describe and compare the characteristics of power data. Thus, it is possible to more accurately match the filtered modal components with the main production equipment.

[0030] Preferably, step A3 includes:

[0031] A301. Calculate the characteristic entropy of each of the modal components;

[0032] A302. Compare the characteristic entropy of each of the modal components with a preset characteristic entropy threshold, and determine the modal components with characteristic entropy lower than the characteristic entropy threshold as effective modal components;

[0033] A303. Perform noise filtering on the effective modal components to obtain filtered modal components.

[0034] Preferably, in step A303, the Wiener adaptive filter algorithm is used to perform noise filtering on the effective modal components.

[0035] Preferably, step A4 includes:

[0036] A401. Extract the time-domain features of each of the filtered modal components;

[0037] A402. Extract the frequency-domain features of each of the filtered modal components based on FFT transformation;

[0038] A403. Use the extracted time-domain features and frequency-domain features to form the actual power feature set of each of the filtered modal components;

[0039] A404. Compare the actual power feature set of each of the filtered modal components with the reference power feature set of each of the main production devices, and determine the corresponding relationship between each of the filtered modal components and each of the main production devices.

[0040] Preferably, the operating state data includes the start time, working duration, stop time, and steady-state power value change index;

[0041] Step A5 includes:

[0042] Identify the start time, working duration, and stop time of each of the main production devices according to the filtered modal components corresponding to each of the main production devices in each of the time periods;

[0043] Calculate the steady-state power value change index of each of the main production devices according to the characteristic entropy of the filtered modal components corresponding to each of the main production devices in each of the time periods.

[0044] Preferably, step A6 includes:

[0045] Calculate the business state outlier value according to the start time, working duration, stop time, and steady-state power value change index;

[0046] Judge whether the current business state of the enterprise is abnormal according to the business state outlier value.

[0047] Preferably, after step A6, the method further includes the steps of:

[0048] A7. If the current operating state of the enterprise is normal, predicting the estimated power consumption time series data of the enterprise according to the total power consumption time series data; the estimated power consumption time series data is the estimated data of the total power consumption time series data of the enterprise in the future;

[0049] A8. Evaluating the potential risk of abnormal operating state of the enterprise according to the estimated power consumption time series data.

[0050] Beneficial effects: The enterprise operating state monitoring method based on enterprise power data provided by this application realizes the accurate identification and analysis of the working states of the main production equipment of the enterprise by performing feature decomposition and modal component analysis on the total power consumption time series data and combining the reference power feature set of the main production equipment. This method overcomes the limitation of traditional methods that only rely on the total power consumption, can more accurately reflect the actual production situation of the enterprise, is not easily manipulated artificially, and thus provides a more reliable technical means for monitoring the enterprise operating state. It realizes the real-time monitoring and early warning of the enterprise operating state, can improve the informatization level, reduce the workload of loan management personnel, ensure the authenticity and reliability of data, and improve the real-time performance and accuracy of data analysis. Description of the Drawings

[0051] Figure 1 It is a flowchart of an enterprise operating state monitoring method based on enterprise power data provided by an embodiment of this application.

[0052] Figure 2 It is a flowchart of another enterprise operating state monitoring method based on enterprise power data provided by an embodiment of this application. Detailed Embodiments

[0053] Next, the technical solutions in the embodiments of this application will be clearly and completely described in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. The components of the embodiments of this application described and illustrated herein can be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of this application provided in the drawings below is not intended to limit the scope of the claimed application, but merely represents the selected embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative efforts fall within the scope of protection of this application.

[0054] It should be noted that similar reference numerals and letters indicate similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of the present application, terms such as "first" and "second" are only used for distinguishing descriptions and cannot be construed as indicating or implying relative importance.

[0055] In the field of enterprise operation status monitoring, traditional methods mainly rely on collecting information such as the financial data, bank statements, and electricity bills of enterprises, as well as on-site inspections by post-loan management personnel. However, this method has problems such as low information collection efficiency, possible data falsification, and information lag, and cannot timely and accurately reflect the current operation status of enterprises, let alone early warning of potential risks.

[0056] In recent years, a method for judging the operation status based on enterprise electricity consumption data has been applied. This method evaluates the operation status of an enterprise by analyzing the changes and trends in its electricity consumption over a period of time. However, this method still faces technical challenges such as a single data dimension, being easily disguised by enterprises through non-production electrical equipment, and difficulty in distinguishing the electricity consumption of key production equipment. Specifically, the amount of data of simple electricity consumption information is small and cannot comprehensively reflect the actual production status of enterprises. At the same time, enterprises may increase the total electricity consumption by turning on electrical equipment unrelated to production and operation, thus disguising themselves as having sufficient orders and full production load. In addition, since enterprises usually have a variety of electrical equipment, but not all equipment is highly related to the operation status, the analysis method based on total electricity consumption is difficult to accurately reflect the true operation status of enterprises.

[0057] If this technical problem cannot be effectively solved, the following serious consequences will occur: First, financial institutions may not be able to timely detect the deterioration of the operation status of enterprises, miss the best opportunity for risk intervention, and increase the risk of loan defaults. Second, due to the lack of data reliability, financial institutions may need to invest more manpower in on-site inspections, increasing management costs. Third, enterprises may take advantage of the limitations of this monitoring method to conceal their true operation status by manipulating electricity consumption data, which not only increases financial risks but also may lead to misallocation of resources.

[0058] In addition, this problem may also affect the credit decision-making efficiency of financial institutions. Due to the inability to obtain accurate and real-time enterprise operation status information, financial institutions may be too conservative or aggressive when approving new loan applications or adjusting existing loan conditions, affecting the effective allocation of financial resources. From a technical perspective, this problem highlights the deficiencies of existing data collection and analysis methods in terms of accuracy, real-time performance, and anti-interference ability, and there is an urgent need for a more intelligent and comprehensive technical solution.

[0059] Based on the analysis of the existing technical problems, the present application proposes a new solution.

[0060] Please refer to Figure 1 , Figure 1 which is a method for monitoring the business operation status based on enterprise power data in some embodiments of the present application, including the steps:

[0061] A1. Obtain the total power consumption time series data of the enterprise, and perform feature decomposition on the total power consumption time series data according to a preset time period length to obtain multiple modal components corresponding to each time period;

[0062] A2. Obtain the reference power feature set of each main production equipment of the enterprise; the reference power feature set includes reference time domain features and reference frequency domain features;

[0063] A3. Determine the effectiveness of each modal component according to the feature entropy of the modal component, and perform noise filtering processing on the effective modal components to obtain filtered modal components;

[0064] A4. Extract the time domain features and frequency domain features of each filtered modal component to form an actual power feature set corresponding to each filtered modal component, and use it to match with the reference power feature set to identify the corresponding relationship between each filtered modal component and each main production equipment;

[0065] A5. Obtain the working status data of each main production equipment according to the matching recognition results corresponding to each time period;

[0066] A6. Judge the current business operation status of the enterprise according to the working status data.

[0067] This method realizes the accurate identification and analysis of the working status of each main production equipment of the enterprise by performing feature decomposition and modal component analysis on the total power consumption time series data, combined with the reference power feature set of the main production equipment. This method overcomes the limitation of the traditional method that only relies on the total power consumption, can more accurately reflect the actual production situation of the enterprise, is not easily manipulated by humans, and thus provides a more reliable technical means for monitoring the business operation status of the enterprise. It realizes the real-time monitoring and early warning of the business operation status of the enterprise, can improve the informatization level, reduce the workload of loan management personnel, ensure the authenticity and reliability of data, and improve the real-time and accuracy of data analysis.

[0068] Specifically, step A1 includes:

[0069] A101. Obtain the total power consumption time series data of the enterprise for a preset duration;

[0070] A102. Divide the total power consumption time series data according to the time period length to obtain multiple total power consumption time series data segments;

[0071] A103. Perform feature decomposition on each time series data segment of the total power consumption to obtain multiple modal components corresponding to each time period.

[0072] By preprocessing and feature decomposing the time series data of the enterprise's total power consumption, the problem of how to extract useful information from complex power consumption data is solved. First, obtaining the time series data of the total power consumption with a preset duration ensures the analysis time range. Then, dividing the data into segments for multiple time periods enables the analysis to be carried out on a finer-grained time scale. Finally, using a modal decomposition algorithm to perform feature decomposition on each data segment, multiple modal components are obtained, and these modal components may correspond to different production equipment or production activities of the enterprise.

[0073] Among them, the time series data of the total power consumption refers to the data sequence of the enterprise's total power consumption changing over time, and specifically, it can be collected in real time using an intelligent electricity meter or a power monitoring device. Here, the time series data of the total power consumption with a preset duration is collected as the analysis object, and the preset duration can be set according to actual needs, such as one day, one week, or one month. It should be noted that the sampling duration (i.e., the preset duration) of the time series data of the total power consumption does not represent the analysis period (i.e., the time interval for executing the enterprise operation status monitoring method based on the enterprise power data of this application), and the analysis period can be set according to actual needs. For example, when the preset duration is one month, the analysis period can be one day or one week, but it is not limited to this.

[0074] Among them, in step A102, the time series data of the total power consumption is divided into multiple time series data segments of the total power consumption, and the duration of each time series data segment of the total power consumption is the time period length. The selection of the time period length needs to consider the production characteristics of the enterprise and the purpose of data analysis. For example, for an enterprise with continuous production, a shorter time period, such as 1 hour or 2 hours, can be selected; for an enterprise with batch production, a time period matching the production batch can be selected. By reasonably dividing the time period, the periodic changes in the enterprise's power consumption pattern can be better captured.

[0075] Among them, feature decomposition refers to the process of decomposing a complex signal into multiple simple components, and a modal component refers to the basic component of a signal obtained through feature decomposition. In step A103, feature decomposition is performed on each time series data segment of the total power consumption, so that each time series data segment of the total power consumption can be decomposed into multiple modal components.

[0076] The modal decomposition algorithm can decompose a complex signal into multiple simple modal components. Commonly used modal decomposition algorithms include empirical mode decomposition (EMD), ensemble empirical mode decomposition (EEMD), and variational mode decomposition (VMD), etc. These algorithms have their own characteristics, and a suitable algorithm can be selected according to actual needs.

[0077] Preferably, in step A103, the variational mode decomposition algorithm is used to perform feature decomposition on each total power consumption time series data segment.

[0078] The variational mode decomposition algorithm is an advanced signal processing technology that can effectively decompose complex time series data into multiple modal components. In solving the problem of feature decomposition of the total power consumption time series data of enterprises, the variational mode decomposition algorithm has the following advantages: it can adaptively determine the number of decomposed modes without manual setting; it has strong robustness to noise and can better extract useful signal features; compared with other modal decomposition methods, such as empirical mode decomposition (EMD) and ensemble empirical mode decomposition (EEMD), the variational mode decomposition algorithm can better handle the mode mixing problem.

[0079] Among them, the variational mode decomposition algorithm generally realizes the decomposition and extraction of modal components through steps such as constructing a constrained variational problem model, introducing Lagrange operators and quadratic penalty factors, iterative optimization, and modal extraction. In this embodiment, the constrained variational problem model can be constructed as:

[0080] (1);

[0081] Among them, is the total power consumption time series data segment, u is the serial number, is the u-th modal component, j is the imaginary symbol, t is the time, is 's set, is the central frequency of the u-th modal component, is 's set, is the impulse function, represents the differential function for finding the differential with respect to time, and s.t. is the constraint condition symbol. By introducing Lagrange operators and quadratic penalty factors and using an iterative algorithm to solve the above constrained variational problem model (the specific process is the prior art and will not be elaborated here), each modal component can be obtained. Each modal component obtained by solving the above constrained variational problem model has a finite bandwidth, and its frequency band is tightly centered around its specific central frequency , thus being able to effectively map the unique power characteristics of different electrical equipment and laying a solid foundation for subsequent analysis.

[0082] It should be noted that the constrained variational problem model is not limited to the above model, and the constrained variational problem model in the prior art can also be selected.

[0083] Among them, the reference power feature set refers to the typical power consumption characteristics of each main production device obtained in advance, including time-domain characteristics and frequency-domain characteristics, which can be specifically reported by the enterprise independently or analyzed by collecting the power consumption data when the device runs alone. The main production device is a production device that has a high degree of correlation with the enterprise's operating status. The main production devices can be determined in advance from all the power-consuming devices of the enterprise by manual or through big data statistical methods, and a query table is formed by mapping each main production device and the corresponding reference power feature set. In step A2, the reference power feature sets of each main production device of the enterprise are extracted from the query table.

[0084] Among them, the time-domain characteristic is a parameter that describes the signal characteristics in the time dimension, which can be specifically extracted by statistical analysis methods.

[0085] Among them, the frequency-domain characteristic is a parameter that describes the signal characteristics in the frequency dimension, which can be specifically extracted by methods such as Fourier transform.

[0086] In some embodiments, the reference time-domain characteristics include at least one of reference mean, reference standard deviation, reference kurtosis, reference mean square value, reference root mean square value, reference peak factor, reference impulse factor, and reference waveform factor;

[0087] The reference frequency-domain characteristics include at least one of reference center frequency, reference bandwidth, reference power spectral density, reference frequency variance, and reference centroid frequency;

[0088] The time-domain characteristics include at least one of mean, standard deviation, kurtosis, mean square value, root mean square value, peak factor, impulse factor, and waveform factor;

[0089] The frequency-domain characteristics include at least one of center frequency, bandwidth, power spectral density, frequency variance, and centroid frequency.

[0090] Among them, the mean and the reference mean reflect the average level of the power data; the standard deviation and the reference standard deviation characterize the dispersion degree of the power data; the kurtosis and the reference kurtosis describe the steepness of the power data distribution; the mean square value, the root mean square value, the reference mean square value, and the reference root mean square value all reflect the energy size of the power data; the peak factor, the impulse factor, the waveform factor, the reference peak factor, the reference impulse factor, and the reference waveform factor are used to describe the waveform characteristics of the power data.

[0091] Among them, the center frequency and the reference center frequency represent the centroid position of the power spectrum; the bandwidth and the reference bandwidth reflect the frequency range of the power spectrum; the power spectral density and the reference power spectral density describe the power distribution per unit frequency; the frequency variance and the reference frequency variance characterize the dispersion degree of the frequency components; the centroid frequency and the reference centroid frequency reflect the centroid position of the power spectrum.

[0092] The combined use of these features realizes a comprehensive description of the characteristics of power data. Through standardized feature definitions, a reliable feature basis is provided for enterprise operation status monitoring. Time-domain features reflect the statistical distribution characteristics of power data, and frequency-domain features reflect the frequency composition characteristics of power data. The combined use of these two types of features can comprehensively describe and compare the characteristics of power data. Thus, it is possible to more accurately match the filtered modal components with the main production equipment.

[0093] Specifically, step A3 includes:

[0094] A301. Calculate the feature entropy of each modal component;

[0095] A302. Compare the feature entropy of each modal component with a preset feature entropy threshold, and determine the modal components with feature entropy lower than the feature entropy threshold as valid modal components;

[0096] A303. Perform noise filtering on the valid modal components to obtain the filtered modal components.

[0097] This method effectively solves the problems of judging the validity of modal components and noise filtering through steps such as calculating feature entropy, setting thresholds, judging validity, and performing noise filtering. This solution can extract more useful information from complex power data, remove redundancy and noise, thereby improving the accuracy and reliability of subsequent enterprise operation status monitoring.

[0098] Among them, feature entropy is an index used to measure the complexity of a signal. Specifically, methods such as sample entropy or approximate entropy can be used for calculation. The feature entropy of each modal component can also be calculated using the following formula:

[0099] ;

[0100] ;

[0101] ;

[0102] Among them, is the feature entropy, is the first reference quantity, is the second reference quantity, N is the data dimension of the modal component (i.e., the number of data contained in the modal component), M is the preset window length (which can be set according to actual needs), is the number of other first data sequences in the first data sequence obtained by sliding window truncation of the modal component with a window length of M and a single data as the step size, whose distance from the i-th first data sequence is less than or equal to s, and s is the preset distance threshold (which can be set according to actual needs), In the second data sequence obtained by sliding window truncation of modal components with a single data as the step size using a window of length M + 1, it is the number of other second data sequences whose distance from the i-th second data sequence is less than or equal to s.

[0103] Among them, the i-th first data sequence can be expressed as:

[0104] ;

[0105] is the i-th first data sequence, , , are respectively the i-th, (i + 1)-th, and (i + M - 1)-th data in the modal component;

[0106] The distance between any i-th first data sequence and the jj-th first data sequence (i ≠ jj) is:

[0107] ;

[0108] is the distance between the i-th first data sequence and the jj-th first data sequence, , are respectively the (i + k)-th and (jj + k)-th data in the modal component.

[0109] Similarly, the i-th second data sequence can be expressed as:

[0110] ;

[0111] is the i-th second data sequence, is the (i + M)-th data in the modal component.

[0112] The distance between any i-th second data sequence and the jj-th second data sequence (i ≠ jj) is:

[0113] ;

[0114] is the distance between the i-th second data sequence and the jj-th second data sequence.

[0115] The magnitude of the characteristic entropy calculated in the above manner reflects the complexity of the electricity load. The larger the characteristic entropy, the more complex and unpredictable the change of the electricity load.

[0116] When the characteristic entropy is too high, it is considered that the corresponding modal component contains too much noise and is not suitable for analyzing the enterprise operation status. Here, a characteristic entropy threshold is set to judge whether the characteristic entropy of each modal component is too high, so that only the modal components with characteristic entropy lower than the characteristic entropy threshold are used as effective modal components for subsequent enterprise operation status analysis to ensure the accuracy of the analysis results. Among them, the setting of the characteristic entropy threshold can be adjusted according to the actual application scenario. For example, a suitable threshold range can be determined through statistical analysis of a large number of sample data.

[0117] Among them, the interference components in the effective modal components can be removed through noise filtering processing. For the noise filtering processing of the effective modal components, various filtering algorithms can be used. For example, methods such as Wiener filtering, Kalman filtering or wavelet denoising can be used. Further, the most suitable filtering method can be selected according to the characteristics of the signal.

[0118] In some preferred embodiments, in step A303, the Wiener adaptive filter algorithm is used to perform noise filtering processing on the effective modal components.

[0119] The Wiener adaptive filter algorithm is an advanced signal processing technology used to extract useful information from noisy signals. For the electricity consumption power data, it can effectively remove the noise in the modal components and improve the data quality. The effective modal components are important signal components obtained through characteristic entropy calculation and threshold comparison. Performing noise filtering on these components can retain key information while reducing unnecessary interference. The noise filtering processing aims to improve the signal-to-noise ratio of the modal components, making subsequent feature extraction and matching more accurate. By removing the noise, the actual electricity consumption situation of the enterprise production equipment can be more clearly reflected.

[0120] By using the Wiener adaptive filter algorithm to perform noise filtering processing on the effective modal components, the problem of noise interference in the modal components is effectively solved. This method can adaptively adjust the filtering parameters to adapt to the characteristics of different modal components, thereby achieving more accurate noise removal.

[0121] Further, the normalized least mean square error (NLMS) algorithm or the least mean square error (LMS) algorithm can be used to implement the Wiener adaptive filter; preferably, the NLMS algorithm is used to implement the Wiener adaptive filter. Compared with the LMS algorithm, it has a faster convergence speed and better stability. The specific methods of using the normalized least mean square error (NLMS) algorithm or the least mean square error (LMS) algorithm to implement the Wiener adaptive filter are all prior arts and will not be elaborated here.

[0122] Further, step A4 includes:

[0123] A401. Extract the time-domain features of each filtered modal component;

[0124] A402. Extract the frequency-domain features of each filtered modal component based on the FFT transform;

[0125] A403. Use the extracted time-domain features and frequency-domain features to form the actual power feature set of each filtered modal component;

[0126] A404. Compare the actual power feature set of each filtered modal component with the reference power feature set of each main production device to determine the corresponding relationship between each filtered modal component and each main production device.

[0127] By extracting the time-domain and frequency-domain features of the filtered modal components and comparing them with the pre-set reference features (i.e., the reference time-domain features and reference frequency-domain features in the reference power feature set), the identification of the corresponding relationship between the filtered modal components and the main production devices is realized. This method not only considers the time-domain information but also introduces frequency-domain analysis, improving the comprehensiveness and accuracy of feature description. By using the FFT transform to extract the frequency-domain features, the periodic components in the signal can be captured, which is very helpful for identifying the features of different production devices. Finally, by comparing the actual power feature set and the reference power feature set, each filtered modal component can be accurately associated with the corresponding main production device. The innovation of this method lies in the comprehensive utilization of time-domain and frequency-domain information, improving the accuracy and reliability of identification, and providing an important basis for subsequent enterprise operation status monitoring.

[0128] Among them, the time-domain features of the filtered modal components such as mean, standard deviation, kurtosis, mean square value, root mean square value, peak factor, impulse factor, waveform factor, etc. can be extracted by statistical analysis methods, which is the prior art and will not be elaborated here.

[0129] Among them, the frequency-domain features of the filtered modal components such as center frequency, bandwidth, power spectral density, frequency variance, centroid frequency, etc. can be extracted by Fourier transform. For example, first use the FFT transform (Fast Fourier Transform) algorithm to convert the filtered modal component into the corresponding frequency spectrum, and then the center frequency and bandwidth can be extracted from the frequency spectrum and the power spectral density can be calculated. Among them, the extraction methods of the center frequency and bandwidth and the calculation method of the power spectral density are the prior art (in fact, when the modal component is extracted using formula (1) in step A1, the center frequency of each modal component can be calculated, so the calculated center frequency can also be directly called here), which will not be elaborated here, and the frequency variance and centroid frequency can be calculated by the following formulas:

[0130] ;

[0131] ;

[0132] Among them, is the frequency variance, is the center frequency, is the sampling frequency of the total power consumption time series data (for example, when collecting the total power consumption time series data, one data is collected per second, then the sampling frequency is 1 Hz), is the angular frequency, is the power spectrum of the filtered modal component (which can be calculated according to the frequency spectrum, and the calculation method is the prior art, and it will not be elaborated here).

[0133] In this embodiment, the working state data includes the start time, working duration, stop time, and steady-state power value change index;

[0134] Therefore, step A5 includes:

[0135] A501. Identify the start time, working duration, and stop time of each main production equipment according to the filtered modal component corresponding to each main production equipment in each time period;

[0136] A502. Calculate the steady-state power value change index of each main production equipment according to the characteristic entropy of the filtered modal component corresponding to each main production equipment in each time period.

[0137] By analyzing the working state data of the main production equipment of the enterprise to monitor the business state of the enterprise. First, use the filtered modal component to identify the start time, working duration, and stop time of each main production equipment, which reflects the usage frequency and intensity of the equipment. Second, by calculating the steady-state power value change index, evaluate the stability of the equipment operation. This method can more comprehensively and accurately reflect the actual production situation of the enterprise, avoid misjudgment that may be caused by simply relying on the total power consumption, and is also difficult to be manipulated artificially. By analyzing these data, abnormal changes in the business state of the enterprise can be detected in a timely manner, providing a more reliable basis for the post-loan management of financial institutions.

[0138] Specifically, the identification of the start time, working duration, and stop time can be achieved through the following steps:

[0139] First, compare the filtered modal component with a preset power threshold. Set a power threshold (which can be set according to actual needs). When the power value exceeds this threshold, it is considered that the corresponding equipment is in the on state; when it is lower than the threshold, it is considered that the corresponding equipment is in the off state.

[0140] Secondly, by analyzing the changing trend of the power value, the start time and stop time of the device can be determined. For example, when the power value suddenly rises from below the power threshold to above the power threshold, this time point can be marked as the start time. Correspondingly, when the power value suddenly drops from above the power threshold to below the power threshold, this time point can be marked as the stop time.

[0141] Thirdly, the working duration can be obtained by calculating the time difference between the start time and the stop time.

[0142] Among them, the change of the characteristic entropy can reflect the fluctuation of the power consumption load of the corresponding device. A stable characteristic entropy indicates that the power change of the corresponding device is relatively stable. Therefore, the change of the steady-state power value of the corresponding main production device can be characterized by the fluctuation of the characteristic entropy of the filtered modal components of the same main production device in each time period. For example, calculate the variance or standard deviation of the characteristic entropy of the filtered modal components of the same main production device in each time period as the change index of the steady-state power value of the corresponding main production device. The larger the change index of the steady-state power value is, the more unstable the power consumption load of the corresponding main production device is.

[0143] Specifically, step A6 includes:

[0144] A601. Calculate the outlier of the business status according to the start time, working duration, stop time and the change index of the steady-state power value;

[0145] A602. Judge whether the current business status of the enterprise is abnormal according to the outlier of the business status.

[0146] Among them, the typical start time, typical working duration, typical stop time and average change index of the steady-state power value of various main production devices of enterprises in the same industry as the monitored enterprise under normal business status can be statistically obtained in advance. In step A601, the average value of the deviations between each start time of the same main production device and the typical start time (denoted as the first average value), the average value of the deviations between each working duration and the typical working duration (denoted as the second average value), the average value of the deviations between each stop time and the typical stop time (denoted as the third average value), and the average value of the deviations between each change index of the steady-state power value and the average change index of the steady-state power value (denoted as the fourth average value) can be calculated. Then, calculate the weighted sum of the first average value, the second average value, the third average value and the fourth average value (the weights of each average value can be adjusted according to actual needs) as the outlier index of the corresponding main production device. Finally, calculate the outlier of the business status according to the outlier indexes of each main production device. For example, calculate the weighted sum (the weights of each main production device can be adjusted according to actual needs) of the outlier indexes of each main production device as the outlier of the business status, but not limited to this.

[0147] After calculating the abnormal value of the operating state, compare it with a pre-set abnormal threshold. If the abnormal value exceeds the abnormal threshold, it is determined that the current operating state of the enterprise is abnormal; if the abnormal value does not exceed the abnormal threshold, it is determined that the current operating state of the enterprise is normal. The abnormal threshold can be determined based on statistical analysis of historical data. For example, the mean of the abnormal values during the normal operation period plus twice the standard deviation can be used as the abnormal threshold.

[0148] In some preferred embodiments, see Figure 2 , after step A6, the method further includes the steps of:

[0149] A7. If the current operating state of the enterprise is normal, predict the estimated power consumption time series data of the enterprise based on the total power consumption time series data; the estimated power consumption time series data is the estimated data of the total power consumption time series data of the enterprise in the future;

[0150] A8. Evaluate the potential abnormal operating state risk of the enterprise based on the estimated power consumption time series data.

[0151] Based on the normal current operating state of the enterprise, predicting the future estimated power consumption time series data using the total power consumption time series data and evaluating the potential abnormal operating state risk based on this achieve the prediction and risk assessment of the future operating state of the enterprise. This method can not only detect potential risks in a timely manner but also provide a more forward-looking decision-making basis for financial institutions, effectively improving the efficiency and accuracy of post-loan management.

[0152] There are various methods to predict the estimated power consumption time series data of the enterprise. For example, time series analysis methods such as autoregressive integrated moving average model (ARIMA) or seasonal autoregressive integrated moving average model (SARIMA) can be used. These models can capture the trends, seasonality, and periodic characteristics of power consumption data, thereby predicting future power consumption situations.

[0153] Another possible implementation is to use machine learning algorithms such as long short-term memory network (LSTM) or gated recurrent unit (GRU). These deep learning models are particularly suitable for processing time series data and can learn long-term dependencies, thereby improving the accuracy of prediction.

[0154] In step A8, to evaluate the potential abnormal risk of the enterprise's operating status based on the estimated power consumption time series data, the abnormal value of the operating status corresponding to the estimated power consumption time series data can be calculated based on steps A1 - A6 using the estimated power consumption time series data. This is denoted as the estimated abnormal value of the operating status (specifically, substituting the estimated power consumption time series data into the total power consumption time series data in steps A1 - A6, and finally obtaining the estimated abnormal value of the operating status through calculation in step A601). Additionally, multiple abnormal risk levels of the operating status and their corresponding abnormal value ranges can be set, and the abnormal risk level of the enterprise's operating status can be determined based on the abnormal value range into which the estimated abnormal value of the operating status falls.

[0155] In this application, the total power consumption time series data of the enterprise obtained through the previous steps and the judged current operating status provide reliable basic data for predicting future power consumption and evaluating potential risks. This synergy makes the entire monitoring and early warning system more comprehensive and accurate, capable of not only monitoring the current status but also predicting future risks.

[0156] In summary, the present invention proposes a method for monitoring the operating status of an enterprise based on the power data of the enterprise. It obtains the total power consumption time series data of the enterprise, performs feature decomposition and noise filtering on the data, extracts features and matches them with reference features, calculates the working status data and judges the operating status of the enterprise, and predicts the potential abnormal risk of the enterprise's operating status. This realizes the in-depth analysis and accurate prediction of the enterprise's power load data. It solves the problems of data formality, information real-time nature, and automation and digitization in the data collection process in current bank post-loan management, reduces the workload of management personnel, and improves the supervision efficiency. It realizes the real-time monitoring and early warning of the enterprise's operating status, can improve the informatization level, reduce the workload of loan management personnel, ensure the authenticity and reliability of data, and improve the real-time nature and accuracy of data analysis. At the same time, based on the evaluation results, it can also provide decision-making support for the enterprise in aspects such as power management, equipment maintenance, and production planning.

[0157] In this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations.

[0158] The above description is only for the embodiments of the present application and is not intended to limit the protection scope of the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

Claims

1. A method for monitoring enterprise operating status based on enterprise power data, characterized in that: Includes steps: A1. Obtain the total power consumption time series data of the enterprise, and perform feature decomposition on the total power consumption time series data according to a preset time period length to obtain multiple modal components corresponding to each time period; A2. Obtain a reference power feature set of each major production equipment of the enterprise; the reference power feature set includes a reference time domain feature and a reference frequency domain feature; A3. Determine the validity of each of the modal components according to the characteristic entropy of the modal components, and perform noise filtering on the valid modal components to obtain filtered modal components; A4. extracting the time domain features and frequency domain features of each of the filtered modal components to form an actual power feature set corresponding to each of the filtered modal components, which is used to match the reference power feature set to identify the corresponding relationship between each of the filtered modal components and each of the main production equipment; A5. Obtain the working status data of each of the main production equipment according to the matching identification results corresponding to each of the time periods; A6. Determine the current operating status of the enterprise based on the working status data.

2. The enterprise operation status monitoring method based on enterprise power data according to claim 1 is characterized in that: Step A1 includes: A101. Obtain the total power consumption time series data of the enterprise for a preset time period; A102. Divide the total power time series data according to the time period length to obtain a plurality of total power time series data segments; A103. Perform feature decomposition on each of the total electric power time series data segments based on a modal decomposition algorithm to obtain multiple modal components corresponding to each of the time periods.

3. The enterprise operation status monitoring method based on enterprise power data according to claim 2 is characterized in that: In step A103, a variational mode decomposition algorithm is used to perform characteristic decomposition on each of the total power time series data segments.

4. The enterprise operation status monitoring method based on enterprise power data according to claim 1 is characterized in that: The reference time domain feature includes at least one of a reference mean, a reference standard deviation, a reference kurtosis, a reference mean square value, a reference root mean square value, a reference peak factor, a reference pulse factor, and a reference waveform factor; The reference frequency domain feature includes at least one of a reference center frequency, a reference bandwidth, a reference power spectrum density, a reference frequency variance, and a reference center of gravity frequency; The time domain characteristics include at least one of mean, standard deviation, kurtosis, mean square value, root mean square value, peak factor, pulse factor, and waveform factor; The frequency domain characteristics include at least one of center frequency, bandwidth, power spectrum density, frequency variance, and center of gravity frequency.

5. The enterprise operation status monitoring method based on enterprise power data according to claim 1 is characterized in that: Step A3 includes: A301. Calculate the characteristic entropy of each of the modal components; A302. Compare the characteristic entropy of each of the modal components with a preset characteristic entropy threshold, and determine the modal component whose characteristic entropy is lower than the characteristic entropy threshold as a valid modal component; A303. Perform noise filtering on the effective modal components to obtain filtered modal components.

6. The enterprise operation status monitoring method based on enterprise power data according to claim 5 is characterized in that: In step A303, a Wiener adaptive filter algorithm is used to perform noise filtering processing on the effective modal components.

7. The enterprise operation status monitoring method based on enterprise power data according to claim 1 is characterized in that: Step A4 includes: A401. Extracting the time domain features of each of the filtered modal components; A402. Extracting the frequency domain characteristics of each of the filtered modal components based on FFT transform; A403. Using the extracted time domain features and the frequency domain features to form an actual power feature set of each of the filtered modal components; A404. Compare the actual power characteristic set of each of the filtered modal components with the reference power characteristic set of each of the main production equipment, and determine the corresponding relationship between each of the filtered modal components and each of the main production equipment.

8. The enterprise operation status monitoring method based on enterprise power data according to claim 1 is characterized in that: The working status data includes the start time, working time, stop time and steady-state power value change index; Step A5 includes: According to the filtered modal components corresponding to each of the main production equipment in each of the time periods, identifying the start time, working time and stop time of each of the main production equipment; According to the characteristic entropy of the filtered modal component corresponding to each of the main production equipment in each of the time periods, the steady-state power value variation index of each of the main production equipment is calculated.

9. The enterprise operation status monitoring method based on enterprise power data according to claim 8 is characterized in that: Step A6 includes: Calculate an abnormal value of the operating state according to the start time, the working time, the stop time and the steady-state power value change index; Whether the current operating status of the enterprise is abnormal is determined according to the operating status abnormal value.

10. The enterprise operation status monitoring method based on enterprise power data according to claim 1 is characterized in that: After step A6, the method further includes the following steps: A7. If the current operating status of the enterprise is normal, predict the estimated power consumption time series data of the enterprise based on the total power consumption time series data; the estimated power consumption time series data is the estimated data of the total power consumption time series data of the enterprise in the future; A8. Evaluate the potential risk of abnormal operating status of the enterprise based on the estimated power consumption time series data.

Images