Equipment credit evaluation system and method for enterprise lease

Abstract

The invention discloses an enterprise lease-oriented equipment credit evaluation system and method, relates to the technical field of equipment data processing, and aims to synchronously quantify the running temperature, power consumption and heat dissipation state of equipment by forming an original thermal state data set Hraw and avoid one-sided judgment caused by traditional single-parameter monitoring. Based on a thermal state feature set Hset of a temperature stability index Tstb, a power consumption fluctuation spectrum index Pspc and a heat dissipation anomaly integral Qint, a three-dimensional feature space reflecting equipment operation stability and energy balance can be constructed, a thermal state feature template set Htpl is further formed through clustering, and dynamic reference establishment in different operation modes is achieved. And comparing the real-time feature vector with the thermal state feature template set Htpl to calculate a thermal balance index Hc, and generating an equipment credit score Sc in combination with a historical credit mean value score Sh, so that quantitative credit of enterprise equipment use behaviors can be realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of device data processing, in particular to a device credit evaluation system and method for enterprise leasing. BACKGROUND

[0002] With the continuous advancement of enterprise digital office and cloud deployment, more and more enterprises obtain IT device resources through leasing to reduce fixed asset investment and maintain the flexibility of technology update.

[0003] At present, the existing device credit evaluation method mainly depends on single data sources such as leasing records, payment behavior or operation logs, and lacks multi-dimensional analysis of the running state of the device. In the actual leasing process, the hot state characteristics of the device running, such as temperature fluctuation, power consumption change and heat dissipation performance, are often regarded as parameters related to the physical health of the device, and are not effectively introduced into the credit evaluation system. The traditional evaluation system cannot distinguish whether the enterprise causes the device to run at high temperature for a long time, the power consumption fluctuates sharply or the heat dissipation efficiency is abnormal due to improper use. These thermal state imbalance phenomena not only reflect the non-standard behavior of the enterprise in using the device, but also directly affect the service life of the device and the subsequent re-rental value. Due to the lack of quantitative analysis model based on thermal energy characteristics, the leasing platform has defects such as information lag and untimely risk identification in credit judgment, which may cause inaccurate credit differentiation between high-risk enterprises and low-risk enterprises. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a device credit evaluation system and method for enterprise leasing, which solves the problems mentioned in the background art.

[0005] To achieve the above purpose, the present application realizes the following technical scheme: a device credit evaluation method for enterprise leasing, comprising the following steps:

[0006] S1, in the collection window Win, the temperature, power consumption and heat dissipation data of the IT device leased by the enterprise are collected, and time synchronization, abnormality elimination and normalization processing are performed to form an original thermal state data set Hraw;

[0007] S2, taking the original thermal state data set Hraw as the extraction basis, extracting the temperature stability index Tstb, the power consumption fluctuation spectrum index Pspc and the heat dissipation abnormal integral Qint, and integrating to form a thermal state feature set Hset; then based on the thermal state feature set Hset, clustering analysis is carried out to establish a thermal state feature template set Htpl;

[0008] S3, taking the real-time feature vector of the current IT device as the basis, comparing with the thermal state feature template set Htpl, calculating the heat balance index Hc, and combining with the historical credit average score Sh to generate the device credit score Sc;

[0009] S4, generating a credit level Lc according to the interval range of the device credit score Sc, and uploading the credit level Lc and the device credit score Sc to a cloud credit evaluation platform for display and updating.

[0010] Preferably, the S1 comprises S11.

[0011] S11, in the collection window Win, taking the enterprise-leased IT device as the collection subject, collecting real-time data through the multi-source sensor components deployed in the device mainboard, power supply circuit and heat dissipation channel; the sensor components comprise a temperature collection sensor, a power monitoring sensor and a heat dissipation detection sensor;

[0012] In the real-time data collection process, monitoring is performed at a sampling frequency Fs as a sampling period:

[0013] The temperature collection sensor is installed at the positions of the device processor, storage chip and power module, periodically detects the local temperature signal, forms a continuous temperature time sequence at the sampling frequency Fs, and generates a temperature sequence Tseq;

[0014] The power monitoring sensor is arranged between the device input power supply line and the mainboard power supply node, obtains the power consumption value varying with the sampling frequency Fs by detecting the instantaneous product of the voltage and current signals and performing time integration, and forms a power consumption change sequence Pchg;

[0015] The heat dissipation detection sensor is arranged at the heat dissipation vent and heat exchange area, obtains the heat dissipation output power per unit time by collecting the airflow temperature difference and flow rate data, and obtains a heat dissipation power curve Qcur;

[0016] Among them, all sensor components sample under a unified time synchronization protocol, and record time stamps in the collected data for consistency of different signals in time sequence, after the real-time data collection is completed, aggregate and cache various data in time sequence to generate an original data cache set Hbuf.

[0017] Preferably, the S1 further comprises S12.

[0018] S12, uploading to the cloud processor based on the original data cache set Hbuf to perform a data preprocessing process, specifically comprising the following processing steps:

[0019] S121, time alignment and clock correction: calculating the time stamp difference point by point for the temperature sequence Tseq, the power consumption change sequence Pchg and the heat dissipation power curve Qcur in the original data cache set Hbuf to obtain a time offset Δt;

[0020] The sliding time window regression algorithm is used to correct all time series in synchronization, with the time deviation among the three as the objective function, so that the corrected temperature sequence Tseq, power consumption change sequence Pchg and heat dissipation power curve Qcur remain consistent under the same time reference.

[0021] In S122, abnormality detection and missing value repair, abnormality detection is performed on the time-corrected temperature sequence Tseq, power consumption change sequence Pchg and heat dissipation power curve Qcur. When the value of any sequence exceeds the preset range [Min, Max], the sequence is marked as an abnormal point and excluded. When a short-time missing interval is detected, it is completed by a neighborhood linear interpolation algorithm.

[0022] In S123, normalization and structured arrangement, the time-corrected temperature sequence Tseq, power consumption change sequence Pchg and heat dissipation power curve Qcur are normalized to eliminate the dimensional differences of different physical quantities, and the values are mapped to the interval [0, 1]. The normalization is performed by using the maximum and minimum normalization method.

[0023] The signal data processed by steps S121, S122 and S123 are reorganized into a structured data table in chronological order to generate the original thermal state data set Hraw.

[0024] Preferably, S2 includes S21.

[0025] In S21, the original thermal state data set Hraw is input, and multi-dimensional thermal state feature indexes reflecting the device running stability and heat dissipation health degree are extracted, including temperature stability index Tstb, power consumption fluctuation spectrum index Pspc and heat dissipation abnormal integral Qint. The temperature stability index Tstb is obtained by sliding time window analysis of the temperature sequence Tseq, with the time window length Wm.

[0026] The power consumption fluctuation spectrum index Pspc is obtained by fast Fourier transform (FFT) of the power consumption change sequence Pchg.

[0027] The heat dissipation abnormal integral Qint is obtained by calculating the heat dissipation deviation area per unit time of the heat dissipation power curve Qcur.

[0028] Let the ideal heat dissipation power be Qref, then the heat dissipation abnormal integral Qint extraction formula is as follows:

[0029] Qref=∫|Qcur-Qref|dt.

[0030] In the formula, dt represents the integral of time t.

[0031] Preferably, S2 further includes S22.

[0032] S22, the temperature stability index Tstb, the power consumption fluctuation spectrum index Pspc and the heat dissipation anomaly integral Qint in the hot state feature set Hset are subjected to numerical standardization processing, and are mapped to the interval [0, 1];

[0033] Subsequently, similarity clustering analysis is performed on all samples based on the temperature stability index Tstb, the power consumption fluctuation spectrum index Pspc and the heat dissipation anomaly integral Qint, and according to the dominance of the three features, the samples are automatically divided into three clustering categories:

[0034] When the temperature stability index Tstb has the highest proportion in the samples, it is divided into the clustering category C1;

[0035] When the power consumption fluctuation spectrum index Pspc has the highest proportion in the samples, it is divided into the clustering category C2;

[0036] When the heat dissipation anomaly integral Qint has the highest proportion in the samples, it is divided into the clustering category C3;

[0037] After clustering, the number of samples and feature statistical parameters in each clustering category Ck (k = 1, 2, 3) are calculated, including the mean and the dispersion; each clustering category Ck represents a thermal operation mode, wherein: the clustering category C1 corresponds to the thermal operation mode of the temperature stability dominant device, the clustering category C2 corresponds to the thermal operation mode of the power consumption fluctuation dominant device, and the clustering category C3 corresponds to the thermal operation mode of the heat dissipation deviation dominant device.

[0038] Preferably, the S2 further comprises S23 and S24;

[0039] S23, for each clustering category Ck, the system calculates the average value of the three feature indexes to obtain a template center vector, including the temperature stability index center value Tstb_c, the power consumption fluctuation spectrum index center value Pspc_c and the heat dissipation anomaly integral center value Qint_c;

[0040] The acquisition formula of the template stability score Hstb is defined based on the principle of thermal equilibrium: Hstb = [Tstb_c + Pspc_c + (1-Qint_c)] / 3; wherein the template stability score Hstb is used to measure the comprehensive thermal equilibrium stability of the clustering category Ck in the temperature stability index center value Tstb_c, the power consumption fluctuation spectrum index center value Pspc_c and the heat dissipation anomaly integral center value Qint_c three dimensions;

[0041] S24, the template center vector and its template stability score Hstb corresponding to the three clustering categories Ck are summarized to form a hot state feature template set Htpl; meanwhile, a unique template identification code Tpl_id is assigned to each template in the hot state feature template set Htpl, and an index structure is established.

[0042] Preferably, the S3 comprises S31;

[0043] S31, in the collection window Win, the original hot state data set Hraw of the enterprise rental IT equipment currently being evaluated is taken as the input to calculate the real-time temperature stability index Tstb_r, the real-time power consumption fluctuation spectrum index Pspc_r, and the real-time heat dissipation abnormal integral Qint_r of the IT equipment; the three real-time indexes obtained are compared with the template center vector corresponding to each template in the hot state feature template set Htpl for similarity, and the similarity comparison is obtained by using the Euclidean distance calculation to obtain the feature deviation distance Dd;

[0044] The feature deviation distance Dd calculation formula is as follows: ;

[0045] The feature deviation distance Dd of each IT equipment and each template is calculated in turn, and the minimum value is selected as the feature deviation distance minimum value Dd_min, and the corresponding template is confirmed as the optimal matching template of the IT equipment;

[0046] The real-time temperature stability index Tstb_r is the current temperature stability index of the equipment calculated based on the temperature sequence Tseq in the collection window Win;

[0047] The real-time power consumption fluctuation spectrum index Pspc_r is the power consumption fluctuation spectrum index calculated based on the power consumption change sequence Pchg in the collection window Win;

[0048] The real-time heat dissipation abnormal integral Qint_r is the heat dissipation abnormal integral calculated based on the heat dissipation power curve Qcur in the collection window Win.

[0049] Preferably, the S3 further comprises S32;

[0050] S32, based on the template stability score Hstb and the feature deviation distance minimum value Dd_min, the heat balance index Hc of the current IT equipment is calculated as Hc=Hstb×(1-Dd_min); and in a plurality of consecutive collection windows, the time weighted average value is calculated to form a time weighted heat balance index Hcw;

[0051] The time weighted heat balance index Hcw calculation formula is as follows:

[0052] ;

[0053] In the formula, Hc_b is the thermal balance index of the bth collection window, and W_b represents the time weight of the bth collection window.

[0054] The device credit score Sc of the current IT device is calculated in combination with the historical credit average score Sh stored by the IT device.

[0055] The device credit score Sc is calculated according to the following formula:

[0056] Sc = a x Hcw + b x Sh.

[0057] In the formula, a and b represent weight coefficients of the time-weighted thermal balance index Hcw and the historical credit average score Sh respectively, and a + b = 1.

[0058] Preferably, the S4 comprises S41.

[0059] S41, the device credit score Sc is mapped and converted according to a preset credit level interval to obtain a credit level Lc, the credit level Lc obtained by determination and the corresponding device credit score Sc are encapsulated to form a credit feedback data packet, and the credit feedback data packet is uploaded to a leasing device credit evaluation cloud platform through a cloud communication interface.

[0060] The credit level Lc is obtained by the following mapping and conversion method:

[0061] When the device credit score Sc is greater than or equal to 0.85, the current IT device is determined to be a high credit level Lc_H.

[0062] When 0.65 is less than the device credit score Sc, the current IT device is determined to be a medium credit level Lc_M.

[0063] When the device credit score Sc is less than 0.65, the current IT device is determined to be a low credit level Lc_L.

[0064] A device credit evaluation system for enterprise leasing comprises a thermal data collection module, a thermal feature extraction module, a score calculation and generation module, and a decision feedback module.

[0065] The thermal data collection module collects temperature, power consumption, and heat dissipation data of the IT device leased by the enterprise within a collection window Win, and performs time synchronization, abnormality elimination, and normalization processing to form a raw thermal data set Hraw.

[0066] The thermal state feature extraction module extracts a temperature stability index Tstb, a power consumption fluctuation spectrum index Pspc and a heat dissipation abnormal integral Qint based on the original thermal state data set Hraw, and integrates to form a thermal state feature set Hset; and then based on the thermal state feature set Hset, clustering analysis is performed to establish a thermal state feature template set Htpl;

[0067] The score calculation generation module compares the real-time feature vector of the current IT equipment with the thermal state feature template set Htpl, calculates a thermal balance index Hc, and combines a historical credit average score Sh to generate an equipment credit score Sc;

[0068] The decision feedback module generates a credit level Lc according to the interval range of the equipment credit score Sc, and uploads the credit level Lc and the equipment credit score Sc to a cloud credit evaluation platform for display and update.

[0069] The present application provides a kind of enterprise leasing-oriented equipment credit evaluation system and method, with the following beneficial effects:

[0070] (1) Through the multi-source thermal state data monitoring and preprocessing process established in the acquisition window Win, the original thermal state data set Hraw is formed, so that the temperature, power consumption and heat dissipation state of equipment operation are quantified synchronously, and the one-sided judgment caused by traditional single parameter monitoring is avoided. The thermal state feature set Hset based on the temperature stability index Tstb, the power consumption fluctuation spectrum index Pspc and the heat dissipation abnormal integral Qint can construct a three-dimensional feature space reflecting the stability and energy balance of equipment operation, and further form a thermal state feature template set Htpl through clustering, to realize the dynamic benchmark establishment under different operation modes. Comparing the real-time feature vector with the thermal state feature template set Htpl to calculate the thermal balance index Hc, and then combining the historical credit average score Sh to generate the equipment credit score Sc, can realize the quantitative credit of enterprise equipment usage behavior.

[0071] (2) By taking the original hot state data set Hraw as input, the system can synchronously extract the temperature stability index Tstb, the power consumption fluctuation spectrum index Pspc and the heat dissipation abnormal integral Qint three core characteristic indexes, so that the temperature fluctuation, load energy distribution and heat dissipation performance of the device running are comprehensively characterized. The multi-dimensional feature extraction process breaks the limitation of traditional single parameter monitoring, and realizes the quantitative description of the energy level of the device running state. Through clustering analysis, the hot state feature set Hset is automatically divided into three clustering categories C1, C2 and C3, and based on the clustering category center vector (temperature stability index center value Tstb_c, power consumption fluctuation spectrum index center value Pspc_c, heat dissipation abnormal integral center value Qint_c) is calculated, a hot state feature template set Htpl is established, so that devices with different thermal behavior modes have independent control standards. Not only can the temperature stability dominant type, power consumption fluctuation dominant type and heat dissipation deviation dominant type devices be dynamically distinguished, but also a unified thermal balance evaluation benchmark can be established through the template stability score Hstb, realizing the consistency of credit evaluation across device types and running cycles.

[0072] (3) By calculating the real-time temperature stability index Tstb_r, the real-time power consumption fluctuation spectrum index Pspc_r and the real-time heat dissipation abnormal integral Qint_r in the collection window Win, and comparing with the template center vector in the hot state feature template set Htpl, the system can obtain the feature deviation distance Dd in real time, and determine the optimal matching template, so that the running state of the device has a dynamic reference benchmark in the thermal behavior dimension. The thermal balance index Hc calculated based on the template stability score Hstb and the minimum value Dd_min of the feature deviation distance can reflect the energy balance degree and stability level of the device running, and the time weighted thermal balance index Hcw formed by the time weighted average of multiple collection windows can ensure that the credit evaluation result not only reflects the instantaneous state, but also has the trend judgment ability. BRIEF DESCRIPTION OF DRAWINGS

[0073] Figure 1 It is a device credit evaluation method for enterprise leasing of the application;

[0074] Figure 2 It is a device credit evaluation system block diagram for enterprise leasing of the application;

[0075] Figure 3 It is a time evolution trend diagram of the thermal balance index Hc and the device credit score Sc. DETAILED DESCRIPTION

[0076] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0077] Embodiment 1

[0078] The present application provides a device credit evaluation method for enterprise leasing, please refer to Figure 1 , comprising the following steps:

[0079] S1, in the collection window Win, the temperature, power consumption and heat dissipation data of the IT equipment of the enterprise leasing are collected, and time synchronization, abnormality elimination and normalization processing are performed, to form a raw thermal data set Hraw;

[0080] S2, taking the raw thermal data set Hraw as the extraction basis, the temperature stability index Tstb, the power consumption fluctuation spectrum index Pspc and the heat dissipation abnormal integral Qint are extracted, and after integration, the thermal state feature set Hset is formed; then based on the thermal state feature set Hset, clustering analysis is carried out, and the thermal state feature template set Htpl is established;

[0081] S3, taking the real-time feature vector of the current IT equipment as the basis, comparing with the thermal state feature template set Htpl, calculating the heat balance index Hc, and combining with the historical credit average score Sh to generate the device credit score Sc;

[0082] S4, according to the interval range of the device credit score Sc, the credit level Lc is generated, and the credit level Lc and the device credit score Sc are uploaded to the cloud credit evaluation platform for display and update.

[0083] In this embodiment, through the multi-source thermal data monitoring and preprocessing process established in the collection window Win, the original thermal data set Hraw is formed, so that the temperature, power consumption and heat dissipation state of the device running are quantified synchronously, and one-sided judgment caused by traditional single parameter monitoring is avoided. Based on the thermal state feature set Hset including the temperature stability index Tstb, the power consumption fluctuation spectrum index Pspc and the heat dissipation abnormal integral Qint, a three-dimensional feature space reflecting the running stability and energy balance of the device can be constructed, and further through clustering, a thermal state feature template set Htpl is formed, so that dynamic benchmark establishment under different running modes is realized. By comparing the real-time feature vector with the thermal state feature template set Htpl to calculate the heat balance index Hc, and then combining the historical credit average score Sh to generate the device credit score Sc, the quantitative credit description of the enterprise device usage behavior can be realized. For example, when the device maintains a high temperature stability index Tstb and power consumption fluctuation spectrum index Pspc in a long-term leasing period, and the heat dissipation abnormal integral Qint is stable at a low level, the device credit score Sc generated by the system will be significantly improved, indicating that the enterprise maintains well and the device runs stably, and the credit level Lc is automatically adjusted upward. Conversely, when the heat dissipation efficiency of the device decreases or the power consumption fluctuation abnormally increases, the heat balance index Hc decreases, and the credit level Lc is immediately adjusted downward. The system can display the abnormal state in real time on the credit evaluation cloud platform and trigger a risk prompt. Thus, the method realizes dynamic quantification of the running state of the enterprise leased IT device, real-time feedback of the credit level and early identification of the risk trend, and significantly improves the accuracy, real-time performance and interpretability of the device credit evaluation.

[0084] Embodiment 2

[0085] Specifically, the S1 includes S11;

[0086] S11, in the collection window Win, taking the enterprise leased IT device as the collection subject, real-time data collection is performed through the multi-source sensor components deployed in the device mainboard, power supply circuit and heat dissipation channel; the sensor components include temperature collection sensors, power monitoring sensors and heat dissipation detection sensors;

[0087] In the real-time data collection process, monitoring is performed at a sampling frequency Fs as a sampling period:

[0088] The temperature collection sensors are installed at the positions of the device processor, storage chip and power module, periodically detect the local temperature signal, form a continuous temperature time sequence at the sampling frequency Fs, and generate a temperature sequence Tseq;

[0089] The power monitoring sensors are arranged between the device input power supply line and the mainboard power supply node, the instantaneous product of the voltage and current signals is detected and time integration is performed to obtain the power consumption value varying with the sampling frequency Fs, and a power consumption change sequence Pchg is formed.

[0090] The heat dissipation detection sensor is arranged at the heat dissipation vent and the heat exchange area, and the heat dissipation output power per unit time is calculated by collecting airflow temperature difference and flow rate data to obtain a heat dissipation power curve Qcur;

[0091] All sensor components are sampled under a unified time synchronization protocol, and a time stamp is recorded in the collected data for consistency of different signals in time sequence. After real-time data collection is completed, all types of data are aggregated and cached in time sequence to generate an original data cache set Hbuf, which provides an input basis for subsequent data preprocessing;

[0092] The temperature sequence Tseq represents a set of temperatures of each key element of the device in the collection window Win over time, which is used to reflect the thermal stability and heat dissipation state of the device;

[0093] The power consumption change sequence Pchg is a time sequence of power input and output during device operation, which is used to characterize load fluctuation characteristics;

[0094] The heat dissipation power curve Qcur is a function curve of heat dissipation output power per unit time over time, which is used to reflect the heat dissipation capacity of the device.

[0095] The S1 further includes S12;

[0096] S12, based on the original data cache set Hbuf, uploads to a cloud processor to perform a data preprocessing process, which is used to improve the accuracy, continuity and calculability of thermal state data, and specifically includes the following processing steps:

[0097] S121, time alignment and clock correction: the time stamp difference of the temperature sequence Tseq, the power consumption change sequence Pchg and the heat dissipation power curve Qcur in the original data cache set Hbuf is calculated point by point to obtain a time offset Δt;

[0098] Using a sliding time window regression algorithm, the time offset Δt is used as the objective function to minimize the time offset between the three, and all time sequences are corrected synchronously to make the corrected temperature sequence Tseq, the power consumption change sequence Pchg and the heat dissipation power curve Qcur consistent under the same time reference;

[0099] S122, anomaly detection and missing data repair: the temperature sequence Tseq, the power consumption change sequence Pchg and the heat dissipation power curve Qcur after time correction are subjected to anomaly detection; when the value of any sequence exceeds the preset range [Min, Max], the sequence is marked as an abnormal point and is excluded; when a short-time missing interval is detected, it is completed by a neighborhood linear interpolation algorithm to ensure the continuity of the sequence;

[0100] S123, normalization and structured arrangement: the temperature sequence Tseq, the power consumption change sequence Pchg and the heat dissipation power curve Qcur after the elimination and repair are normalized, the numerical value is mapped to the interval [0, 1], and the dimensional difference of different physical quantities is eliminated, wherein the normalization is processed by using the maximum and minimum normalization method;

[0101] The signal data processed by steps S121, S122 and S123 are reorganized into a structured data table in chronological order to generate a raw thermal state data set Hraw;

[0102] Wherein, the normalization ensures that different thermal characteristic indicators, including temperature stability indicator Tstb, power consumption fluctuation spectrum indicator Pspc and heat dissipation abnormal integral Qint, are calculated in a unified numerical space, improving model stability and calculation comparability.

[0103] In this embodiment, by arranging multi-source sensor assemblies in the collection window Win, temperature sequence Tseq, power consumption change sequence Pchg and heat dissipation power curve Qcur are formed respectively, and the system can capture key thermal physical changes in the whole process of device operation, so that the collection results cover the core areas such as device mainboard, power supply and heat dissipation system, and truly reflect the thermal behavior characteristics of IT equipment. Based on the raw data buffer set Hbuf, the cloud processor performs time alignment, abnormality elimination and missing data repair and other preprocessing operations, which can effectively eliminate the distortion caused by sensor drift, network delay and sampling error. For example, in an enterprise office environment, some devices may have decreased heat dissipation efficiency or frequent power fluctuations due to long-term operation, and traditional single-point detection is prone to missed judgment. However, by unified time correction and continuity repair of power consumption change sequence Pchg and heat dissipation power curve Qcur, this method can accurately identify thermal abnormal trends and maintain data integrity. The raw thermal state data set Hraw generated after normalization allows temperature stability indicator Tstb, power consumption fluctuation spectrum indicator Pspc and heat dissipation abnormal integral Qint to be directly calculated and compared in a unified numerical space, thereby providing a highly consistent input basis for subsequent thermal balance models. Thus, this step realizes the whole process optimization from raw collection to structured data, ensures that the thermal characteristic data of enterprise rental IT equipment under different operating conditions are complete and comparable, and lays a foundation for the stability and accuracy of subsequent credit evaluation.

[0104] Embodiment 3

[0105] Specifically, the S2 includes S21;

[0106] S21, taking the original thermal state data set Hraw as input, extracting a multi-dimensional thermal state feature index reflecting the device running stability and heat dissipation health degree from it, including a temperature stability index Tstb, a power consumption fluctuation spectrum index Pspc and a heat dissipation abnormal integral Qint; the temperature stability index Tstb is obtained by performing sliding time window analysis on the temperature sequence Tseq, setting the time window length as Wm, and calculating the temperature mean square error in each time window;

[0107] The specific extraction formula is as follows:

[0108] Tstb=1-Var(Tseq,Wm) / Varmax;

[0109] In the formula, Var(Tseq,Wm) represents the variance of the temperature sequence Tseq in the window Wm, and Varmax represents the maximum value of the device historical temperature variance;

[0110] Wherein, when the temperature stability index Tstb is closer to 1, the temperature fluctuation is smaller, and the device thermal stability is higher;

[0111] The power consumption fluctuation spectrum index Pspc is obtained by performing fast Fourier transform (FFT) on the power consumption change sequence Pchg, setting the main energy concentration frequency band power as Pf, and the total frequency band power as Pt, and the power consumption fluctuation spectrum index Pspc extraction formula is as follows:

[0112] Pspc=Pf / Pt;

[0113] Wherein, when the power consumption fluctuation spectrum index Pspc is higher, it means that the device energy consumption is concentrated and the running load is stable; if the power consumption fluctuation spectrum index Pspc is lower, it means that the load fluctuation is larger, and there may be abnormal tasks or overload running;

[0114] The heat dissipation abnormal integral Qint is obtained by calculating the heat dissipation deviation area in unit time based on the heat dissipation power curve Qcur;

[0115] Setting the ideal heat dissipation power as Qref, the heat dissipation abnormal integral Qint extraction formula is as follows:

[0116] Qref=∫|Qcur-Qref|dt;

[0117] In the formula, dt represents the integral of time t; when the heat dissipation abnormal integral Qint value is smaller, it means that the heat dissipation performance is stable and the thermal management control is good.

[0118] The S2 further includes S22;

[0119] S22, the temperature stability index Tstb, power consumption fluctuation spectrum index Pspc and heat dissipation abnormal integral Qint in the thermal state feature set Hset are subjected to numerical standardization processing, which is mapped to the interval [0, 1], so as to ensure that each feature index has comparability in the same numerical space;

[0120] Subsequently, similarity clustering analysis is performed on all samples based on the temperature stability index Tstb, power consumption fluctuation spectrum index Pspc and heat dissipation abnormal integral Qint, and the samples are automatically divided into three clustering categories according to the dominance of the three features:

[0121] When the temperature stability index Tstb has the highest proportion in the sample, it is divided into clustering category C1;

[0122] When the power consumption fluctuation spectrum index Pspc has the highest proportion in the sample, it is divided into clustering category C2;

[0123] When the heat dissipation abnormal integral Qint has the highest proportion in the sample, it is divided into clustering category C3;

[0124] After clustering, the number of samples and feature statistical parameters in each clustering category Ck (k=1, 2, 3) are calculated, including mean and dispersion, which are used for subsequent template center vector (Tstb_c, Pspc_c, Qint_c) calculation; each clustering category Ck represents a thermal operation mode, wherein: the clustering category C1 corresponds to the thermal operation mode of the temperature stability dominant device, the clustering category C2 corresponds to the thermal operation mode of the power consumption fluctuation dominant device, and the clustering category C3 corresponds to the thermal operation mode of the heat dissipation deviation dominant device.

[0125] The S2 further includes S23 and S24;

[0126] S23, for each clustering category Ck, the system calculates the average value of the three feature indexes to obtain the template center vector, including the temperature stability index center value Tstb_c, the power consumption fluctuation spectrum index center value Pspc_c and the heat dissipation abnormal integral center value Qint_c;

[0127] Based on the principle of thermal equilibrium, the acquisition formula of the template stability score Hstb is defined as: Hstb=[Tstb_c+Pspc_c+(1-Qint_c)] / 3; wherein the template stability score Hstb is used to measure the comprehensive thermal balance stability of the clustering category Ck in the temperature stability index center value Tstb_c, the power consumption fluctuation spectrum index center value Pspc_c and the heat dissipation abnormal integral center value Qint_c three dimensions; when the template stability score Hstb is close to 1, it indicates that the device runs stably, the power consumption is concentrated, and the heat dissipation is less abnormal, reflecting that the enterprise performs well in device use and maintenance;

[0128] S24, the template center vectors corresponding to the three clustering categories Ck and their template stability scores Hstb are summarized to form a hot state feature template set Htpl; meanwhile, a unique template identification code Tpl_id is assigned to each template in the hot state feature template set Htpl, and an index structure is established to support subsequent calling and comparison operations; in addition, a similarity matrix between templates is calculated for dynamic template updating and new sample matching, so that the hot state feature template set Htpl remains timely and representative.

[0129] In this embodiment, by taking the original hot state data set Hraw as input, the system can simultaneously extract the temperature stability index Tstb, the power consumption fluctuation spectrum index Pspc, and the heat dissipation abnormal integral Qint, three core feature indexes, so that the temperature fluctuation, load energy distribution, and heat dissipation performance of the device operation are comprehensively characterized. This multi-dimensional feature extraction process breaks the limitations of traditional single-parameter monitoring and realizes quantitative description of the energy level of the device operation state. For example, in daily rental use of enterprises, if the device causes frequent fluctuations in power consumption due to task concentration while the temperature remains stable, the traditional model is difficult to identify potential energy distribution abnormalities, but through the joint calculation of the power consumption fluctuation spectrum index Pspc and the temperature stability index Tstb, the abnormal energy consumption characteristics can be revealed, thereby more accurately evaluating the operation stability. Further, the system automatically divides the hot state feature set Hset into three clustering categories C1, C2, and C3 through clustering analysis, and calculates the template center vectors (temperature stability index center value Tstb_c, power consumption fluctuation spectrum index center value Pspc_c, and heat dissipation abnormal integral center value Qint_c) based on each clustering category, establishes a hot state feature template set Htpl, so that devices with different heat behavior modes have independent control standards. Thus, this method not only can dynamically distinguish temperature stability dominant, power consumption fluctuation dominant, and heat dissipation deviation dominant devices, but also can establish a unified thermal balance evaluation benchmark through the template stability score Hstb, realizing consistency in credit evaluation across device types and operation cycles. For example, for a server that has been running for a long time but has degraded heat dissipation performance, its heat dissipation abnormal integral Qint_c is significantly high, and the template stability score Hstb is reduced, which ultimately reflects a decrease in the credit score in credit evaluation.

[0130] Embodiment 4

[0131] Please refer to Figure 1 and Figure 3 , specifically: the S3 includes S31;

[0132] S31, inputting the original hot state data set Hraw of the enterprise rental IT equipment currently being evaluated in the collection window Win, calculating the real-time temperature stability index Tstb_r, the real-time power consumption fluctuation spectrum index Pspc_r and the real-time heat dissipation abnormal integral Qint_r of the IT equipment; comparing the three real-time indexes obtained with the template center vector corresponding to each template in the hot state feature template set Htpl, and obtaining the feature deviation distance Dd by using the Euclidean distance calculation;

[0133] The feature deviation distance Dd calculation formula is as follows: ;

[0134] The feature deviation distance Dd of each IT equipment and each template is calculated in turn, and the minimum value is selected as the feature deviation distance minimum value Dd_min, and the corresponding template is confirmed as the optimal matching template of the IT equipment; the template stability score Hstb of the template is recorded synchronously as the reference input for subsequent heat balance index Hc calculation;

[0135] The real-time temperature stability index Tstb_r is the current temperature stability index of the equipment calculated based on the temperature sequence Tseq in the collection window Win.

[0136] The real-time power consumption fluctuation spectrum index Pspc_r is the power consumption fluctuation spectrum index calculated based on the power consumption change sequence Pchg in the collection window Win.

[0137] The real-time heat dissipation abnormal integral Qint_r is the heat dissipation abnormal integral calculated based on the heat dissipation power curve Qcur in the collection window Win.

[0138] The S3 further includes S32.

[0139] S32, based on the template stability score Hstb and the feature deviation distance minimum value Dd_min, calculating the heat balance index Hc of the current IT equipment Hc=Hstb×(1-Dd_min); and calculating the time weighted average value to form the time weighted heat balance index Hcw in a plurality of continuous collection windows.

[0140] The time weighted heat balance index Hcw calculation formula is as follows:

[0141] ;

[0142] In the formula, Hc_b is the thermal balance index of the bth collection window, and W_b represents the time weight of the bth collection window, which is automatically determined according to the equipment operation time period set in the enterprise rental system, and is used to reflect the difference in working intensity of the equipment in different operation time periods; wherein, a high-load working period corresponds to a larger weight, and a low-load or non-working period corresponds to a smaller weight.

[0143] In combination with the historical credit average score Sh stored by the IT equipment, the device credit score Sc of the current IT equipment is calculated.

[0144] The device credit score Sc is calculated according to the following formula:

[0145] Sc=α×Hcw+β×Sh.

[0146] In the formula, α and β represent the weight coefficients of the time-weighted thermal balance index Hcw and the historical credit average score Sh respectively, and α+β=1, and the specific values are set by the user.

[0147] The S4 comprises S41.

[0148] S41, the device credit score Sc is mapped and converted according to the preset credit level interval to obtain the credit level Lc, and the obtained credit level Lc and the corresponding device credit score Sc are encapsulated to form a credit feedback data packet, which is uploaded to the rental equipment credit evaluation cloud platform through a cloud communication interface.

[0149] The credit level Lc is obtained by the following mapping and conversion method:

[0150] When the device credit score Sc is greater than or equal to 0.85, the current IT equipment is determined to be a high credit level Lc_H.

[0151] When 0.65≤device credit score Sc<0.85, the current IT equipment is determined to be a medium credit level Lc_M.

[0152] When the device credit score Sc is less than 0.65, the current IT equipment is determined to be a low credit level Lc_L.

[0153] In this embodiment, by calculating the real-time temperature stability index Tstb_r, the real-time power consumption fluctuation spectrum index Pspc_r and the real-time heat dissipation abnormal integral Qint_r in the collection window Win, and performing similarity comparison with the template center vector in the thermal state feature template set Htpl, the system can obtain the feature deviation distance Dd in real time, and determine the optimal matching template, so that the running state of the device has a dynamic reference benchmark in the thermal behavior dimension. The thermal balance index Hc calculated based on the template stability score Hstb and the minimum value Dd_min of the feature deviation distance can reflect the energy balance degree and stability level of the device running, and the time-weighted thermal balance index Hcw formed by time-weighted average of multiple collection windows can ensure that the credit evaluation result not only reflects the instantaneous state, but also has the ability of trend judgment. For example, when the server rented by the enterprise runs at high load during the day and stands by at low load at night, the system will give greater weight to the thermal balance index Hc of the high load period according to the time weight W_b, so that the device credit score Sc can more truly reflect the actual use intensity and maintenance level of the enterprise. Combined with the weighted fusion of the historical credit average score Sh, the obtained device credit score Sc not only considers the current running behavior, but also takes into account the long-term performance, ensuring the stability and continuity of the evaluation result. Further, the system maps the device credit score Sc to the credit level Lc and uploads it to the cloud credit evaluation platform, which can realize the visual display and dynamic update of the credit result. For example, when the thermal balance index Hc of a certain device continuously decreases due to the blockage of the heat dissipation system, the system will automatically adjust the credit level Lc from the high credit level Lc_H to the medium credit level Lc_M, and mark the abnormal trend in the cloud interface, so that the lessee can arrange maintenance in advance.

[0154] Embodiment 5

[0155] An enterprise rental-oriented device credit evaluation system, please refer to Figure 2 , specifically: including a thermal state data collection module, a thermal state feature extraction module, a score calculation and generation module, and a decision feedback module;

[0156] The thermal state data collection module collects temperature, power consumption and heat dissipation data of the IT device rented by the enterprise within the collection window Win, and performs time synchronization, abnormal elimination and normalization processing to form the original thermal state data set Hraw;

[0157] The thermal state feature extraction module extracts the temperature stability index Tstb, the power consumption fluctuation spectrum index Pspc and the heat dissipation abnormal integral Qint based on the original thermal state data set Hraw, and integrates them to form the thermal state feature set Hset; then based on the thermal state feature set Hset, clustering analysis is performed to establish the thermal state feature template set Htpl;

[0158] The score calculation generation module compares the real-time feature vector of the current IT device with the hot-state feature template set Htpl, calculates a heat balance index Hc, and generates a device credit score Sc in combination with a historical credit average score Sh;

[0159] The decision feedback module generates a credit level Lc according to the interval range of the device credit score Sc, and uploads the credit level Lc and the device credit score Sc to a cloud credit evaluation platform for display and update.

[0160] Although the embodiments of the present application have been shown and described, it is to be understood that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the present application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for evaluating device credit for enterprise leasing, characterized by: The method comprises the following steps: S1, collecting temperature, power consumption and heat dissipation data of the IT equipment rented by the enterprise in a collection window Win, and performing time synchronization, abnormality elimination and normalization processing to form a raw thermal state data set Hraw; S2, extracting the temperature stability index Tstb, the power consumption fluctuation spectrum index Pspc and the heat dissipation abnormal integral Qint based on the raw thermal state data set Hraw, and integrating to form a thermal state feature set Hset; then performing clustering analysis based on the thermal state feature set Hset to establish a thermal state feature template set Htpl; S3, comparing the real-time feature vector of the current IT equipment with the thermal state feature template set Htpl based on the real-time feature vector of the current IT equipment, calculating a heat balance index Hc, and combining a historical credit average score Sh to generate a device credit score Sc; S4, generating a credit level Lc according to the interval range of the device credit score Sc, and uploading the credit level Lc and the device credit score Sc to a cloud credit evaluation platform for display and updating.

2. The device credit evaluation method for enterprise-oriented leasing according to claim 1, characterized in that: The S1 comprises S11; S11, in the collection window Win, taking the IT equipment rented by the enterprise as the collection subject, and collecting real-time data through the multi-source sensor components deployed in the device mainboard, power supply circuit and heat dissipation channel; the sensor components comprise a temperature collection sensor, a power monitoring sensor and a heat dissipation detection sensor; In the real-time data collection process, the sampling frequency Fs is used as the sampling period for monitoring: The temperature collection sensor is installed at the positions of the device processor, storage chip and power module, periodically detects the local temperature signal, forms a continuous temperature time sequence with the sampling frequency Fs, and generates a temperature sequence Tseq; The power monitoring sensor is arranged between the device input power supply line and the mainboard power supply node, obtains the power consumption value varying with the sampling frequency Fs by detecting the instantaneous product of the voltage and current signals and performing time integration, and forms a power consumption change sequence Pchg; The heat dissipation detection sensor is arranged at the heat dissipation vent and heat exchange area, calculates the heat dissipation output power per unit time by collecting the airflow temperature difference and flow rate data, and obtains a heat dissipation power curve Qcur; Wherein, all sensor components sample under a unified time synchronization protocol, and record time stamps in the collected data for consistency of different signals in time sequence; after the real-time data collection is completed, the various data are aggregated and cached in time sequence to generate an original data cache set Hbuf.

3. The method of claim 2, wherein: The S1 further comprises S12; S12, uploading the original data cache set Hbuf to a cloud processor to perform a data preprocessing process, which comprises the following processing steps: S121, time alignment and clock correction: calculating the time stamp difference of each point of the temperature sequence Tseq, the power consumption change sequence Pchg and the heat dissipation power curve Qcur in the original data cache set Hbuf to obtain a time offset Δt; The sliding time window regression algorithm is used to correct all time series in synchronization, with the time deviation among the three as the objective function, so that the corrected temperature sequence Tseq, power consumption change sequence Pchg and heat dissipation power curve Qcur are consistent under the same time reference. S122, anomaly detection and missing data repair: perform anomaly detection on the time-corrected temperature sequence Tseq, power consumption change sequence Pchg and heat dissipation power curve Qcur; when the value of any sequence exceeds the preset range [Min, Max], mark the sequence as an abnormal point and eliminate it; when a short-time missing interval is detected, complete it by using the neighborhood linear interpolation algorithm; S123, normalization and structured arrangement: normalize the temperature sequence Tseq, power consumption change sequence Pchg and heat dissipation power curve Qcur after elimination and repair, map the values to the interval [0, 1], and eliminate the dimensional differences of different physical quantities, wherein the normalization is processed by using the maximum and minimum normalization method; The signal data processed by steps S121, S122 and S123 are reorganized into a structured data table in chronological order to generate the original thermal state data set Hraw.

4. The method of claim 3, wherein the method further comprises: The S2 includes S21; S21, taking the original thermal state data set Hraw as input, extract multi-dimensional thermal state feature indexes reflecting the device running stability and heat dissipation health degree, including temperature stability index Tstb, power consumption fluctuation spectrum index Pspc and heat dissipation abnormal integral Qint; the temperature stability index Tstb is obtained by sliding time window analysis on the temperature sequence Tseq, and the time window length is Wm; The power consumption fluctuation spectrum index Pspc is obtained by fast Fourier transform (FFT) on the power consumption change sequence Pchg; The heat dissipation abnormal integral Qint is obtained by calculating the heat dissipation deviation area per unit time of the heat dissipation power curve Qcur; Let the ideal heat dissipation power be Qref, then the heat dissipation abnormal integral Qint extraction formula is as follows: Qref=∫|Qcur-Qref|dt; In the formula, dt represents the integral of time t.

5. The method of claim 4, wherein: The S2 also includes S22; S22, perform numerical standardization processing on the temperature stability index Tstb, power consumption fluctuation spectrum index Pspc and heat dissipation abnormal integral Qint in the thermal state feature set Hset, and map them to the interval [0, 1]; Then, take the temperature stability index Tstb, power consumption fluctuation spectrum index Pspc and heat dissipation abnormal integral Qint as the division basis to perform similarity clustering analysis on all samples, and automatically divide the samples into three clustering categories according to the dominance of the three features: When the temperature stability index Tstb has the highest proportion in the sample, it is classified into clustering category C1; When the power consumption fluctuation spectrum index Pspc has the highest proportion in the sample, it is classified into clustering category C2; When the heat dissipation abnormal integral Qint has the highest proportion in the sample, it is classified into clustering category C3; After clustering, the number of samples in each cluster category Ck (k=1, 2, 3) and the statistical parameters of the characteristics are calculated, including the mean and the dispersion; each cluster category Ck represents a thermal operation mode, wherein: the cluster category C1 corresponds to the thermal operation mode of the temperature-stable-dominant equipment, the cluster category C2 corresponds to the thermal operation mode of the power consumption fluctuation-dominant equipment, and the cluster category C3 corresponds to the thermal operation mode of the heat dissipation deviation-dominant equipment.

6. The method of claim 5, wherein: The S2 further includes S23 and S24; S23, for each cluster category Ck, the system calculates the average value of three characteristic indicators to obtain a template center vector, including a temperature stability index center value Tstb_c, a power consumption fluctuation spectrum index center value Pspc_c, and a heat dissipation abnormal integral center value Qint_c; Based on the principle of thermal equilibrium, the acquisition formula of the template stability score Hstb is defined as: Hstb=[Tstb_c+Pspc_c+(1-Qint_c)] / 3; wherein the template stability score Hstb is used to measure the comprehensive thermal balance stability of the cluster category Ck in the three dimensions of the temperature stability index center value Tstb_c, the power consumption fluctuation spectrum index center value Pspc_c, and the heat dissipation abnormal integral center value Qint_c; S24, the template center vectors corresponding to the three cluster categories Ck and their template stability scores Hstb are collected to form a thermal state characteristic template set Htpl; at the same time, a unique template identification code Tpl_id is assigned to each template in the thermal state characteristic template set Htpl, and an index structure is established.

7. The method of claim 6, wherein the method further comprises: The S3 includes S31; S31, in the collection window Win, the original thermal state data set Hraw of the enterprise rental IT equipment currently being evaluated is taken as the input to calculate the real-time temperature stability index Tstb_r, the real-time power consumption fluctuation spectrum index Pspc_r, and the real-time heat dissipation abnormal integral Qint_r of the IT equipment; the three real-time indicators obtained are compared with the template center vectors corresponding to each template in the thermal state characteristic template set Htpl for similarity; the similarity comparison is obtained by using the Euclidean distance to calculate the characteristic deviation distance Dd; The calculation formula of the characteristic deviation distance Dd is as follows: ； The characteristic deviation distance Dd of each IT equipment and each template is calculated in turn, and the minimum value is selected as the minimum value of the characteristic deviation distance Dd_min, and the corresponding template is confirmed as the optimal matching template of the IT equipment; The real-time temperature stability index Tstb_r is the current temperature stability index of the equipment calculated based on the temperature sequence Tseq in the collection window Win; The real-time power consumption fluctuation spectrum index Pspc_r is the power consumption fluctuation spectrum index calculated based on the power consumption change sequence Pchg in the collection window Win; The real-time heat dissipation abnormal integral Qint_r is the heat dissipation abnormal integral calculated based on the heat dissipation power curve Qcur in the collection window Win.

8. The method of claim 7, wherein the method further comprises: The S3 further includes S32; S32, based on the template stability score Hstb and the minimum feature deviation distance Dd_min, calculate the heat balance index Hc of the current IT equipment Hc=Hstb×(1-Dd_min); and in a plurality of consecutive acquisition windows, calculate the time-weighted average value to form the time-weighted heat balance index Hcw; The time-weighted heat balance index Hcw is calculated according to the following formula: ； In the formula, Hc_b is the heat balance index of the bth acquisition window, and W_b represents the time weight of the bth acquisition window. Then, in combination with the historical credit average score Sh stored by the IT equipment, the device credit score Sc of the current IT equipment is calculated. The device credit score Sc is calculated according to the following formula: Sc=α×Hcw+β×Sh; In the formula, α and β represent the weight coefficients of the time-weighted heat balance index Hcw and the historical credit average score Sh respectively, and α+β=1.

9. The method of claim 8, wherein: The S4 includes S41; S41, map and convert the device credit score Sc according to the preset credit level interval to obtain the credit level Lc, encapsulate the credit level Lc obtained by the determination and the corresponding device credit score Sc to form a credit feedback data packet, and upload the credit feedback data packet to the leasing equipment credit evaluation cloud platform through a cloud communication interface; The credit level Lc is obtained by the following mapping conversion method: When the device credit score Sc is greater than or equal to 0.85, the current IT equipment is determined to be a high credit level Lc_H; When 0.65≤device credit score Sc<0.85, the current IT equipment is determined to be a medium credit level Lc_M; When the device credit score Sc is less than 0.65, the current IT equipment is determined to be a low credit level Lc_L.

10. An enterprise leasing-oriented device credit evaluation system, applied to the enterprise leasing-oriented device credit evaluation method in any one of claims 1-9, characterized in that: It comprises a thermal data acquisition module, a thermal feature extraction module, a score calculation and generation module, and a decision feedback module. The thermal data acquisition module acquires temperature, power consumption and heat dissipation data of the enterprise-rented IT equipment in the acquisition window Win, and performs time synchronization, abnormality elimination and normalization processing to form a raw thermal data set Hraw; The thermal feature extraction module extracts the temperature stability index Tstb, the power consumption fluctuation spectrum index Pspc and the heat dissipation abnormal integral Qint based on the raw thermal data set Hraw, and integrates them to form a thermal feature set Hset; then, based on the thermal feature set Hset, clustering analysis is performed to establish a thermal feature template set Htpl; The score calculation and generation module calculates the heat balance index Hc based on the real-time feature vector of the current IT equipment, compares it with the thermal feature template set Htpl, and generates the device credit score Sc in combination with the historical credit average score Sh; The decision feedback module generates the credit level Lc according to the interval range of the device credit score Sc, and uploads the credit level Lc and the device credit score Sc to the cloud credit evaluation platform for display and update.

Images