An end-cloud collaborative device carbon efficiency real-time prediction method and system

Abstract

This application provides a method and system for real-time prediction of equipment energy and carbon efficiency through end-to-cloud collaboration, relating to the field of energy and carbon efficiency prediction technology. The method includes: real-time acquisition of current, vibration, and temperature data from the equipment end, extracting electrical, mechanical, and thermodynamic energy efficiency characteristic parameters; calculation of cross-domain coupling coefficients, efficiency health indices, and characteristic parameter volatility at the equipment end, generating an energy efficiency diagnostic report and uploading it to a cloud platform; dynamic scheduling of the basic prediction model by the cloud platform, outputting predicted equipment power efficiency values; energy efficiency degradation correction of the predicted equipment power efficiency values ​​by the cloud platform, obtaining corrected predicted equipment power efficiency values; and carbon efficiency conversion calculation based on the real-time carbon emission factor of the power grid and the corrected predicted equipment power efficiency values, outputting predicted equipment energy and carbon efficiency values. This solves the technical problems of insufficient accuracy, poor real-time performance, and lack of consideration for coordinated energy and carbon management in existing technologies.

Description

Technical Field

[0001] This invention relates to the field of energy and carbon efficiency prediction, and in particular to a method and system for real-time prediction of device energy and carbon efficiency through edge-cloud collaboration. Background Technology

[0002] Improving the energy efficiency of industrial equipment and controlling carbon emissions have a key impact on the low-carbon transformation of the industrial sector, and are also important means for enterprises to reduce operating costs and enhance their sustainable development capabilities.

[0003] However, traditional equipment management suffers from problems such as a disconnect between energy efficiency monitoring and carbon emission assessment, and outdated data processing and analysis models. On the one hand, existing equipment energy efficiency monitoring focuses on data from a single physical domain, failing to form a comprehensive analysis across multiple dimensions such as electrical, mechanical, and thermodynamics, thus making it impossible to achieve an integrated assessment of energy efficiency and carbon emissions. On the other hand, the lack of efficient collaboration between the equipment and the cloud results in insufficient real-time performance and accuracy of energy and carbon efficiency assessments, making it difficult to support low-carbon operation decisions for equipment.

[0004] Therefore, there is an urgent need for a real-time prediction method for device energy and carbon efficiency through edge-cloud collaboration, in order to solve the problems of insufficient accuracy, poor real-time performance, and lack of consideration for coordinated energy and carbon management in existing technologies. Summary of the Invention

[0005] This invention addresses the technical problems of insufficient accuracy, poor real-time performance, and lack of consideration for coordinated energy and carbon efficiency management in existing technologies by providing a method and system for real-time prediction of device energy and carbon efficiency through edge-cloud collaboration.

[0006] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0007] In a first aspect, the present invention provides a method for real-time prediction of device energy and carbon efficiency through edge-cloud collaboration, comprising:

[0008] The equipment end-side collects current data, vibration data and temperature data of the equipment in real time, and extracts electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters and thermodynamic energy efficiency characteristic parameters through spectrum analysis, frequency domain analysis and time series analysis.

[0009] Based on the electrical energy efficiency characteristic parameters, the mechanical energy efficiency characteristic parameters, and the thermodynamic energy efficiency characteristic parameters, the device end calculates the cross-domain coupling coefficient, the efficiency health index, and the characteristic parameter fluctuation rate, and generates an energy efficiency diagnostic report containing anomaly type identifiers, which is then uploaded to the cloud platform.

[0010] The cloud platform receives the energy efficiency diagnostic report, dynamically schedules the basic prediction model from the preset prediction model array, inputs the electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters and thermodynamic energy efficiency characteristic parameters from the energy efficiency diagnostic report, and outputs the predicted value of the equipment's power consumption efficiency.

[0011] The cloud platform corrects the predicted power efficiency of the equipment by adjusting the energy efficiency attenuation based on the cumulative running time of the equipment and the pre-stored equipment energy efficiency attenuation benchmark curve, thus obtaining the corrected predicted power efficiency of the equipment.

[0012] The cloud platform, based on the real-time carbon emission factor of the power grid and the predicted power efficiency of the corrected equipment, outputs the predicted carbon efficiency of the equipment through carbon energy efficiency conversion calculation.

[0013] Secondly, the present invention provides a real-time prediction system for device energy and carbon efficiency through edge-cloud collaboration, comprising:

[0014] The data acquisition module is used to collect current, vibration and temperature data of the equipment in real time at the equipment end. Through spectrum analysis, frequency domain analysis and time series analysis, it extracts electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters and thermodynamic energy efficiency characteristic parameters.

[0015] An anomaly identification module is used on the device side to calculate the cross-domain coupling coefficient, efficiency health index, and characteristic parameter fluctuation rate based on the electrical energy efficiency characteristic parameters, the mechanical energy efficiency characteristic parameters, and the thermodynamic energy efficiency characteristic parameters, and to generate an energy efficiency diagnosis report containing an anomaly type identifier and upload it to the cloud platform.

[0016] The efficiency prediction module is used by the cloud platform to receive the energy efficiency diagnosis report, dynamically schedule the basic prediction model from the preset prediction model array, input the electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters and thermodynamic energy efficiency characteristic parameters in the energy efficiency diagnosis report, and output the predicted value of the equipment's power consumption efficiency.

[0017] The efficiency correction module is used by the cloud platform to correct the predicted power efficiency of the device based on the device's cumulative runtime and the pre-stored device energy efficiency attenuation baseline curve, so as to obtain the corrected predicted power efficiency of the device.

[0018] The calculation output module is used by the cloud platform to calculate and output the predicted carbon efficiency of the equipment based on the real-time carbon emission factor of the power grid and the predicted power efficiency of the corrected equipment through carbon energy efficiency conversion.

[0019] The beneficial effects of this invention are:

[0020] Compared to existing technologies, this application firstly collects real-time current, vibration, and temperature data from the equipment at the device end. Through spectrum analysis, frequency domain analysis, and time series analysis, it extracts electrical, mechanical, and thermodynamic energy efficiency characteristic parameters, achieving comprehensive capture of the equipment's operating status data across multiple physical domains (electrical, mechanical, and thermodynamic), accurately characterizing the energy efficiency loss and operating status of each domain. Secondly, based on the electrical, mechanical, and thermodynamic energy efficiency characteristic parameters, the device end calculates the cross-domain coupling coefficient, efficiency health index, and characteristic parameter volatility, generating an energy efficiency diagnostic report containing anomaly type identifiers and uploading it to the cloud platform. This achieves deep fusion analysis of the equipment's multi-domain energy efficiency characteristics (electrical, mechanical, and thermodynamic), quantifies cross-domain coupling relationships, reduces the amount of data uploaded to the cloud platform, and ensures real-time transmission. Secondly, the cloud platform receives energy efficiency diagnostic reports, dynamically schedules basic prediction models from a pre-set prediction model array, inputs electrical, mechanical, and thermodynamic energy efficiency characteristic parameters from the energy efficiency diagnostic reports, and outputs predicted equipment power efficiency values. By dynamically scheduling basic prediction models adapted to the current operating state of the equipment, it fully leverages the scenario adaptation advantages of multi-model collaboration, achieving accurate and robust prediction of equipment power efficiency. Furthermore, based on the equipment's cumulative operating time and pre-stored equipment energy efficiency degradation benchmark curves, the cloud platform corrects the predicted equipment power efficiency values ​​for energy efficiency degradation, obtaining corrected predicted equipment power efficiency values. This takes into account the natural energy efficiency degradation characteristics that occur throughout the equipment's life cycle with cumulative operating time, improving the accuracy of power efficiency prediction. Finally, based on the real-time carbon emission factor of the power grid and the corrected predicted equipment power efficiency values, the cloud platform calculates carbon efficiency conversion and outputs predicted equipment carbon efficiency values, achieving an integrated and accurate characterization of energy efficiency and carbon emissions, providing data support for low-carbon operation management of equipment.

[0021] Through the above technical solution, this application effectively addresses the pain points of traditional technologies, such as the disconnect between energy efficiency and carbon emission assessment, the limitations of single physical domain analysis, and insufficient end-to-cloud collaboration, by acquiring multi-domain data and extracting targeted features at the device end, calculating cross-domain coupling indicators, and uploading simplified diagnostic reports. Combined with the dynamic model scheduling, energy efficiency attenuation correction, and carbon energy efficiency conversion of the cloud platform, it achieves comprehensive perception of equipment operating status through multi-dimensional feature extraction and cross-domain coupling analysis. It also reduces data transmission volume and improves real-time performance by completing data preprocessing, feature extraction, and preliminary diagnosis at the device end. Furthermore, it improves the accuracy of equipment power consumption efficiency prediction by leveraging dynamic model scheduling and attenuation correction. Finally, through real-time carbon emission factor coupling conversion from the power grid, it achieves real-time, accurate, and integrated prediction of equipment energy and carbon efficiency, providing reliable data support and decision-making basis for the low-carbon operation and management of industrial equipment. Attached Figure Description

[0022] Figure 1A flowchart illustrating a method for real-time prediction of device energy and carbon efficiency through edge-cloud collaboration provided by the present invention.

[0023] Figure 2 This is a schematic diagram of the structure of a real-time prediction system for device carbon efficiency through edge-cloud collaboration, provided by the present invention.

[0024] In the attached diagram, the components represented by each number are as follows:

[0025] Data acquisition module 11, anomaly identification module 12, efficiency prediction module 13, efficiency correction module 14, calculation output module 15. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0028] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0029] Example 1, as Figure 1 As shown, this embodiment of the invention provides a method for real-time prediction of device energy and carbon efficiency through edge-cloud collaboration, including:

[0030] S10: Real-time acquisition of current, vibration, and temperature data from the equipment at the device end. Through spectrum analysis, frequency domain analysis, and time series analysis, electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, and thermodynamic energy efficiency characteristic parameters are extracted.

[0031] Traditional energy efficiency monitoring methods are mostly limited to data from a single physical domain, such as collecting only current or temperature, and lack targeted signal analysis methods. This results in incomplete extraction of energy efficiency-related features, making it difficult to cover multiple dimensions of energy efficiency influencing factors such as electrical losses, mechanical losses, and thermal losses, thereby affecting the accuracy of energy efficiency assessment and energy-carbon efficiency prediction.

[0032] Meanwhile, during equipment operation, current data, vibration data, and temperature data are directly related to the electrical operating status, mechanical transmission status, and thermodynamic heat dissipation status, respectively. These are core data reflecting the comprehensive energy efficiency of the equipment. Through targeted signal processing methods such as spectrum analysis, frequency domain analysis, and time series analysis, key characteristic parameters representing the energy efficiency status of different domains can be extracted from the above three types of data, providing a reliable data foundation for multi-domain coupling analysis and accurate energy and carbon efficiency prediction.

[0033] To address the aforementioned issues, this application collects real-time current, vibration, and temperature data from the equipment at the device end. Through spectrum analysis, frequency domain analysis, and time series analysis, it extracts electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, and thermodynamic energy efficiency characteristic parameters.

[0034] Specifically, step S10 in the method includes:

[0035] The device end side collects the device's current data through a current transformer;

[0036] The device end side uses a triaxial vibration acceleration sensor to collect vibration data of the device;

[0037] The device end-side uses temperature sensors to collect the winding temperature and bearing temperature of the device as temperature data.

[0038] The device performs a fast Fourier transform on the collected current data to calculate the total harmonic distortion rate and the amplitude of each harmonic current, which are used as electrical energy efficiency characteristic parameters.

[0039] The device end performs spectral analysis on the collected vibration data and extracts the characteristic frequency amplitudes related to the mechanical faults of the device as mechanical energy efficiency characteristic parameters.

[0040] The device performs moving average filtering and differentiation on the collected temperature data to calculate the temperature rise rate of the device, which is used as a thermodynamic energy efficiency characteristic parameter.

[0041] In this embodiment, the current data of the equipment is first collected at the equipment end via a current transformer. A current transformer is an instrument used to transform current, characterized by high accuracy and strong anti-interference capability, suitable for real-time acquisition of the operating current of industrial equipment. Specifically, the current transformer can be connected in series in the equipment's power supply circuit at the equipment end to ensure that the collected current data accurately reflects the equipment's power load status.

[0042] Secondly, a triaxial vibration accelerometer is used at the equipment end to collect vibration data. This triaxial accelerometer can simultaneously collect vibration acceleration signals in the X, Y, and Z directions, comprehensively reflecting the mechanical vibration state of the equipment and avoiding the omission of fault characteristics caused by collecting data in only one direction. Specifically, the triaxial vibration accelerometer can be installed on critical vibration parts of the equipment, such as motor bearing end covers and pump housings, and secured with bolts to ensure a tight fit with the equipment surface, minimizing signal attenuation.

[0043] Secondly, temperature sensors at the equipment end collect the winding and bearing temperatures as temperature data. These temperatures are key indicators reflecting the thermodynamic state of the equipment. Excessively high winding temperatures accelerate insulation aging, while excessively high bearing temperatures exacerbate wear; both negatively impact equipment energy efficiency. For example, the temperature sensors should employ a contact measurement method, embedding the winding temperature sensor inside the motor stator winding and mounting the bearing temperature sensor on the bearing housing surface to collect the winding and bearing temperatures as temperature data.

[0044] Furthermore, since current harmonics increase electrical losses and reduce power efficiency during equipment operation, the equipment side performs a Fast Fourier Transform (FFT) on the collected current data to calculate the total harmonic distortion (THD) and the amplitude of each harmonic current, which serve as electrical energy efficiency characteristic parameters. The FFT is an algorithm that converts time-domain current signals into frequency-domain signals, decomposing them into the fundamental current and each harmonic current. The THD is an indicator of the degree of current waveform distortion, calculated as the square root of the sum of the squares of the effective values ​​of each harmonic current and the effective value of the fundamental current, expressed as a percentage. The amplitude of each harmonic current refers to the peak value of the corresponding harmonic current.

[0045] For example, the formula for calculating the total harmonic distortion rate is: ,in, Total harmonic distortion (THD) This is the effective value of the fundamental current. ~In represents the effective value of the 2nd to nth harmonic current.

[0046] For example, performing a Fast Fourier Transform on the collected 1-second current data yields an RMS value of 450A for the fundamental current, 15A for the 3rd harmonic current, and 10A for the 5th harmonic current. The total harmonic distortion rate is then: ≈4.01%.

[0047] Furthermore, the equipment side performs spectral analysis on the collected vibration data to extract characteristic frequency amplitudes related to mechanical faults, which serve as mechanical energy efficiency characteristic parameters. Specifically, spectral analysis uses signal processing algorithms such as Fast Fourier Transform to convert the time-domain vibration acceleration signal into a power spectral density signal in the frequency domain. This signal quantifies the energy distribution of different frequency components. By identifying the frequencies corresponding to the peak values ​​in the power spectrum, characteristic frequencies are obtained, which can then be used to locate the frequency components related to mechanical faults, providing a reliable basis for fault diagnosis. The characteristic frequency amplitude refers to the vibration acceleration amplitude corresponding to that characteristic frequency, and its magnitude is positively correlated with the severity of the mechanical fault: the larger the characteristic frequency amplitude, the more severe the corresponding mechanical fault, and the greater the mechanical losses generated during equipment operation.

[0048] Specifically, the vibration signal spectrum generated by equipment operation contains a large number of frequency components, most of which are background noise from normal equipment operation, such as inherent structural vibrations and environmental interference signals. Only a few specific frequency components are directly related to mechanical faults; these specific frequencies are called characteristic frequencies. Characteristic frequencies are jointly determined by the structural physical characteristics of the core mechanical components of the equipment and the fault occurrence mechanism. For example, the characteristic frequency of bearing faults can be derived by combining design parameters such as the number of bearing rolling elements, pitch circle diameter, and rolling element diameter with classical mechanical fault diagnosis theory. The characteristic frequency of rotor imbalance or misalignment faults is usually the fundamental frequency or its integer multiples corresponding to the rotor speed. The characteristic frequency of gear faults is related to the meshing frequency corresponding to the product of the number of gear teeth and the rotational speed, as well as its sidebands.

[0049] For example, the fault characteristic frequency can be obtained through two methods: theoretical calculation or experimental calibration. On the one hand, the characteristic frequency of typical faults can be directly calculated based on the design drawings and nameplate parameters of the core mechanical components of the equipment, combined with the classic theoretical formulas in the field of mechanical fault diagnosis. On the other hand, common mechanical faults such as bearing wear, rotor imbalance, and abnormal gear meshing can be artificially simulated before the equipment leaves the factory or in a laboratory environment. Vibration data under fault conditions can be collected and spectrum analysis can be performed to identify the peak value of the characteristic frequency corresponding to the fault. At the same time, the fluctuation range of the frequency under actual working conditions can be recorded. Finally, the frequency range can be preset as the fault characteristic frequency screening range on the equipment side to ensure that the equipment side can accurately locate the fault-related frequency components during real-time analysis and avoid background noise interference.

[0050] For example, a motor bearing rotates at 1500 rpm. Spectrum analysis reveals a significant peak at 160 Hz. The corresponding frequency is taken as the characteristic frequency. This characteristic frequency is compared with the fault characteristic frequency screening range of different faults. It is found to be consistent with the characteristic frequency of the bearing outer ring fault. The amplitude of the corresponding characteristic frequency is 2.5g, which is taken as the mechanical energy efficiency characteristic parameter.

[0051] Finally, since the temperature data may contain environmental noise, the device performs moving average filtering and differentiation processing on the collected temperature data to calculate the device's temperature rise rate, which serves as a thermodynamic efficiency characteristic parameter. Differentiation processing involves calculating the ratio of the temperature change between two adjacent sampling times to the time interval, i.e., the temperature rise rate, which reflects how fast the device's temperature rises and can be used to detect thermal anomalies. Moving average filtering effectively smooths noise and preserves the temperature change trend; the filter window size can be adjusted according to the sampling frequency. The formula for moving average filtering is: T_filtered(k) = [T(k) + T(k-1)+…+T(k-m+1)] / m, where T_filtered(k) is the filtered temperature value, m is the filter window size, and T(k) is the original temperature value at time k. The formula for the temperature rise rate is: v = (T_filtered(k)-T_filtered(k-1)) / Δt, where Δt is the sampling time interval.

[0052] For example, if the temperature acquisition interval is 0.1 seconds and the filter window size is 5, the temperature at time k after filtering is 85℃ and the temperature at time k-1 is 84.8℃. Then, using the formula for calculating the temperature rise rate, the temperature rise rate is calculated as (85-84.8) / 0.1=2℃ / s.

[0053] In summary, compared to existing technologies, this application acquires real-time current, vibration, and temperature data from the equipment at the device end. Through spectrum analysis, frequency domain analysis, and time series analysis, it extracts electrical, mechanical, and thermodynamic energy efficiency characteristic parameters. This achieves comprehensive capture of operational status data across multiple physical domains (electrical, mechanical, and thermodynamic). The multi-domain energy efficiency characteristic parameters extracted using targeted signal analysis methods accurately characterize the energy loss and operational status of each domain, providing a reliable data foundation for subsequent cross-domain coupling analysis, health status assessment, and accurate prediction of energy and carbon efficiency.

[0054] S20: The device end calculates the cross-domain coupling coefficient, efficiency health index and characteristic parameter fluctuation rate based on the electrical energy efficiency characteristic parameters, the mechanical energy efficiency characteristic parameters and the thermodynamic energy efficiency characteristic parameters, and generates an energy efficiency diagnosis report containing anomaly type identifiers and uploads it to the cloud platform.

[0055] Traditional equipment energy efficiency monitoring often involves isolated analysis of characteristic parameters in a single physical domain, failing to quantify the interactions between electrical, mechanical, and thermodynamic domains, and lacking a systematic assessment of the overall energy efficiency and health status of the equipment.

[0056] To address the aforementioned issues, the device side of this application calculates the cross-domain coupling coefficient, efficiency health index, and characteristic parameter volatility based on the electrical energy efficiency characteristic parameters, the mechanical energy efficiency characteristic parameters, and the thermodynamic energy efficiency characteristic parameters, and generates an energy efficiency diagnostic report containing anomaly type identifiers, which is then uploaded to the cloud platform.

[0057] Specifically, step S20 in the method includes:

[0058] The end-side obtains the electromagnetic-mechanical coupling coefficient by calculating the coherence function value between a specific harmonic current amplitude and the corresponding characteristic frequency vibration amplitude, based on the amplitude of each harmonic current in the electrical energy efficiency characteristic parameters and the amplitude of the vibration characteristic frequency in the mechanical energy efficiency characteristic parameters.

[0059] The end-side obtains the electro-thermal coupling coefficient by calculating the correlation coefficient between harmonic loss power and temperature rise rate based on the total harmonic distortion rate in the electrical energy efficiency characteristic parameters and the temperature rise rate in the thermodynamic energy efficiency characteristic parameters.

[0060] The end-side calculates the harmonic loss power by summing the squares of the harmonic current amplitudes and the equivalent resistance of the equipment based on the electrical energy efficiency characteristic parameters.

[0061] The end side calculates the vibration loss power through energy mapping based on the vibration characteristic frequency amplitude and vibration energy conversion coefficient in the mechanical energy efficiency characteristic parameters.

[0062] The end side calculates the thermal gradient anomaly value by using the difference between the measured temperature rise rate and the theoretical temperature rise rate in the thermodynamic energy efficiency characteristic parameters.

[0063] The ratio of the harmonic loss power to the current total active power is used to obtain the harmonic loss ratio; the ratio of the vibration loss power to the current total active power is used to obtain the vibration loss ratio; and the ratio of the thermal gradient anomaly value to the equipment's rated temperature rise rate is used to obtain the thermal anomaly coefficient.

[0064] The harmonic loss ratio, vibration loss ratio and thermal anomaly coefficient are combined to form an energy efficiency and health feature vector.

[0065] The energy efficiency health feature vector is input into a pre-trained health assessment model to calculate the efficiency health index.

[0066] The end-side calculates the characteristic parameter volatility based on the historical data of electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, and thermodynamic energy efficiency characteristic parameters within the sliding time window, using the standard deviation.

[0067] In this embodiment, the electromagnetic-mechanical coupling coefficient is first obtained by calculating the coherence function value between the amplitude of each harmonic current in the electrical energy efficiency characteristic parameters and the amplitude of the vibration characteristic frequency in the mechanical energy efficiency characteristic parameters. This is because during equipment operation, harmonic currents in the electrical domain generate electromagnetic forces, which in turn induce vibrations in the mechanical domain, and the two are coupled. The specific harmonic current refers to the harmonic order corresponding to the mechanical characteristic frequency of the equipment. For example, the 3rd harmonic corresponds to the bearing fault characteristic frequency. The electromagnetic-mechanical coupling coefficient is obtained by calculating the coherence function value between the amplitude of this harmonic current and the amplitude of the corresponding vibration characteristic frequency.

[0068] For example, the formula for calculating the coherence function is: γ²(f) = |G_xy(f)|² / (G_xx(f) × G_yy(f)), where γ²(f) is the coherence function value, G_xy(f) is the cross-power spectral density of the harmonic current signal and the vibration signal, G_xx(f) is the auto-power spectral density of the harmonic current signal, and G_yy(f) is the auto-power spectral density of the vibration signal. The coherence function value is used to quantify the degree of linear correlation between the two signals, and its value ranges from 0 to 1. The closer the coherence function value is to 1, the stronger the coupling relationship between the harmonic current and the vibration.

[0069] For example, using the formula for calculating the coherence function, the coherence function value of the third harmonic current amplitude and the 160Hz vibration characteristic frequency amplitude is calculated to be 0.85, and 0.85 is used as the electromagnetic-mechanical coupling coefficient.

[0070] Secondly, based on the total harmonic distortion (THD) parameter in the electrical energy efficiency characteristic parameters and the temperature rise rate in the thermodynamic energy efficiency characteristic parameters, the correlation coefficient between harmonic power loss and the temperature rise rate is calculated to obtain the electro-thermal coupling coefficient. Specifically, a higher THD corresponds to greater harmonic power loss, leading to increased equipment heating and a faster temperature rise rate; the two are positively correlated. The correlation coefficient can be calculated using the Pearson correlation coefficient method to perform linear correlation analysis between harmonic power loss and the temperature rise rate. This correlation coefficient is used as the electro-thermal coupling coefficient to quantify the degree of linear correlation between harmonic power loss and the temperature rise rate, with a value ranging from -1 to 1. A positive value indicates a positive correlation, and a larger absolute value indicates a stronger correlation.

[0071] For example, the harmonic loss power of a certain device over a period of time is 5kW, 6kW, 7kW, and 8kW, respectively, and the corresponding temperature rise rates are 1.5℃ / s, 1.8℃ / s, 2.1℃ / s, and 2.4℃ / s, respectively. The Pearson correlation coefficient method is used to perform linear correlation analysis on the two sets of data, and the correlation coefficient is 1.0. 1.0 is used as the electro-thermal coupling coefficient.

[0072] Secondly, at the end side, based on the amplitude of each harmonic current and the equivalent resistance of the equipment in the electrical energy efficiency characteristic parameters, the harmonic loss power is calculated by summing the squares. Specifically, the equivalent resistance of the equipment is equivalent to a single lumped resistance in the electrical circuit of the equipment, which can be obtained through equipment nameplate parameters or experimental measurement. Harmonic loss power refers to the additional power loss of the equipment due to current harmonics, mainly manifested as resistive loss. The power loss generated by each harmonic current across the equivalent resistance is the product of the square of the effective value of the harmonic current and the equivalent resistance, and the total harmonic loss power is the sum of the power losses of each harmonic.

[0073] For example, the formula for calculating harmonic power loss is: P_h=Σ(I_n²×R_eq), where I_n is the amplitude of the nth harmonic current and R_eq is the equivalent resistance of the equipment. For example, if the equivalent resistance of the equipment is 0.5Ω, the amplitude of the 3rd harmonic current is 15A, and the amplitude of the 5th harmonic current is 10A, then the power loss of the 3rd harmonic is 15²×0.5=112.5W, the power loss of the 5th harmonic is 10²×0.5=50W, and the total harmonic power loss is 112.5+50=162.5W.

[0074] Furthermore, the vibration loss power is calculated at the end-side based on the vibration characteristic frequency amplitude and vibration energy conversion coefficient in the mechanical energy efficiency characteristic parameters through energy mapping. Specifically, energy loss occurs during equipment vibration; the larger the vibration characteristic frequency amplitude, the greater the vibration energy and the higher the power loss. The vibration energy conversion coefficient is a proportionality coefficient that converts vibration acceleration amplitude into vibration power. It can be calibrated experimentally, such as by applying vibration of known power to a standard vibration table, measuring the corresponding vibration acceleration amplitude, and fitting the result to obtain the vibration energy conversion coefficient.

[0075] For example, the vibration loss power can be calculated by multiplying the vibration characteristic frequency amplitude by the vibration energy conversion coefficient. For instance, if the vibration characteristic frequency amplitude is 2.5g and the vibration energy conversion coefficient is calibrated to be 10W / g, then the vibration loss power is 2.5×10=25W.

[0076] Furthermore, the thermal gradient anomaly value is calculated by subtracting the measured and theoretical temperature rise rates from the thermodynamic efficiency characteristic parameters at the end-point. The measured temperature rise rate is the actual measured value after filtering, while the theoretical temperature rise rate refers to the temperature rise rate calculated based on the heat dissipation design parameters under normal operating conditions. This rate can be determined through factory test data or simulation calculations. When thermal anomalies exist, such as cooling fan failure or blocked heat dissipation channels, the measured temperature rise rate will be higher than the theoretical temperature rise rate. The thermal gradient anomaly value is the difference between the measured and theoretical temperature rise rates; a larger difference indicates a more severe thermal anomaly.

[0077] For example, if the theoretical temperature rise rate of the equipment under the current load is 1℃ / s obtained from the equipment factory test data, and the actual temperature rise rate is 2℃ / s, then the thermal gradient anomaly value is 2-1=1℃ / s.

[0078] Furthermore, the ratio of harmonic loss power to the current total active power yields the harmonic loss percentage; the ratio of vibration loss power to the current total active power yields the vibration loss percentage; and the ratio of the thermal gradient anomaly value to the equipment's rated temperature rise rate yields the thermal anomaly coefficient. Specifically, the total active power is the total input power of the equipment during current operation, which can be collected by power sensors or calculated based on voltage and current data. The equipment's rated temperature rise rate is the allowable temperature increase rate under rated load, indicated on the equipment nameplate. The harmonic loss percentage and vibration loss percentage reflect the proportions of electrical and mechanical losses in the total power, respectively; a higher percentage indicates lower energy efficiency. The thermal anomaly coefficient is a normalized thermal anomaly index, ranging from 0 to ∞; a value exceeding 1 indicates the presence of a thermal anomaly.

[0079] For example, if the current total active power of the equipment is 100kW, the rated temperature rise rate of the equipment is 3℃ / s, the harmonic loss power is 0.1625kW, the vibration loss power is 0.025kW, and the thermal gradient anomaly value is 1℃ / s, then the proportion of harmonic loss is 0.1625 / 100×100%=0.1625%, the proportion of vibration loss is 0.025 / 100×100%=0.025%, and the thermal anomaly coefficient is 1 / 3≈0.333.

[0080] Furthermore, the proportion of harmonic losses, the proportion of vibration losses, and the thermal anomaly coefficient are combined to form an energy efficiency health feature vector. Specifically, the energy efficiency health feature vector integrates a multi-dimensional vector of electrical, mechanical, and thermodynamic losses and abnormal states of the equipment, which can comprehensively characterize the energy efficiency health status of the equipment. The energy efficiency health feature vector has three dimensions, and each element corresponds to the proportion of harmonic losses, the proportion of vibration losses, and the thermal anomaly coefficient, respectively. For example, based on the above calculation results, the energy efficiency health feature vector can be represented as [0.1625%, 0.025%, 0.333].

[0081] Furthermore, the energy efficiency health feature vector is input into the pre-trained health assessment model to calculate the efficiency health index. The efficiency health index ranges from 0 to 1, with values ​​closer to 1 indicating better energy efficiency and health status of the equipment.

[0082] For example, the health assessment model can be constructed based on the following technical path: 1. Model Construction: A random forest algorithm adapted to the nonlinear mapping requirements of multi-dimensional feature vectors can be used. It mainly consists of 100 CART decision trees. The input layer is a 3-dimensional energy efficiency health feature vector. Feature splitting is performed using Gini coefficients. Two features are randomly selected from each tree for splitting. The depth of the decision tree is limited to 8 layers to avoid overfitting. The output layer is a normalized efficiency health index with a value of 0-1. The prediction results of each decision tree are fused through a voting method to obtain the final output. 2. Data Preparation: Collect historical energy efficiency health feature vectors of the device under different health states, such as normal, slightly abnormal, moderately abnormal, and severely abnormal states, to form a sample feature dataset. Simultaneously, obtain the historical efficiency health index calibrated with measured energy efficiency data under the corresponding states as supervision annotations to form a sample label dataset. The sample feature dataset and the sample label dataset are randomly divided in a ratio of 7:1.5:1.5 as the training set, validation set, and test set, respectively. 3. Model Training: The historical energy efficiency and health feature vectors in the training set are used as input features, and the corresponding historical efficiency and health indices are used as supervision labels. The stochastic gradient descent method is used to minimize the mean square error between the predicted value and the label value. When the mean square error of the validation set does not decrease for 5 consecutive iterations and is lower than 0.001, the model is considered to have converged. Training is stopped and the parameters are saved to obtain the trained health assessment model.

[0083] For example, by inputting the energy efficiency health feature vector [0.1625%, 0.025%, 0.333] into a pre-trained health assessment model, the calculated efficiency health index is 0.92.

[0084] Finally, the end-side calculation of the characteristic parameter volatility is based on historical data of electrical, mechanical, and thermodynamic energy efficiency parameters within a sliding time window, using standard deviation. Specifically, the sliding time window selects historical characteristic parameter data over a recent period. The window size can be dynamically set according to actual needs, such as 1 minute or 5 minutes, to ensure that short-term trends in the characteristic parameters are reflected. Standard deviation measures the dispersion of the data; a larger standard deviation indicates more drastic changes in the characteristic parameters and poorer stability, potentially indicating abnormal equipment conditions. The characteristic parameter volatility is the average of the standard deviations of each characteristic parameter within the sliding time window, comprehensively reflecting the overall stability of the three types of characteristic parameters.

[0085] For example, if the sliding time window size is 1 minute, the standard deviation of the historical data of electrical energy efficiency characteristic parameters is 0.2, the standard deviation of the historical data of mechanical energy efficiency characteristic parameters is 0.1, and the standard deviation of the historical data of thermodynamic energy efficiency characteristic parameters is 0.05, then the fluctuation rate of the characteristic parameters is (0.2+0.1+0.05) / 3=0.1167.

[0086] Furthermore, the step of "generating an energy efficiency diagnostic report containing an anomaly type identifier and uploading it to the cloud platform" includes:

[0087] The electromagnetic-mechanical coupling coefficient is compared with a first preset threshold. When the electromagnetic-mechanical coupling coefficient exceeds the first preset threshold, a vibration anomaly type identifier is generated.

[0088] The electro-thermal coupling coefficient is compared with a second preset threshold. When the electro-thermal coupling coefficient exceeds the second preset threshold, a thermal anomaly type identifier is generated.

[0089] The terminal integrates the electrical energy efficiency characteristic parameters, the mechanical energy efficiency characteristic parameters, the thermodynamic energy efficiency characteristic parameters, the electromagnetic-mechanical coupling coefficient, the electro-thermal coupling coefficient, the efficiency health index, the characteristic parameter volatility, and the corresponding abnormality type identifier to form an energy efficiency diagnostic report, which is then uploaded to the cloud platform.

[0090] In this embodiment, the electromagnetic-mechanical coupling coefficient is first compared with a first preset threshold. When the electromagnetic-mechanical coupling coefficient exceeds the first preset threshold, a vibration anomaly type identifier is generated. Specifically, the first preset threshold is a critical value for judging whether the electromagnetic-mechanical coupling relationship is abnormal. It can be determined based on statistical data of the electromagnetic-mechanical coupling coefficient during normal operation of the equipment, for example, taking 1.2 times the maximum value during normal operation. When the electromagnetic-mechanical coupling coefficient exceeds the first preset threshold, it indicates that the coupling relationship between harmonic current and vibration is too strong, and the equipment has vibration anomalies, such as rotor imbalance or bearing failure. At this time, a vibration anomaly type identifier is generated; if it does not exceed the threshold, no identifier is generated.

[0091] For example, if the maximum value of the electromagnetic-mechanical coupling coefficient of a motor is 0.7 when it is running normally, the first preset threshold can be set to 0.7×1.2=0.84. If the current electromagnetic-mechanical coupling coefficient is 0.85, which exceeds the first preset threshold, then a vibration abnormality type identifier is generated.

[0092] Secondly, the electro-thermal coupling coefficient is compared with a second preset threshold. When the electro-thermal coupling coefficient exceeds the second preset threshold, a thermal anomaly type identifier is generated. Specifically, the second preset threshold is a critical value for judging whether the electro-thermal coupling relationship is abnormal. It can be determined based on statistical data of the electro-thermal coupling coefficient during normal operation of the equipment, for example, taking 1.2 times the maximum value of the electro-thermal coupling coefficient during normal operation. When the electro-thermal coupling coefficient exceeds the second preset threshold, it indicates that the coupling relationship between harmonic loss and temperature rise is too strong, and the equipment has a thermal anomaly, such as poor heat dissipation or winding short circuit. In this case, a thermal anomaly type identifier is generated; if it does not exceed the threshold, no identifier is generated.

[0093] For example, if the maximum value of the electro-thermal coupling coefficient of a device is 0.9 when it is running normally, the second preset threshold can be set to 0.9×1.2=1.08. If the current electro-thermal coupling coefficient is 1.0, which does not exceed the second preset threshold, then no thermal anomaly type identifier will be generated.

[0094] Finally, the device-side system integrates electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, thermodynamic energy efficiency characteristic parameters, electromagnetic-mechanical coupling coefficient, electro-thermal coupling coefficient, efficiency health index, characteristic parameter volatility, and corresponding anomaly type identifiers to form an energy efficiency diagnostic report, which is then uploaded to the cloud platform. Specifically, the energy efficiency diagnostic report is a centralized presentation of the data processing results from the device-side, containing core information about the device's operating status and providing comprehensive input for cloud platform model scheduling and prediction. For example, the energy efficiency diagnostic report can use a structured format, such as JSON, to ensure that the cloud platform can parse it quickly.

[0095] In summary, compared to existing technologies, the device-side implementation of this application calculates the cross-domain coupling coefficient, efficiency health index, and characteristic parameter volatility based on the electrical, mechanical, and thermodynamic energy efficiency characteristic parameters, and generates an energy efficiency diagnostic report containing anomaly type identifiers, which is then uploaded to the cloud platform. This achieves deep fusion analysis of the electrical, mechanical, and thermodynamic energy efficiency characteristics of the equipment, quantifies cross-domain coupling relationships, reduces the amount of data uploaded to the cloud platform, ensures real-time transmission, and provides highly targeted and reliable pre-diagnostic basis for cloud-based dynamic model scheduling and accurate prediction of power consumption efficiency.

[0096] S30: The cloud platform receives the energy efficiency diagnosis report, dynamically schedules the basic prediction model from the preset prediction model array, inputs the electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters and thermodynamic energy efficiency characteristic parameters from the energy efficiency diagnosis report, and outputs the predicted value of the equipment's power consumption efficiency.

[0097] Traditional equipment power efficiency prediction often relies on a single model, which is difficult to adapt to the complex nonlinear mapping requirements of different operating conditions and multi-domain energy efficiency characteristics of equipment. This results in low prediction accuracy and insufficient robustness. Furthermore, the dynamic changes in equipment operating status and the coupling characteristics of electrical, mechanical, and thermodynamic domain features further exacerbate the limitations of single models. Therefore, cloud platforms need to leverage the advantages of multi-model collaboration by dynamically scheduling a pre-set array of differentiated prediction models based on the energy efficiency diagnostic reports uploaded by the equipment, thereby overcoming the bottleneck of single model scenario adaptation and achieving accurate prediction of equipment power efficiency.

[0098] To address the aforementioned issues, the cloud platform described in this application receives the energy efficiency diagnostic report, dynamically schedules the basic prediction model from a pre-set prediction model array, inputs the electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, and thermodynamic energy efficiency characteristic parameters from the energy efficiency diagnostic report, and outputs the predicted value of the equipment's power consumption efficiency.

[0099] Specifically, step S30 in the method includes:

[0100] The cloud platform calculates the model scheduling coefficient based on the efficiency health index, cross-domain coupling coefficient, and characteristic parameter volatility in the energy efficiency diagnostic report through weighted fusion.

[0101] Multiply the model scheduling coefficient by the total number of models in the prediction model array and round down to determine the number of basic prediction models that need to be scheduled.

[0102] A corresponding number of basic prediction models are scheduled from the prediction model array to form a basic prediction model combination;

[0103] The electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters and thermodynamic energy efficiency characteristic parameters are input into each of the basic prediction models in the basic prediction model combination to obtain multiple preliminary prediction results;

[0104] Based on the anomaly type identifier, electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters and thermodynamic energy efficiency characteristic parameters in the energy efficiency diagnosis report, weight coefficients are assigned to each basic prediction model in the basic prediction model combination.

[0105] The cloud platform performs weighted fusion of multiple preliminary prediction results based on the weighting coefficients to obtain the predicted value of equipment power efficiency.

[0106] In this embodiment, the cloud platform first calculates the model scheduling coefficient based on the efficiency health index, cross-domain coupling coefficient, and characteristic parameter volatility from the energy efficiency diagnostic report through weighted fusion. The cross-domain coupling coefficient refers to the average of the electromagnetic-mechanical coupling coefficient and the electro-thermal coupling coefficient. Specifically, the model scheduling coefficient is used to determine the number of basic prediction models that need to be scheduled. During weighted fusion, different weight coefficients need to be assigned to the efficiency health index, cross-domain coupling coefficient, and characteristic parameter volatility. For example, the efficiency health index has a weight of 0.4, the cross-domain coupling coefficient has a weight of 0.3, and the characteristic parameter volatility has a weight of 0.3. The weight coefficients can be dynamically determined based on the actual application scenario, and the sum of the weights must be 1.

[0107] For example, if the weighting coefficients of the efficiency health index, cross-domain coupling coefficient, and characteristic parameter volatility are 0.4, 0.3, and 0.3 respectively, the efficiency health index is 0.92, the electromagnetic-mechanical coupling coefficient is 0.85, the electro-thermal coupling coefficient is 1.0, and the characteristic parameter volatility is 0.1167, then the cross-domain coupling coefficient is (0.85+1.0) / 2=0.925, and the model scheduling coefficient is 0.4×0.92+0.3×0.925+0.3×0.1167=0.6805.

[0108] Secondly, the number of basic prediction models to be scheduled is determined by multiplying the model scheduling coefficient by the total number of models in the prediction model array and rounding down. Specifically, the total number of basic prediction models in the prediction model array is a preset fixed value, such as 10 or 20, which can be dynamically determined according to the complexity of the equipment operating conditions and the required prediction accuracy. The product of the model scheduling coefficient and the total number of models in the prediction model array reflects the number of basic prediction models adapted to the current equipment state. For example, if the total number of models in the prediction model array is 10 and the model scheduling coefficient is 0.6805, then the product is 0.6805 × 10 = 6.805, which, after rounding down, yields 7 basic prediction models to be scheduled.

[0109] Next, a corresponding number of basic prediction models are scheduled from the prediction model array to form a basic prediction model portfolio. Specifically, the cloud platform randomly selects a number of basic prediction models from the prediction model array. For example, 7 models are randomly selected from 10 basic prediction models in the prediction model array to form a basic prediction model portfolio.

[0110] Furthermore, electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, and thermodynamic energy efficiency characteristic parameters are input into various basic prediction models in the basic prediction model combination to obtain multiple preliminary prediction results. Specifically, each basic prediction model takes electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, and thermodynamic energy efficiency characteristic parameters as inputs, performs calculations through its own model structure, and outputs the corresponding predicted value of equipment power consumption efficiency as a preliminary prediction result. For example, when the electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, and thermodynamic energy efficiency characteristic parameters are input into seven basic prediction models, the seven preliminary prediction results output are 92%, 91.5%, 92.3%, 91.8%, 92.1%, 91.7%, and 92.2%, respectively.

[0111] Furthermore, based on the anomaly type identifiers, electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, and thermodynamic energy efficiency characteristic parameters in the energy efficiency diagnostic report, weight coefficients are assigned to each basic prediction model in the basic prediction model combination. Specifically, the weight coefficients are used to measure the reliability of the preliminary prediction results of each basic prediction model; the higher the reliability, the larger the weight coefficient, and the weight coefficients satisfy the condition that the sum of the weights of all models is 1.

[0112] Finally, the cloud platform performs weighted fusion of multiple preliminary prediction results based on weighting coefficients to obtain the predicted value of equipment power efficiency. Specifically, weighted fusion involves multiplying each preliminary prediction result by its corresponding weighting coefficient, and then summing all the products to obtain the final predicted value of equipment power efficiency. In this way, the predictive advantages of multiple basic prediction models can be combined, the prediction bias of a single model can be reduced, and the stability and accuracy of the prediction results can be improved.

[0113] For example, if the weight coefficients of the seven basic prediction models are 0.15, 0.13, 0.16, 0.14, 0.15, 0.13, and 0.14, respectively, and the preliminary prediction results are 92%, 91.5%, 92.3%, 91.8%, 92.1%, 91.7%, and 92.2%, respectively, then the predicted value of equipment power efficiency is approximately 92.0% = 0.15×92% + 0.13×91.5% + 0.16×92.3% + 0.14×91.8% + 0.15×92.1% + 0.13×91.7% + 0.14×92.2%.

[0114] Furthermore, the phrase "assigning weight coefficients to each basic prediction model in the basic prediction model combination based on the anomaly type identifier, electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, and thermodynamic energy efficiency characteristic parameters in the energy efficiency diagnostic report" includes:

[0115] Based on the anomaly type identifier, the cloud platform retrieves recent historical data similar to the current anomaly type from the historical database;

[0116] The cloud platform calculates the prediction accuracy of each basic prediction model on the recent historical data;

[0117] The cloud platform is based on the entropy weight method, which calculates the corresponding weight coefficients according to the prediction accuracy of each basic prediction model. The weight coefficients satisfy the condition that the sum of the weight coefficients of each basic prediction model is 1.

[0118] In this embodiment, the cloud platform first retrieves recent historical data similar to the current anomaly type from the historical database based on the anomaly type identifier. Specifically, the historical database stores all energy efficiency diagnostic reports generated during the equipment's past operation and the corresponding measured equipment power efficiency values, with the data categorized and stored according to the anomaly type identifier. The time range for recent historical data can be set to the last 3 months or the last 6 months to ensure data timeliness and avoid significant differences between historical data and the current state due to factors such as equipment aging. For example, if the current anomaly type identifier is a vibration anomaly type identifier, then the historical database is retrieved for all historical data containing the vibration anomaly type identifier for the last 3 months, including the corresponding electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, thermodynamic energy efficiency characteristic parameters, historical preliminary prediction results, and measured equipment power efficiency values.

[0119] Secondly, the cloud platform calculates the prediction accuracy of each basic prediction model on recent historical data. Specifically, prediction accuracy is an indicator that measures the predictive performance of the basic prediction model on historical data. It is calculated as follows: Prediction accuracy = (Number of samples where the error between the preliminary prediction result and the measured power efficiency value of the equipment is within the allowable range / Total number of samples) × 100%. The allowable error range can be set according to the prediction accuracy requirements, such as ±1%.

[0120] For example, for each basic prediction model, the preliminary prediction results from recent historical data are compared with the corresponding measured power efficiency values ​​of the equipment. The number of samples with errors within the allowable range is counted, and then the prediction accuracy is calculated. For instance, if a basic prediction model has prediction errors within the allowable range for 95 samples out of 100 recent historical data points, then the prediction accuracy of the basic prediction model is 95 / 100 = 95%.

[0121] Finally, the cloud platform calculates the corresponding weight coefficients based on the prediction accuracy of each basic prediction model using the entropy weight method. The sum of the weight coefficients of all basic prediction models is equal to 1. Specifically, the entropy weight method is an objective weighting method that assigns weights based on the dispersion of the prediction accuracy of each basic prediction model. The greater the difference in prediction accuracy, the more the weight allocation reflects the advantage of the superior model. The calculation steps are as follows: First, the prediction accuracy of each basic prediction model is normalized to obtain a normalized index value; second, the information entropy of each basic prediction model is calculated; finally, the entropy weight is calculated based on the information entropy to obtain the weight coefficients. The formula for calculating information entropy is: E_i = -k × Σ(p_ij × lnp_ij), and the formula for calculating entropy weight is: ω_i = (1 - E_i) / (m - ΣE_i), where k is a constant, k = 1 / lnm, m is the number of models, and p_ij is the normalized prediction accuracy of the i-th basic prediction model in the j-th sample.

[0122] For example, if the prediction accuracies of the seven basic prediction models are 95%, 93%, 96%, 94%, 95%, 93%, and 94%, respectively, the corresponding weight coefficients calculated by the entropy weight method are 0.15, 0.13, 0.16, 0.14, 0.15, 0.13, and 0.14, respectively, and the sum of the weights is 1.

[0123] Furthermore, the construction process of the "predictive model array" includes:

[0124] Historical electrical energy efficiency characteristic parameters, historical mechanical energy efficiency characteristic parameters, and historical thermodynamic energy efficiency characteristic parameters are collected from historical operating data to form a sample characteristic dataset;

[0125] Obtain the measured equipment power efficiency values ​​corresponding to the historical electrical energy efficiency characteristic parameters, historical mechanical energy efficiency characteristic parameters and historical thermodynamic energy efficiency characteristic parameters, and form a sample label dataset;

[0126] Based on machine learning algorithms, construct a predetermined number of basic prediction models;

[0127] The sample feature dataset and sample label dataset are used to train and validate each basic prediction model;

[0128] The various basic prediction models that have been verified to converge are integrated to form a prediction model array.

[0129] In this embodiment, historical electrical energy efficiency characteristic parameters, historical mechanical energy efficiency characteristic parameters, and historical thermodynamic energy efficiency characteristic parameters are first collected from historical operating data to form a sample feature dataset. Specifically, historical operating data refers to various types of data collected during the long-term operation of the equipment, including current data, vibration data, and temperature data under normal operating conditions and various abnormal operating conditions. Using the same feature extraction method as in step S10, historical electrical energy efficiency characteristic parameters, historical mechanical energy efficiency characteristic parameters, and historical thermodynamic energy efficiency characteristic parameters are extracted from the historical operating data to form the sample feature dataset. For example, the sample feature dataset contains 10,000 sets of historical electrical energy efficiency characteristic parameters, historical mechanical energy efficiency characteristic parameters, and historical thermodynamic energy efficiency characteristic parameters.

[0130] Secondly, the measured power efficiency values ​​of the equipment, corresponding to historical electrical energy efficiency characteristic parameters, historical mechanical energy efficiency characteristic parameters, and historical thermodynamic energy efficiency characteristic parameters, are obtained to form a sample label dataset. Specifically, the measured power efficiency values ​​are the actual power efficiency of the equipment obtained by measuring with a professional power analyzer. During measurement, the acquisition time of the corresponding characteristic parameters must be recorded simultaneously to ensure a one-to-one correspondence between sample features and sample labels. The sample label dataset has the same number of samples as the sample feature dataset, with each sample label being a numerical value, such as 90%, 92%, etc. For example, the sample label dataset contains 10,000 measured power efficiency values ​​of the equipment, each corresponding one-to-one with the 10,000 samples in the sample feature dataset.

[0131] Secondly, based on machine learning algorithms, a preset number of basic prediction models are constructed. The preset number can be dynamically determined according to the complexity of the equipment's operating conditions, and can be set to 10-20 to ensure the diversity of the prediction model array and cover the prediction needs under different operating conditions. For example, a basic prediction model can be constructed using the following technical path: Ten basic prediction models can be constructed using a deep neural network (DNN) architecture. The structure and construction logic of a single basic prediction model are as follows: Each basic prediction model consists of an input layer, a hidden layer, a regularization layer, and an output layer. The dimension of the input layer is N, where N is the total dimension of the electrical energy efficiency feature parameters, mechanical energy efficiency feature parameters, and thermodynamic energy efficiency feature parameters. The hidden layer consists of three fully connected hidden layers, with 64 neurons in the first layer, 32 in the second layer, and 16 in the third layer. The ReLU activation function is used, and the training stability is optimized through a batch normalization layer between each layer to accelerate convergence. A Dropout layer is inserted between the second and third hidden layers in the regularization layer, with a dropout probability of 0.2. The output layer has a dimension of 1 and uses the Sigmoid activation function to map the output value to the 0-1 range, outputting the predicted value of the power efficiency of a single device.

[0132] Furthermore, sample feature datasets and sample label datasets are used to train and validate each basic prediction model. The training and validation process is identical for all basic prediction models. Taking any one basic prediction model as an example: the sample feature dataset and sample label dataset are divided into a training set and a validation set in a 7:3 ratio. The training set is used for model parameter training, and the validation set is used for model performance validation. During training, historical electrical energy efficiency parameters, historical mechanical energy efficiency parameters, and historical thermodynamic energy efficiency parameters from the training set are used as input features, and the corresponding measured equipment power efficiency values ​​are used as supervision labels. Optimization algorithms such as backpropagation and gradient descent are used to adjust the parameters of the basic prediction model, minimizing the mean squared error between the predicted value and the label value. During validation, the prediction accuracy and mean squared error of the basic prediction model on the validation set are calculated. If the indicators of the basic prediction model on the validation set meet the preset requirements, such as prediction accuracy ≥ 90% and mean squared error ≤ 0.001, the basic prediction model is considered to have completed training, and the trained basic prediction model is obtained.

[0133] Finally, the validated and converged basic prediction models are integrated to form a prediction model array. Specifically, all validated and converged basic prediction models are selected, integrated, and stored in the model library of the cloud platform to form a prediction model array.

[0134] In summary, compared to existing technologies, the cloud platform described in this application receives the energy efficiency diagnostic report, dynamically schedules basic prediction models from a pre-set prediction model array, inputs the electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, and thermodynamic energy efficiency characteristic parameters from the energy efficiency diagnostic report, and outputs predicted values ​​for equipment power efficiency. Thus, by dynamically scheduling basic prediction models adapted to the current operating state of the equipment, the platform fully leverages the scenario adaptation advantages of multi-model collaboration, achieving accurate and robust prediction of equipment power efficiency, and providing a reliable data foundation for subsequent energy efficiency degradation correction and energy-carbon efficiency conversion.

[0135] S40: The cloud platform corrects the predicted power efficiency of the device based on the device's cumulative runtime and the pre-stored device energy efficiency attenuation benchmark curve, thereby obtaining the corrected predicted power efficiency of the device.

[0136] Throughout the entire life cycle of an equipment, as the cumulative operating time increases, natural losses such as component wear and insulation aging will cause the energy efficiency to decline in a regular manner. However, traditional equipment power efficiency prediction does not take into account this objective decline characteristic, resulting in a large deviation between the predicted value and the actual operating energy efficiency of the equipment.

[0137] To address the aforementioned issues, the cloud platform described in this application corrects the predicted power efficiency of the device based on the device's cumulative runtime and a pre-stored device energy efficiency attenuation baseline curve, thereby obtaining a corrected predicted power efficiency value.

[0138] Specifically, step S40 in the method includes:

[0139] The cloud platform obtains the cumulative runtime of the device;

[0140] Based on the cumulative runtime, the cloud platform queries the pre-stored device energy efficiency attenuation benchmark curve to obtain the benchmark energy efficiency value for the current runtime;

[0141] The cloud platform determines the attenuation correction weight coefficient based on the cumulative runtime of the device, wherein the attenuation correction weight coefficient increases as the cumulative runtime of the device increases;

[0142] The cloud platform weights and fuses the predicted power efficiency of the equipment with the benchmark energy efficiency value according to the attenuation correction weighting coefficient to calculate the corrected predicted power efficiency of the equipment.

[0143] In this embodiment, the cloud platform first obtains the device's cumulative runtime. Specifically, the device's cumulative runtime refers to the total operating time of the device from its factory installation to the current moment, stored in the device's controller or the device file on the cloud platform. The cloud platform obtains the device's cumulative runtime in real time through communication with the device side, ensuring the accuracy of the data. For example, the cloud platform obtains that the current cumulative runtime of a certain device is 10,000 hours through communication with the device side.

[0144] Secondly, based on the cumulative runtime, the cloud platform queries the pre-stored equipment energy efficiency degradation benchmark curve to obtain the benchmark energy efficiency value for the current runtime. Specifically, the equipment energy efficiency degradation benchmark curve is a curve fitted based on equipment lifecycle test data and historical operating data. The horizontal axis represents the cumulative runtime, and the vertical axis represents the benchmark energy efficiency value. The benchmark energy efficiency value is the theoretical energy efficiency value of the equipment under the corresponding cumulative runtime, reflecting the natural degradation trend of the equipment.

[0145] For example, the cloud platform can use interpolation to query the corresponding baseline energy efficiency value based on the current cumulative runtime. For instance, the baseline energy efficiency value in the device energy efficiency degradation baseline curve for a cumulative runtime of 10,000 hours is 88%.

[0146] Secondly, the cloud platform determines the attenuation correction weight coefficient based on the cumulative operating time of the device. This weight coefficient increases with the cumulative operating time. Specifically, the attenuation correction weight coefficient adjusts the influence of the baseline energy efficiency value in the correction calculation. A shorter cumulative operating time results in less energy efficiency degradation and a smaller weight coefficient; conversely, a longer cumulative operating time leads to more severe energy efficiency degradation and a larger weight coefficient.

[0147] For example, the attenuation correction weight coefficient ranges from 0 to 1 and can be determined using a piecewise function or a linear function. For instance, a linear function can be used: ω_decay = min(0.00005×T, 0.5), where T is the cumulative runtime, 0.00005 is the coefficient, and 0.5 is the maximum weight coefficient, avoiding over-correction. For example, if the cumulative runtime is 10,000 hours, then the attenuation correction weight coefficient ω_decay = min(0.00005×10000, 0.5) = 0.5.

[0148] Finally, the cloud platform weights and merges the predicted equipment power efficiency value with the baseline energy efficiency value according to the attenuation correction weight coefficient to calculate the corrected predicted equipment power efficiency value. Specifically, the weighted fusion calculation formula is: Corrected predicted equipment power efficiency value = (1 - attenuation correction weight coefficient) × predicted equipment power efficiency value + attenuation correction weight coefficient × baseline energy efficiency value. Through weighted calculation, the influence of the predicted equipment power efficiency value and the baseline energy efficiency value is balanced, so that the corrected predicted equipment power efficiency value reflects both the predicted energy efficiency under the current operating conditions and the influence of natural attenuation of the equipment.

[0149] For example, if the predicted power efficiency of the equipment is 92%, the baseline energy efficiency is 88%, and the attenuation correction weighting factor is 0.5, then the corrected predicted power efficiency of the equipment = (1-0.5)×92%+0.5×88%=90%.

[0150] In summary, compared to existing technologies, the cloud platform described in this application corrects the predicted power efficiency of the device based on the device's cumulative operating time and a pre-stored device energy efficiency degradation benchmark curve, resulting in a corrected predicted power efficiency value. This approach considers the natural energy efficiency degradation characteristics that occur throughout the device's lifespan with cumulative operating time, and specifically corrects the predicted power efficiency value, making the corrected predicted power efficiency value more closely match the actual operating energy efficiency status of the device, thus improving the accuracy of power efficiency prediction.

[0151] S50: The cloud platform, based on the real-time carbon emission factor of the power grid and the predicted power efficiency of the corrected equipment, outputs the predicted carbon efficiency of the equipment through carbon energy efficiency conversion calculation.

[0152] Existing equipment energy efficiency assessments mostly focus on power consumption efficiency itself, without taking into account the dynamic changes in the real-time carbon emission factors of the power grid. For example, if the adjustment of the power generation energy structure leads to fluctuations in carbon emission levels, it is impossible to quantify the effective energy output efficiency of the equipment corresponding to a unit of carbon emission.

[0153] To address the aforementioned issues, the cloud platform described in this application, based on the real-time carbon emission factor of the power grid and the predicted power efficiency of the corrected equipment, outputs the predicted carbon efficiency of the equipment through carbon energy efficiency conversion calculation.

[0154] Specifically, step S50 in the method includes:

[0155] The cloud platform determines the carbon emission conversion coefficient corresponding to the real-time carbon emission factor of the power grid by querying a pre-stored carbon emission factor conversion mapping table.

[0156] The cloud platform multiplies the corrected equipment power efficiency prediction value with the carbon emission conversion coefficient to calculate the equipment carbon efficiency prediction value, wherein the equipment carbon efficiency prediction value is used to characterize the effective energy output efficiency of the equipment corresponding to a unit of carbon emission.

[0157] In this embodiment, the cloud platform first determines the carbon emission conversion coefficient corresponding to the real-time carbon emission factor of the power grid by querying a pre-stored carbon emission factor conversion mapping table. Specifically, the carbon emission factor conversion mapping table is based on the power grid carbon emission factor and stores the correspondence between the real-time carbon emission factor of the power grid and the carbon emission conversion coefficient. The real-time carbon emission factor of the power grid can be obtained from real-time data released by the National Energy Administration and power grid companies, and the unit is... / kWh, the carbon emission conversion factor is a proportionality coefficient used to convert the electrical efficiency of equipment into energy-carbon efficiency, with units of kWh / It is the reciprocal of the carbon emission factor of the power grid.

[0158] For example, if the real-time carbon emission factor of the power grid is 0.8 If the carbon emission factor conversion mapping table corresponds to / kWh, then the carbon emission conversion coefficient is 1 / 0.8 = 1.25kWh / .

[0159] Secondly, the cloud platform multiplies the corrected predicted equipment power efficiency value by the carbon emission conversion coefficient to calculate the predicted equipment carbon efficiency value. This predicted value characterizes the effective energy output efficiency of the equipment per unit of carbon emission. Specifically, the formula for calculating the predicted equipment carbon efficiency value is: Predicted Equipment Carbon Efficiency Value = Corrected Predicted Equipment Power Efficiency Value × Carbon Emission Conversion Coefficient. This achieves the conversion from equipment power efficiency to energy output efficiency per unit of carbon emission, directly quantifying the effective energy output efficiency of the equipment.

[0160] Specifically, the unit for the predicted carbon efficiency of the equipment is kWh / A higher predicted carbon efficiency value for equipment indicates a greater effective energy output per unit of carbon emissions, signifying better low-carbon performance of the equipment. For example, a predicted carbon efficiency value of 1.125 kWh / This means that for every 1 kg of carbon dioxide emitted, the equipment can output 1.125 kWh of effective energy. The predicted carbon efficiency of the equipment provides a direct basis for the control of carbon emissions from the equipment and can be used for scenarios such as evaluating the effect of energy-saving retrofitting of equipment and selecting low-carbon equipment.

[0161] For example, if the predicted power efficiency of the corrected equipment is 90%, the carbon emission conversion factor is 1.25 kWh / Therefore, the predicted carbon efficiency of the equipment is 90% × 1.25 = 1.125 kWh / kWh. .

[0162] In summary, compared to existing technologies, the cloud platform described in this application, based on the real-time carbon emission factor of the power grid and the predicted power efficiency of the corrected equipment, outputs a predicted carbon efficiency value for the equipment through carbon-energy efficiency conversion calculations. Thus, by converting the dynamic changes of the real-time carbon emission factor of the power grid with the accurate predicted power efficiency of the corrected equipment, a predicted carbon efficiency value for the equipment that quantifies the effective energy output efficiency of the equipment per unit of carbon emission is output. This achieves an integrated and accurate characterization of energy efficiency and carbon emissions, providing data support for the low-carbon operation and management of equipment.

[0163] In summary, the embodiments of this application have at least the following technical effects:

[0164] Compared to existing technologies, this application first collects real-time current, vibration, and temperature data from the equipment at the device end. Through spectrum analysis, frequency domain analysis, and time series analysis, it extracts electrical, mechanical, and thermodynamic energy efficiency characteristic parameters. This achieves comprehensive capture of operational status data across multiple physical domains (electrical, mechanical, and thermodynamic). The multi-domain energy efficiency characteristic parameters extracted using targeted signal analysis methods accurately characterize the energy loss and operational status of each domain, providing a reliable data foundation for subsequent cross-domain coupling analysis, health status assessment, and accurate prediction of energy and carbon efficiency.

[0165] Secondly, based on the electrical, mechanical, and thermodynamic energy efficiency characteristic parameters, the device-side of this application calculates the cross-domain coupling coefficient, efficiency health index, and characteristic parameter volatility, and generates an energy efficiency diagnostic report containing anomaly type identifiers, which is then uploaded to the cloud platform. This achieves deep fusion analysis of the electrical, mechanical, and thermodynamic energy efficiency characteristics of the equipment, quantifies cross-domain coupling relationships, reduces the amount of data uploaded to the cloud platform, ensures real-time transmission, and provides highly targeted and reliable pre-diagnostic basis for cloud-based dynamic model scheduling and accurate prediction of power consumption efficiency.

[0166] Furthermore, the cloud platform described in this application receives the energy efficiency diagnostic report, dynamically schedules basic prediction models from a pre-set prediction model array, inputs the electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, and thermodynamic energy efficiency characteristic parameters from the energy efficiency diagnostic report, and outputs the predicted value of equipment power consumption efficiency. In this way, by dynamically scheduling basic prediction models adapted to the current operating state of the equipment, the advantages of multi-model collaboration in scenario adaptation are fully utilized, achieving accurate and robust prediction of equipment power consumption efficiency, and providing a reliable data foundation for subsequent energy efficiency degradation correction and energy-carbon efficiency conversion.

[0167] Furthermore, the cloud platform described in this application corrects the predicted power efficiency of the device based on the device's cumulative operating time and a pre-stored device energy efficiency degradation benchmark curve, resulting in a corrected predicted power efficiency value. This approach considers the natural energy efficiency degradation characteristics that occur throughout the device's lifespan with cumulative operating time, and specifically corrects the predicted power efficiency value, making the corrected predicted power efficiency value more closely match the actual operating energy efficiency status of the device, thus improving the accuracy of power efficiency prediction.

[0168] Finally, the cloud platform described in this application, based on the real-time carbon emission factor of the power grid and the predicted power efficiency of the corrected equipment, outputs a predicted energy-carbon efficiency value for the equipment through carbon-energy efficiency conversion calculation. Thus, by converting the dynamic changes of the real-time carbon emission factor of the power grid with the accurate predicted power efficiency of the corrected equipment, a predicted energy-carbon efficiency value for the equipment, which quantifies the effective energy output efficiency of the equipment per unit of carbon emission, is output. This achieves an integrated and accurate characterization of energy efficiency and carbon emissions, providing data support for the low-carbon operation and management of equipment.

[0169] Through the above technical solution, this application effectively addresses the pain points of traditional technologies, such as the disconnect between energy efficiency and carbon emission assessment, the limitations of single physical domain analysis, and insufficient end-to-cloud collaboration, by acquiring multi-domain data and extracting targeted features at the device end, calculating cross-domain coupling indicators, and uploading simplified diagnostic reports. Combined with the dynamic model scheduling, energy efficiency attenuation correction, and carbon energy efficiency conversion of the cloud platform, it achieves comprehensive perception of equipment operating status through multi-dimensional feature extraction and cross-domain coupling analysis. It also reduces data transmission volume and improves real-time performance by completing data preprocessing, feature extraction, and preliminary diagnosis at the device end. Furthermore, it improves the accuracy of equipment power consumption efficiency prediction by leveraging dynamic model scheduling and attenuation correction. Finally, through real-time carbon emission factor coupling conversion from the power grid, it achieves real-time, accurate, and integrated prediction of equipment energy and carbon efficiency, providing reliable data support and decision-making basis for the low-carbon operation and management of industrial equipment.

[0170] Example 2, as Figure 2As shown, based on the same inventive concept as the end-to-cloud collaborative device energy carbon efficiency real-time prediction method provided in Embodiment 1, this embodiment of the invention also provides an end-to-cloud collaborative device energy carbon efficiency real-time prediction system, including:

[0171] Data acquisition module 11 is used to collect current data, vibration data and temperature data of the equipment in real time at the equipment end. Through spectrum analysis, frequency domain analysis and time series analysis, electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters and thermodynamic energy efficiency characteristic parameters are extracted.

[0172] The anomaly identification module 12 is used to calculate the cross-domain coupling coefficient, efficiency health index and characteristic parameter fluctuation rate based on the electrical energy efficiency characteristic parameters, the mechanical energy efficiency characteristic parameters and the thermodynamic energy efficiency characteristic parameters on the device side, and generate an energy efficiency diagnosis report containing anomaly type identifier and upload it to the cloud platform.

[0173] The efficiency prediction module 13 is used for the cloud platform to receive the energy efficiency diagnosis report, dynamically schedule the basic prediction model from the preset prediction model array, input the electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters and thermodynamic energy efficiency characteristic parameters in the energy efficiency diagnosis report, and output the predicted value of the equipment's power consumption efficiency.

[0174] Efficiency correction module 14 is used by the cloud platform to correct the predicted power efficiency of the device based on the device's cumulative running time and the pre-stored device energy efficiency attenuation benchmark curve, so as to obtain the corrected predicted power efficiency of the device.

[0175] The calculation output module 15 is used by the cloud platform to calculate and output the predicted value of the device's carbon efficiency based on the real-time carbon emission factor of the power grid and the predicted value of the power consumption efficiency of the corrected device through carbon energy efficiency conversion.

[0176] Specifically, the data acquisition module 11 is used for:

[0177] The device end side collects the device's current data through a current transformer;

[0178] The device end side uses a triaxial vibration acceleration sensor to collect vibration data of the device;

[0179] The device end-side uses temperature sensors to collect the winding temperature and bearing temperature of the device as temperature data.

[0180] The device performs a fast Fourier transform on the collected current data to calculate the total harmonic distortion rate and the amplitude of each harmonic current, which are used as electrical energy efficiency characteristic parameters.

[0181] The device end performs spectral analysis on the collected vibration data and extracts the characteristic frequency amplitudes related to the mechanical faults of the device as mechanical energy efficiency characteristic parameters.

[0182] The device performs moving average filtering and differentiation on the collected temperature data to calculate the temperature rise rate of the device, which is used as a thermodynamic energy efficiency characteristic parameter.

[0183] The anomaly detection module 12 is specifically used for:

[0184] The end-side obtains the electromagnetic-mechanical coupling coefficient by calculating the coherence function value between a specific harmonic current amplitude and the corresponding characteristic frequency vibration amplitude, based on the amplitude of each harmonic current in the electrical energy efficiency characteristic parameters and the amplitude of the vibration characteristic frequency in the mechanical energy efficiency characteristic parameters.

[0185] The end-side obtains the electro-thermal coupling coefficient by calculating the correlation coefficient between harmonic loss power and temperature rise rate based on the total harmonic distortion rate in the electrical energy efficiency characteristic parameters and the temperature rise rate in the thermodynamic energy efficiency characteristic parameters.

[0186] The end-side calculates the harmonic loss power by summing the squares of the harmonic current amplitudes and the equivalent resistance of the equipment based on the electrical energy efficiency characteristic parameters.

[0187] The end side calculates the vibration loss power through energy mapping based on the vibration characteristic frequency amplitude and vibration energy conversion coefficient in the mechanical energy efficiency characteristic parameters.

[0188] The end side calculates the thermal gradient anomaly value by using the difference between the measured temperature rise rate and the theoretical temperature rise rate in the thermodynamic energy efficiency characteristic parameters.

[0189] The ratio of the harmonic loss power to the current total active power is used to obtain the harmonic loss ratio; the ratio of the vibration loss power to the current total active power is used to obtain the vibration loss ratio; and the ratio of the thermal gradient anomaly value to the equipment's rated temperature rise rate is used to obtain the thermal anomaly coefficient.

[0190] The harmonic loss ratio, vibration loss ratio and thermal anomaly coefficient are combined to form an energy efficiency and health feature vector.

[0191] The energy efficiency health feature vector is input into a pre-trained health assessment model to calculate the efficiency health index.

[0192] The end-side calculates the characteristic parameter volatility based on the historical data of electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, and thermodynamic energy efficiency characteristic parameters within the sliding time window, using the standard deviation.

[0193] Furthermore, the step of "generating an energy efficiency diagnostic report containing an anomaly type identifier and uploading it to the cloud platform" includes:

[0194] The electromagnetic-mechanical coupling coefficient is compared with a first preset threshold. When the electromagnetic-mechanical coupling coefficient exceeds the first preset threshold, a vibration anomaly type identifier is generated.

[0195] The electro-thermal coupling coefficient is compared with a second preset threshold. When the electro-thermal coupling coefficient exceeds the second preset threshold, a thermal anomaly type identifier is generated.

[0196] The terminal integrates the electrical energy efficiency characteristic parameters, the mechanical energy efficiency characteristic parameters, the thermodynamic energy efficiency characteristic parameters, the electromagnetic-mechanical coupling coefficient, the electro-thermal coupling coefficient, the efficiency health index, the characteristic parameter volatility, and the corresponding abnormality type identifier to form an energy efficiency diagnostic report, which is then uploaded to the cloud platform.

[0197] Specifically, the efficiency prediction module 13 is used for:

[0198] The cloud platform calculates the model scheduling coefficient based on the efficiency health index, cross-domain coupling coefficient, and characteristic parameter volatility in the energy efficiency diagnostic report through weighted fusion.

[0199] Multiply the model scheduling coefficient by the total number of models in the prediction model array and round down to determine the number of basic prediction models that need to be scheduled.

[0200] A corresponding number of basic prediction models are scheduled from the prediction model array to form a basic prediction model combination;

[0201] The electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters and thermodynamic energy efficiency characteristic parameters are input into each of the basic prediction models in the basic prediction model combination to obtain multiple preliminary prediction results;

[0202] Based on the anomaly type identifier, electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters and thermodynamic energy efficiency characteristic parameters in the energy efficiency diagnosis report, weight coefficients are assigned to each basic prediction model in the basic prediction model combination.

[0203] The cloud platform performs weighted fusion of multiple preliminary prediction results based on the weighting coefficients to obtain the predicted value of equipment power efficiency.

[0204] Specifically, the phrase "assigning weight coefficients to each basic prediction model in the basic prediction model combination based on the anomaly type identifier, electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, and thermodynamic energy efficiency characteristic parameters in the energy efficiency diagnostic report" includes:

[0205] Based on the anomaly type identifier, the cloud platform retrieves recent historical data similar to the current anomaly type from the historical database;

[0206] The cloud platform calculates the prediction accuracy of each basic prediction model on the recent historical data;

[0207] The cloud platform is based on the entropy weight method, which calculates the corresponding weight coefficients according to the prediction accuracy of each basic prediction model. The weight coefficients satisfy the condition that the sum of the weight coefficients of each basic prediction model is 1.

[0208] Specifically, the construction process of the "prediction model array" includes:

[0209] Historical electrical energy efficiency characteristic parameters, historical mechanical energy efficiency characteristic parameters, and historical thermodynamic energy efficiency characteristic parameters are collected from historical operating data to form a sample characteristic dataset;

[0210] Obtain the measured equipment power efficiency values ​​corresponding to the historical electrical energy efficiency characteristic parameters, historical mechanical energy efficiency characteristic parameters and historical thermodynamic energy efficiency characteristic parameters, and form a sample label dataset;

[0211] Based on machine learning algorithms, construct a predetermined number of basic prediction models;

[0212] The sample feature dataset and sample label dataset are used to train and validate each basic prediction model;

[0213] The various basic prediction models that have been verified to converge are integrated to form a prediction model array.

[0214] Specifically, the efficiency correction module 14 is used for:

[0215] The cloud platform obtains the cumulative runtime of the device;

[0216] Based on the cumulative runtime, the cloud platform queries the pre-stored device energy efficiency attenuation benchmark curve to obtain the benchmark energy efficiency value for the current runtime;

[0217] The cloud platform determines the attenuation correction weight coefficient based on the cumulative runtime of the device, wherein the attenuation correction weight coefficient increases as the cumulative runtime of the device increases;

[0218] The cloud platform weights and fuses the predicted power efficiency of the equipment with the benchmark energy efficiency value according to the attenuation correction weighting coefficient to calculate the corrected predicted power efficiency of the equipment.

[0219] Specifically, the calculation output module 15 is used for:

[0220] The cloud platform determines the carbon emission conversion coefficient corresponding to the real-time carbon emission factor of the power grid by querying a pre-stored carbon emission factor conversion mapping table.

[0221] The cloud platform multiplies the corrected equipment power efficiency prediction value with the carbon emission conversion coefficient to calculate the equipment carbon efficiency prediction value, wherein the equipment carbon efficiency prediction value is used to characterize the effective energy output efficiency of the equipment corresponding to a unit of carbon emission.

[0222] In summary, the embodiments of this application have at least the following technical effects:

[0223] Compared to existing technologies, this application firstly uses a data acquisition module to collect real-time current, vibration, and temperature data from the equipment. Through spectrum analysis, frequency domain analysis, and time series analysis, it extracts electrical, mechanical, and thermodynamic energy efficiency characteristic parameters, achieving comprehensive capture of the equipment's operational status data across multiple physical domains (electrical, mechanical, and thermodynamic), accurately characterizing the energy efficiency loss and operational status of each domain. Secondly, through an anomaly identification module, the equipment side calculates cross-domain coupling coefficients, efficiency health indices, and characteristic parameter volatility based on the electrical, mechanical, and thermodynamic energy efficiency characteristic parameters. It then generates an energy efficiency diagnostic report containing anomaly type identifiers and uploads it to the cloud platform. This achieves deep fusion analysis of the equipment's multi-domain energy efficiency characteristics (electrical, mechanical, and thermodynamic), quantifies cross-domain coupling relationships, reduces the amount of data uploaded to the cloud platform, and ensures real-time transmission. Furthermore, through the efficiency prediction module, the cloud platform receives energy efficiency diagnostic reports, dynamically schedules basic prediction models from a pre-set prediction model array, inputs electrical, mechanical, and thermodynamic energy efficiency characteristic parameters from the energy efficiency diagnostic report, and outputs predicted equipment power efficiency values. By dynamically scheduling basic prediction models adapted to the current operating state of the equipment, the platform fully leverages the scenario adaptation advantages of multi-model collaboration, achieving accurate and robust prediction of equipment power efficiency. Further, through the efficiency correction module, the cloud platform corrects the predicted equipment power efficiency values ​​based on the equipment's cumulative operating time and a pre-stored equipment energy efficiency degradation baseline curve, obtaining corrected predicted equipment power efficiency values. This takes into account the natural energy efficiency degradation characteristics that occur throughout the equipment's lifecycle with cumulative operating time, improving the accuracy of power efficiency prediction. Finally, through the calculation output module, the cloud platform, based on the real-time carbon emission factor of the power grid and the corrected predicted equipment power efficiency values, calculates carbon efficiency conversion and outputs predicted equipment carbon efficiency values, achieving an integrated and accurate characterization of energy efficiency and carbon emissions, providing data support for low-carbon operation management of equipment. In this way, real-time, accurate, and integrated prediction of equipment carbon efficiency is achieved, providing reliable data support and decision-making basis for the low-carbon operation and management of industrial equipment.

[0224] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0225] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0226] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0227] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0228] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0229] Although preferred embodiments of the invention have been described, those skilled in the art, once they have learned the basic inventive concept, can make other changes and modifications to these embodiments.

[0230] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of this invention and its equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for real-time prediction of device carbon efficiency through edge-cloud collaboration, characterized in that, The method includes: The equipment end-side collects current data, vibration data and temperature data of the equipment in real time, and extracts electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters and thermodynamic energy efficiency characteristic parameters through spectrum analysis, frequency domain analysis and time series analysis. Based on the electrical energy efficiency characteristic parameters, the mechanical energy efficiency characteristic parameters, and the thermodynamic energy efficiency characteristic parameters, the device end calculates the cross-domain coupling coefficient, the efficiency health index, and the characteristic parameter fluctuation rate, and generates an energy efficiency diagnostic report containing anomaly type identifiers and uploads it to the cloud platform. The cross-domain coupling coefficient includes the electromagnetic-mechanical coupling coefficient and the electro-thermal coupling coefficient. The cloud platform receives the energy efficiency diagnostic report, dynamically schedules the basic prediction model from the preset prediction model array, inputs the electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters and thermodynamic energy efficiency characteristic parameters from the energy efficiency diagnostic report, and outputs the predicted value of the equipment's power consumption efficiency. The cloud platform corrects the predicted power efficiency of the equipment by adjusting the energy efficiency attenuation based on the cumulative running time of the equipment and the pre-stored equipment energy efficiency attenuation benchmark curve, thus obtaining the corrected predicted power efficiency of the equipment. The cloud platform, based on the real-time carbon emission factor of the power grid and the predicted power efficiency of the corrected equipment, outputs the predicted carbon efficiency of the equipment through carbon energy efficiency conversion calculation. Specifically, the device-side calculates the cross-domain coupling coefficient, efficiency health index, and characteristic parameter volatility based on the electrical energy efficiency characteristic parameters, the mechanical energy efficiency characteristic parameters, and the thermodynamic energy efficiency characteristic parameters, including: Based on the amplitude of each harmonic current in the electrical energy efficiency characteristic parameters and the amplitude of the vibration characteristic frequency in the mechanical energy efficiency characteristic parameters, the device end side calculates the coherence function value between a specific harmonic current amplitude and the corresponding characteristic frequency vibration amplitude to obtain the electromagnetic-mechanical coupling coefficient. Based on the total harmonic distortion rate in the electrical energy efficiency characteristic parameters and the temperature rise rate in the thermodynamic energy efficiency characteristic parameters, the device end side calculates the correlation coefficient between harmonic loss power and temperature rise rate to obtain the electro-thermal coupling coefficient. The device end-side calculates the harmonic loss power by summing the squares of the harmonic current amplitudes and the equivalent resistance of the device based on the electrical energy efficiency characteristic parameters. The equipment end-side calculates the vibration loss power through energy mapping based on the vibration characteristic frequency amplitude and vibration energy conversion coefficient in the mechanical energy efficiency characteristic parameters. The device end-side calculates the thermal gradient anomaly value by using the difference between the measured temperature rise rate and the theoretical temperature rise rate in the thermodynamic energy efficiency characteristic parameters. The ratio of the harmonic loss power to the current total active power is used to obtain the harmonic loss ratio; the ratio of the vibration loss power to the current total active power is used to obtain the vibration loss ratio; and the ratio of the thermal gradient anomaly value to the equipment's rated temperature rise rate is used to obtain the thermal anomaly coefficient. The harmonic loss ratio, vibration loss ratio and thermal anomaly coefficient are combined to form an energy efficiency and health feature vector. The energy efficiency health feature vector is input into a pre-trained health assessment model to calculate the efficiency health index. The device end-side calculates the characteristic parameter volatility based on the historical data of electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, and thermodynamic energy efficiency characteristic parameters within a sliding time window, using the standard deviation.

2. The method for real-time prediction of device energy and carbon efficiency through edge-cloud collaboration according to claim 1, characterized in that, The equipment terminal collects real-time current, vibration, and temperature data. Through spectrum analysis, frequency domain analysis, and time series analysis, it extracts electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, and thermodynamic energy efficiency characteristic parameters, including: The device end side collects the device's current data through a current transformer; The device end side uses a triaxial vibration acceleration sensor to collect vibration data of the device; The device end-side uses temperature sensors to collect the winding temperature and bearing temperature of the device as temperature data. The device performs a fast Fourier transform on the collected current data to calculate the total harmonic distortion rate and the amplitude of each harmonic current, which are used as electrical energy efficiency characteristic parameters. The device end performs spectral analysis on the collected vibration data and extracts the characteristic frequency amplitudes related to the mechanical faults of the device as mechanical energy efficiency characteristic parameters. The device performs moving average filtering and differentiation on the collected temperature data to calculate the temperature rise rate of the device, which is used as a thermodynamic energy efficiency characteristic parameter.

3. The method for real-time prediction of device energy carbon efficiency through edge-cloud collaboration according to claim 1, characterized in that, Generate an energy efficiency diagnostic report containing anomaly type identifiers and upload it to the cloud platform, including: The electromagnetic-mechanical coupling coefficient is compared with a first preset threshold. When the electromagnetic-mechanical coupling coefficient exceeds the first preset threshold, a vibration anomaly type identifier is generated. The electro-thermal coupling coefficient is compared with a second preset threshold. When the electro-thermal coupling coefficient exceeds the second preset threshold, a thermal anomaly type identifier is generated. The device integrates the electrical energy efficiency characteristic parameters, the mechanical energy efficiency characteristic parameters, the thermodynamic energy efficiency characteristic parameters, the electromagnetic-mechanical coupling coefficient, the electro-thermal coupling coefficient, the efficiency health index, the characteristic parameter volatility, and the corresponding abnormality type identifier to form an energy efficiency diagnostic report, which is then uploaded to the cloud platform.

4. The method for real-time prediction of device energy carbon efficiency through edge-cloud collaboration according to claim 1, characterized in that, The cloud platform receives the energy efficiency diagnostic report, dynamically schedules basic prediction models from a pre-set prediction model array, inputs electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, and thermodynamic energy efficiency characteristic parameters from the energy efficiency diagnostic report, and outputs predicted values ​​of equipment power consumption efficiency, including: The cloud platform calculates the model scheduling coefficient based on the efficiency health index, cross-domain coupling coefficient, and characteristic parameter volatility in the energy efficiency diagnostic report through weighted fusion. Multiply the model scheduling coefficient by the total number of models in the prediction model array and round down to determine the number of basic prediction models that need to be scheduled. A corresponding number of basic prediction models are scheduled from the prediction model array to form a basic prediction model combination; The electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters and thermodynamic energy efficiency characteristic parameters are input into each of the basic prediction models in the basic prediction model combination to obtain multiple preliminary prediction results; Based on the anomaly type identifier, electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters and thermodynamic energy efficiency characteristic parameters in the energy efficiency diagnosis report, weight coefficients are assigned to each basic prediction model in the basic prediction model combination. The cloud platform performs weighted fusion of multiple preliminary prediction results based on the weighting coefficients to obtain the predicted value of equipment power efficiency.

5. The method for real-time prediction of device energy and carbon efficiency through edge-cloud collaboration according to claim 4, characterized in that, Based on the anomaly type identifier, electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, and thermodynamic energy efficiency characteristic parameters in the energy efficiency diagnostic report, weight coefficients are assigned to each basic prediction model in the basic prediction model combination, including: Based on the anomaly type identifier, the cloud platform retrieves recent historical data similar to the current anomaly type from the historical database; The cloud platform calculates the prediction accuracy of each basic prediction model on the recent historical data; The cloud platform is based on the entropy weight method, which calculates the corresponding weight coefficients according to the prediction accuracy of each basic prediction model. The weight coefficients satisfy the condition that the sum of the weight coefficients of each basic prediction model is 1.

6. The method for real-time prediction of device energy carbon efficiency through edge-cloud collaboration according to claim 4, characterized in that, The process of constructing the predictive model array includes: Historical electrical energy efficiency characteristic parameters, historical mechanical energy efficiency characteristic parameters, and historical thermodynamic energy efficiency characteristic parameters are collected from historical operating data to form a sample characteristic dataset; Obtain the measured equipment power efficiency values ​​corresponding to the historical electrical energy efficiency characteristic parameters, historical mechanical energy efficiency characteristic parameters and historical thermodynamic energy efficiency characteristic parameters, and form a sample label dataset; Based on machine learning algorithms, construct a predetermined number of basic prediction models; The sample feature dataset and sample label dataset are used to train and validate each basic prediction model; The various basic prediction models that have been verified to converge are integrated to form a prediction model array.

7. The method for real-time prediction of device energy carbon efficiency through edge-cloud collaboration according to claim 1, characterized in that, The cloud platform corrects the predicted power efficiency of the equipment based on the cumulative operating time of the equipment and the pre-stored equipment energy efficiency degradation benchmark curve, resulting in a corrected predicted power efficiency value, including: The cloud platform obtains the cumulative runtime of the device; Based on the cumulative runtime, the cloud platform queries the pre-stored device energy efficiency attenuation benchmark curve to obtain the benchmark energy efficiency value for the current runtime; The cloud platform determines the attenuation correction weight coefficient based on the cumulative runtime of the device, wherein the attenuation correction weight coefficient increases as the cumulative runtime of the device increases; The cloud platform weights and fuses the predicted power efficiency of the equipment with the benchmark energy efficiency value according to the attenuation correction weighting coefficient to calculate the corrected predicted power efficiency of the equipment.

8. The method for real-time prediction of device energy carbon efficiency through edge-cloud collaboration according to claim 1, characterized in that, The cloud platform, based on the real-time carbon emission factor of the power grid and the predicted power efficiency of the corrected equipment, calculates carbon efficiency conversion and outputs the predicted carbon efficiency of the equipment, including: The cloud platform determines the carbon emission conversion coefficient corresponding to the real-time carbon emission factor of the power grid by querying a pre-stored carbon emission factor conversion mapping table. The cloud platform multiplies the predicted power efficiency of the corrected equipment with the carbon emission conversion coefficient to calculate the predicted energy-carbon efficiency of the equipment. The predicted carbon efficiency of the equipment is used to characterize the effective energy output efficiency of the equipment corresponding to a unit of carbon emission.

9. A real-time prediction system for device carbon efficiency through edge-cloud collaboration, characterized in that, A method for real-time prediction of device energy carbon efficiency through edge-cloud collaboration as described in any one of claims 1-8 includes: The data acquisition module is used to collect current, vibration and temperature data of the equipment in real time at the equipment end. Through spectrum analysis, frequency domain analysis and time series analysis, it extracts electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters and thermodynamic energy efficiency characteristic parameters. An anomaly identification module is used on the device side to calculate the cross-domain coupling coefficient, efficiency health index and characteristic parameter fluctuation rate based on the electrical energy efficiency characteristic parameters, the mechanical energy efficiency characteristic parameters and the thermodynamic energy efficiency characteristic parameters, and to generate an energy efficiency diagnosis report containing an anomaly type identifier and upload it to the cloud platform. The cross-domain coupling coefficient includes the electromagnetic-mechanical coupling coefficient and the electro-thermal coupling coefficient. The efficiency prediction module is used by the cloud platform to receive the energy efficiency diagnosis report, dynamically schedule the basic prediction model from the preset prediction model array, input the electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters and thermodynamic energy efficiency characteristic parameters in the energy efficiency diagnosis report, and output the predicted value of the equipment's power consumption efficiency. The efficiency correction module is used by the cloud platform to correct the predicted power efficiency of the device based on the device's cumulative runtime and the pre-stored device energy efficiency attenuation baseline curve, so as to obtain the corrected predicted power efficiency of the device. The calculation output module is used by the cloud platform to calculate and output the predicted value of the device's carbon efficiency based on the real-time carbon emission factor of the power grid and the predicted value of the power consumption efficiency of the corrected device through carbon energy efficiency conversion. Specifically, the device-side calculates the cross-domain coupling coefficient, efficiency health index, and characteristic parameter volatility based on the electrical energy efficiency characteristic parameters, the mechanical energy efficiency characteristic parameters, and the thermodynamic energy efficiency characteristic parameters, including: Based on the amplitude of each harmonic current in the electrical energy efficiency characteristic parameters and the amplitude of the vibration characteristic frequency in the mechanical energy efficiency characteristic parameters, the device end side calculates the coherence function value between a specific harmonic current amplitude and the corresponding characteristic frequency vibration amplitude to obtain the electromagnetic-mechanical coupling coefficient. Based on the total harmonic distortion rate in the electrical energy efficiency characteristic parameters and the temperature rise rate in the thermodynamic energy efficiency characteristic parameters, the device end side calculates the correlation coefficient between harmonic loss power and temperature rise rate to obtain the electro-thermal coupling coefficient. The device end-side calculates the harmonic loss power by summing the squares of the harmonic current amplitudes and the equivalent resistance of the device based on the electrical energy efficiency characteristic parameters. The equipment end-side calculates the vibration loss power through energy mapping based on the vibration characteristic frequency amplitude and vibration energy conversion coefficient in the mechanical energy efficiency characteristic parameters. The device end-side calculates the thermal gradient anomaly value by using the difference between the measured temperature rise rate and the theoretical temperature rise rate in the thermodynamic energy efficiency characteristic parameters. The ratio of the harmonic loss power to the current total active power is used to obtain the harmonic loss ratio; the ratio of the vibration loss power to the current total active power is used to obtain the vibration loss ratio; and the ratio of the thermal gradient anomaly value to the equipment's rated temperature rise rate is used to obtain the thermal anomaly coefficient. The harmonic loss ratio, vibration loss ratio and thermal anomaly coefficient are combined to form an energy efficiency and health feature vector. The energy efficiency health feature vector is input into a pre-trained health assessment model to calculate the efficiency health index. The device end-side calculates the characteristic parameter volatility based on the historical data of electrical energy efficiency characteristic parameters, mechanical energy efficiency characteristic parameters, and thermodynamic energy efficiency characteristic parameters within a sliding time window, using the standard deviation.

Images