An energy-saving variable frequency fan aerodynamic performance abnormal energy efficiency detection method and system

Abstract

The present application relates to the running state monitoring technology of industrial variable frequency fan, disclose a kind of energy-saving type variable frequency fan aerodynamic performance abnormal energy efficiency detection method and system.Method includes: the mechanical angular velocity and electromagnetic torque of driving motor are obtained, acceleration phase and deceleration phase are matched and covered in the same effective analysis speed window according to angular acceleration division, time domain data is mapped to speed domain, and interpolation torque function, interpolation angular acceleration function and mean curve of two-phase state are constructed;With the quotient of total moment of inertia deduction torque difference integral and angular acceleration difference integral, the quasi-steady torque coefficient is obtained by fitting mean curve;With the ratio of aerodynamic difference coefficient and quasi-steady torque coefficient square and after gas density correction, aerodynamic degradation index is obtained, and alarm is triggered when its relative health benchmark occurs systematic deviation.The algebraic relationship of the index is that the comprehensive quadratic resistance coefficient of pipe network is removed, and the change of pipe network resistance and the degradation of blade aerodynamic performance can be distinguished without adding fluid measuring element.

Description

Technical Field

[0001] This invention relates to the technology for monitoring the operating status of industrial variable frequency fans, and in particular to a method and system for detecting abnormal aerodynamic performance of energy-saving variable frequency fans. Background Technology

[0002] Industrial variable frequency fans are major energy-consuming devices in ventilation and dust removal systems, boiler induced draft fans, mine ventilation systems, and large building HVAC systems. They commonly use variable frequency speed control to adjust speed according to real-time load commands to reduce throttling losses. However, fans operate for extended periods in dusty, high-humidity, or corrosive airflow environments. Inevitably, microparticles adhere to the blade surface, leading to scaling, corrosion, and slight wear. This alters the blade airfoil and surface roughness, thickens the airflow boundary layer, and shifts the separation point forward. Consequently, the efficiency of the fan in converting motor shaft power into fluid pressure and kinetic energy continuously declines. Initially, this degradation does not disrupt the rotor's mass distribution balance, and mechanical fault diagnosis methods based on vibration acceleration spectrum, envelope demodulation, or acoustic emission signals are largely ineffective. Often, identifiable characteristics only appear in vibration signals when degradation accumulates to the point of large-scale stall, structural fatigue, or even blade fracture. Before this point, the fan has been operating in an inefficient range for an extended period, with ineffective power consumption continuously accumulating.

[0003] To detect energy efficiency degradation before mechanical damage occurs, existing technologies mostly follow a performance mapping approach. This involves using pitot tubes, averaging pitot tubes, or high-precision differential pressure transmitters to measure the static pressure difference and volumetric flow rate at the fan inlet and outlet. Combined with the motor's output power and speed, this provides a real-time pressure-flow operating point, which is then compared to the factory efficiency curve to determine if the fan has deviated from its high-efficiency range or entered a stall or surge region. Other solutions involve machining pressure taps at specific geometric locations such as the fan's inlet cone throat and arranging pressure sensor arrays. By comparing the measured static pressure difference with the stall static pressure, stall protection and operational control are achieved. On the data processing side, normal behavior models or unsupervised time-series anomaly detection methods are frequently used. These methods use error vectors and likelihood values ​​to filter out outliers from multivariate operating data that deviate from the normal distribution.

[0004] The above solutions, when implemented in actual field operations, are hampered by the same unavoidable problem. Industrial air duct networks are inherently time-varying resistance systems. The opening and closing of variable air volume (VAV) terminal dampers, changes in damper openings, and the gradual clogging of dust collector bags and filters all cause significant changes in the overall quadratic resistance coefficient of the network. When a decrease in airflow, an increase in static pressure, or an increase in power consumption per unit airflow is observed at a given speed, this set of characteristics could originate from a decrease in the aerodynamic work capacity of the blades themselves, or it could simply be due to an increase in the resistance of the external network. These two factors exhibit high homogeneity in steady-state operating data, and existing detection logic cannot mathematically separate these two causes, resulting in frequent false alarms and missed alarms, making it impossible to schedule targeted blade cleaning and maintenance.

[0005] The problem is also directly related to the steady-state assumptions upon which existing methods rely. To avoid the drastic fluctuations in data during the acceleration and deceleration phases of frequency conversion damaging model stability, existing systems typically employ rigid steady-state discrimination logic, treating periods where the rate of change in speed or pressure fluctuation exceeds a threshold as invalid transient noise and eliminating them. However, for wind turbines that frequently adjust to unit load and ambient temperature, the actual steady-state time is very low, resulting in extremely sparse data for evaluation and a widening monitoring blind zone. Furthermore, the tiny pressure taps of precision pressure-tapping elements are easily clogged in high-dust, high-humidity, or sulfur-containing flue gas environments, and the diaphragms are also prone to damage. Their failure rate is even higher than that of the wind turbine itself. Coupled with the construction costs and downtime associated with pipe cutting and modification, this type of detection method is difficult to widely implement.

[0006] Therefore, how to distinguish the changes in external pipeline resistance from the degradation of the blade's aerodynamic performance during the dynamic process of a wind turbine continuously accelerating and decelerating in accordance with frequency conversion commands, without relying on additional fluid measurement elements, is an urgent problem to be solved in this field. Summary of the Invention

[0007] The purpose of this invention is to provide an energy-saving variable frequency fan aerodynamic performance abnormality detection method and system, in order to solve the technical problem that the existing technology cannot distinguish the change of external pipeline resistance from the degradation of the blade aerodynamic performance from highly homogeneous operating data without adding fluid measurement elements such as Pitot tubes and differential pressure transmitters, thus leading to false alarms and missed alarms.

[0008] To achieve the above objectives, the present invention provides a method for detecting abnormal aerodynamic performance of an energy-saving variable frequency fan, comprising the following steps: Continuous time-series status data of industrial variable frequency fan units during operation is obtained, and the status data includes at least the mechanical angular velocity and electromagnetic torque of the drive motor; The angular acceleration is calculated based on the mechanical angular velocity, and the continuous time-series state data is segmented into motion phases according to a preset phase determination threshold. The acceleration phase time-series data and deceleration phase time-series data are identified and extracted. Within the set optimization time span, matching acceleration phase time series data and deceleration phase time series data covering the same effective analysis speed window are used to construct an effective dual analysis group; The time-domain data in the effective dual analysis group is converted to the speed domain, and the interpolated torque function and interpolated angular acceleration function of the acceleration phase and deceleration phase at the same speed scale are constructed respectively. The mean torque curve and mean angular acceleration curve of the two phase states are obtained. Within the effective analysis speed window, the integral of the torque difference between the interpolated torque functions of the two phases and the integral of the angular acceleration difference between the interpolated angular acceleration functions are calculated respectively. The aerodynamic difference coefficient, which is not affected by the fluid excitation intensity, is obtained by subtracting the quotient of the integral of the torque difference and the integral of the angular acceleration difference from the total moment of inertia. The mean torque curve and the mean angular acceleration curve are fitted together to extract the quasi-steady-state torque coefficient, which includes compensation for the inherent mechanical losses of the equipment. The aerodynamic degradation index is calculated based on the ratio of the aerodynamic difference coefficient to the square of the quasi-steady-state torque coefficient, and an aerodynamic performance abnormality alarm is triggered when the aerodynamic degradation index deviates systematically from the healthy benchmark.

[0009] The aforementioned basic scheme transforms the variable frequency acceleration and deceleration transient data from previously excluded interference objects into a source of features, so that the evaluation of wind turbine aerodynamic performance no longer relies on absolute steady state as a prerequisite, thus expanding the range of data available for evaluation and the time coverage of monitoring.

[0010] As a preferred embodiment of the present invention, the phase state determination threshold is determined by the inherent no-load limit angular acceleration of the industrial variable frequency fan unit combined with a proportional coefficient. When the effective dual analysis group is not obtained for a continuous period exceeding a set time, the speed setpoint of the variable frequency drive controller is superimposed with a low-frequency sinusoidal disturbance command for active excitation. The maximum angular acceleration generated by the sinusoidal disturbance command is greater than the phase state determination threshold, and the effective analysis speed window is reconstructed from its actual extreme speed value. Using the inherent quantity of the equipment, rather than an arbitrarily modifiable acceleration / deceleration time setpoint, as the segmentation scale avoids the judgment scale drifting with control parameters. The active excitation parameter and the phase state determination threshold are combined to obtain evaluation samples that can be identified by the phase state segmentation process, even in situations where the process is too stable and lacks natural acceleration / deceleration conditions.

[0011] As a preferred embodiment of the present invention, the process of converting the time-domain data in the effective dual analysis group to the speed domain includes: performing spline interpolation on the acceleration phase time series data with monotonically increasing mechanical angular velocity as the independent variable to generate a smooth acceleration phase interpolated torque function and acceleration phase interpolated angular acceleration function; performing equivalent interpolation on the deceleration phase time series data with monotonically decreasing mechanical angular velocity as the independent variable and performing coordinate inversion to generate a smooth deceleration phase interpolated torque function and deceleration phase interpolated angular acceleration function; adding the interpolated torque functions of the acceleration phase and deceleration phase and averaging them to form a torque mean curve, and adding the interpolated angular acceleration functions of the acceleration phase and deceleration phase and averaging them to form an angular acceleration mean curve. The speed domain mapping allows the comparison of the two phase states to be performed on the same speed scale, avoiding deviations introduced by inconsistent operating points, and simultaneously obtaining quasi-steady-state operating characteristics independent of the difference curve from the mean curve.

[0012] As a preferred embodiment of the present invention, the calculation process of the aerodynamic difference coefficient includes: calculating the definite integral of the difference between the interpolated torque functions of the accelerating and decelerating phases within the effective analysis speed window to obtain the hysteresis torque area; calculating the definite integral of the difference between the interpolated angular acceleration functions of the accelerating and decelerating phases within the effective analysis speed window to obtain the angular acceleration difference area; and subtracting the quotient of the hysteresis torque area and the angular acceleration difference area from the calculated total moment of inertia to obtain the aerodynamic difference coefficient. This aerodynamic difference coefficient characterizes the aerodynamic information contributed by the flow hysteresis caused by fluid inertia. Interval integration has a suppressive effect on local spikes in the sampling sequence, and using the angular acceleration difference area as the quotient eliminates the influence of differences in the severity of acceleration and deceleration during different equipment operations on the absolute value of the characteristic quantity.

[0013] As a preferred embodiment of the present invention, the quasi-steady-state torque coefficient is determined by performing a quadratic term coefficient obtained by fitting a quadratic polynomial least squares method within the speed range using the difference between the mean torque curve and the product of the reduced total moment of inertia and the mean angular acceleration curve. The first-order term and constant term in the fitting formula are used to absorb the inherent mechanical losses of the equipment as the speed changes. The calculation and alarm triggering process of the aerodynamic degradation index includes: dividing the aerodynamic difference coefficient by the square of the quasi-steady-state torque coefficient and multiplying it by the ratio of the current measured gas density to the reference density for correction to obtain the aerodynamic degradation index for the current period; calculating the deviation rate of the aerodynamic degradation index relative to the preset health benchmark and inputting the deviation rate into an exponentially weighted moving average filter to extract the long-term evolution trend; when the filtered long-term evolution trend exceeds the degradation sensitivity threshold and the linear regression slope of the recent trend is not zero and the direction is consistent, an alarm signal for early degradation of blade aerodynamic performance is output. In the above algebraic relationship, the comprehensive quadratic resistance coefficient of the pipeline network is canceled out, so that the fluctuation of the operating conditions on the pipeline side will not cause a systematic shift in the index, while the gas density ratio correction eliminates the influence of the changes in the physical properties of the medium with the seasons and operating conditions.

[0014] To implement the above method, the present invention also provides an energy-saving variable frequency fan aerodynamic performance abnormality energy efficiency detection system, comprising a variable frequency drive controller and an aerodynamic performance intelligent detection controller connected in sequence; the aerodynamic performance intelligent detection controller is equipped with: The high-frequency data acquisition and preprocessing module is used to read the mechanical angular velocity and electromagnetic torque from the variable frequency drive controller at high frequency, and to perform filtering and validity verification. The motion phase identification and segmentation module is used to calculate angular acceleration and dynamically segment continuously maintained acceleration phase time series data and deceleration phase time series data according to the sign of angular acceleration and the phase determination threshold. The dual phase state optimization matching module is used to pair acceleration phases and deceleration phases that cover the same effective analysis speed window and whose average angular acceleration differences meet the conditions when independently searching historical phase states as effective dual analysis groups. The dynamic hysteresis loop construction module is used to map the effective dual analysis group to the speed domain and generate the interpolated torque function, the interpolated angular acceleration function, and the phase mean curve; The area integral decoupling module is used to calculate the integral of torque difference and angular acceleration difference between two phase states. Based on their ratio and the converted total moment of inertia, the aerodynamic difference coefficient is extracted, and the aerodynamic degradation index is calculated by combining the quasi-steady-state torque coefficient obtained by fitting the mean curve. The anomaly detection and early warning module is used to monitor the deviation trend of the aerodynamic degradation index relative to the healthy benchmark, trigger an alarm signal, and output it to the upper-level monitoring system.

[0015] As a preferred technical solution of the present invention, the phase determination threshold used in the motion phase identification and segmentation module is the unit's no-load limit angular acceleration multiplied by a proportional coefficient; the dual phase optimization matching module is also configured to actively acquire transition state data by sending a superimposed low-frequency sinusoidal speed disturbance command to the frequency converter drive controller when long-term data is missing under natural operating conditions.

[0016] As a preferred technical solution of the present invention, the dynamic hysteresis loop construction module uses a cubic spline interpolation algorithm to process the original time domain data, so that under a unified speed scale, the acceleration phase torque curve and the deceleration phase torque curve form a dynamic aerodynamic hysteresis loop region with an enclosing area.

[0017] As a preferred technical solution of the present invention, the area integral decoupling module calculates the numerical integral ratio or linear regression slope of the torque difference relative to the angular acceleration difference, and deducts the torque consumed by the mechanical rigid components in the form of equivalent inertia, thus separating out the characteristic quantity determined only by the aerodynamic coefficient of the fan body and the fluid inertia.

[0018] As a preferred embodiment of the present invention, the anomaly detection and early warning module is pre-configured with adaptive unsupervised baseline learning logic. This logic is used to statistically analyze the expected mean of the aerodynamic degradation index corresponding to a set number of effective dual analysis groups after initial equipment commissioning or blade repair to establish the health benchmark. In silent monitoring mode, an exponentially weighted moving average is used to track degradation deviations. The detection function is deployed in software within existing control equipment, without requiring modifications to the physical structure of the on-site air duct.

[0019] Compared with the prior art, the present invention has the following beneficial effects: At the same rotational speed, the difference between the electromagnetic torques of the accelerating and decelerating phases is calculated. The torque converted from mechanical friction and motor losses, as well as the numerically dominant quasi-steady-state aerodynamic torque component, cancel each other out because they are all single-valued functions of rotational speed. The remaining torque difference is contributed only by the flow lag caused by fluid inertia and is proportional to the difference in acceleration and deceleration angular acceleration. Normalizing this remaining difference by integrating the angular acceleration difference over the rotational speed domain yields an aerodynamic difference coefficient independent of excitation intensity. The quotient is then calculated by the square of the quasi-steady-state torque coefficient obtained by fitting the phase mean curve of the same set of dual data. The comprehensive quadratic drag coefficient of the pipeline network is mathematically canceled out, so that the final aerodynamic degradation index depends only on the pipeline network converted fluid inertia and the aerodynamic coupling coefficient and pressure generation coefficient of the fan body. Since the equivalent fluid inertia of a fixed-installation pipeline network can be considered constant after correction for gas density, changes in operating conditions such as damper opening adjustment, terminal damper operation, and filter bag dust accumulation, which only alter the pipeline resistance, will not cause a systematic shift in this index. However, changes in the aerodynamic coefficient of the fan body caused by blade scaling, wear, or corrosion will cause a repeatable shift in this index relative to the healthy baseline. By combining the synchronously output pipeline resistance auxiliary index, the two types of causes can be distinguished, reducing false alarms and false negatives of existing detection methods under these two operating conditions.

[0020] All the raw data required for testing is only the mechanical speed and electromagnetic torque feedback in the internal status register of the frequency converter drive controller. It can be read at high frequency via a standard industrial fieldbus. There is no need to install Pitot tubes, averaging pitot tubes, differential pressure transmitters or pressure sensor arrays on the process pipeline, nor is it necessary to open or cut the pipeline. This avoids the failure of monitoring function caused by pressure tap blockage or diaphragm damage. The detection logic is deployed in software form within the existing edge computing gateway, programmable logic controller or frequency converter extension logic unit, resulting in low hardware investment and subsequent maintenance workload.

[0021] The variable frequency acceleration / deceleration transition state data, previously excluded as interference, has become a source of features. This allows for the continuous acquisition of evaluation samples even in field conditions where the wind turbine is rarely in an absolutely steady state, improving the utilization rate of high-frequency acquisition data and the monitoring time coverage. In situations where the process is too smooth and lacks natural acceleration / deceleration, evaluation samples can also be actively acquired through small-amplitude speed disturbances that satisfy phase state determination constraints. Integral and normalization processing over the speed range suppresses the influence of random noise and differences in excitation intensity. Combined with healthy baseline learning and exponentially weighted moving average trend extraction, systematic shifts in the wind turbine's aerodynamic coefficients can be identified before degradation causes rotor mass imbalance and abnormal mechanical vibration, providing a basis for planned blade cleaning or surface repair. Attached Figure Description

[0022] Figure 1This is a schematic diagram of the architecture of the energy-saving variable frequency fan aerodynamic performance abnormality energy efficiency detection system provided in an embodiment of the present invention; Figure 2 This is a flowchart illustrating the energy efficiency detection method for abnormal aerodynamic performance of an energy-saving variable frequency fan provided in an embodiment of the present invention.

[0023] Figure reference numerals: 100, Industrial variable frequency fan unit; 110, Drive motor; 120, Fan impeller; 130, Industrial air duct network; 200, Variable frequency drive controller; 300, Intelligent aerodynamic performance detection controller; 301, High-frequency data acquisition and preprocessing module; 302, Motion phase identification and segmentation module; 303, Dual phase optimization and matching module; 304, Dynamic hysteresis loop construction module; 305, Area integration decoupling module; 306, Anomaly judgment and early warning module; 400, Upper-level monitoring system. Detailed Implementation

[0024] To make the objectives, technical solutions, and technical effects of the present invention clearer, the specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. The embodiments described herein are only for explaining the present invention and do not constitute a limitation on the scope of protection of the present invention.

[0025] like Figure 1 As shown, the energy-saving variable frequency fan aerodynamic performance abnormality energy efficiency detection system provided in this embodiment includes an industrial variable frequency fan unit 100, a variable frequency drive controller 200, an aerodynamic performance intelligent detection controller 300, and a host monitoring system 400. The industrial variable frequency fan unit 100 includes a drive motor 110 and a fan impeller 120 driven by the drive motor 110. The fan impeller 120 is installed in an industrial air duct network 130. The drive motor 110 can be a three-phase AC asynchronous motor or a permanent magnet synchronous motor; the drive motor 110 and the fan impeller 120 can be directly connected via a main shaft, or coupled via a coupling, pulley, or reducer. The fan impeller 120 can be centrifugal or axial. The industrial air duct network 130 has a defined length, cross-sectional area, and volume. It is generally equipped with components such as regulating dampers, dust collection bags, and terminal air valves. The operation of these components and the dust accumulation process will cause the overall resistance of the network to change during operation, which constitutes the main source of interference to the aerodynamic performance evaluation.

[0026] The variable frequency drive controller 200 is electrically connected to the stator winding of the drive motor 110, undertaking energy conversion and low-level motor control tasks. The variable frequency drive controller 200 employs a field-oriented control or direct torque control strategy. While regulating the current loop and speed loop, it can calculate in real time the instantaneous mechanical angular velocity, output electromagnetic torque, and stator excitation current and torque current components of the drive motor 110, mapping these values ​​to an internal status register that can be read externally. If the selected variable frequency drive controller 200 does not directly provide the electromagnetic torque value, it can be determined by its output torque current component and rotor flux linkage. The estimate can be obtained by subtracting the motor's electrical losses from the input active power and dividing by the mechanical angular velocity, where... This represents the number of pole pairs of the motor. This represents the rotor flux linkage amplitude. This represents the torque current component.

[0027] The aerodynamic performance intelligent detection controller 300 is the resident carrier of the algorithm logic in this embodiment. It can be an edge industrial control computer or edge computing gateway equipped with a real-time operating system, a programmable logic controller, or a programmable expansion logic unit embedded in the frequency converter drive controller 200. The aerodynamic performance intelligent detection controller 300 establishes a millisecond-level communication link with the frequency converter drive controller 200 via an industrial fieldbus. The communication protocol can be Modbus-TCP, Modbus-RTU, Profinet, or EtherCAT. As before... Figure 1 As shown, the aerodynamic performance intelligent detection controller 300 internally includes a high-frequency data acquisition and preprocessing module 301, a motion phase identification and segmentation module 302, a dual phase optimization and matching module 303, a dynamic hysteresis loop construction module 304, an area integration decoupling module 305, and an anomaly judgment and early warning module 306. Each module is implemented in software logic, and intermediate results are transmitted between them through a ring buffer and a shared data structure. The upper-level monitoring system 400 can be a distributed control system or a data acquisition and monitoring control system. It receives and presents the exponential trend data and alarm signals output by the aerodynamic performance intelligent detection controller 300, and archives historical data for long-term comparison.

[0028] Before explaining the specific actions of each module, it is necessary to establish a lumped-parameter dynamic model of the wind turbine unit under variable frequency transient conditions. This model is the basis for the subsequent algorithms. Let the electromagnetic torque output by the drive motor 110 be... The total moment of inertia of the drive motor 110 rotor, transmission chain, and fan impeller 120, referred to the motor shaft, is: The mechanical angular velocity is According to Newton's second law in rotational form, the dynamic equation of the motor shaft system is:

[0029] In the formula This is the sum of the frictional torque of the bearings and seals, the rotor wind-induced torque, and the loss torque of the drive motor 110 converted to the motor shaft due to mechanical additional losses. This loss term is determined by the contact friction and wind-induced torque mechanisms, exhibits the same value at the same rotational speed, and can be considered a single-valued function of the rotational speed, independent of the direction of rotational speed change. The direction-dependent residual component is a second-order small quantity and can be absorbed by subsequent baseline learning. (Pneumatic load torque) The reaction force generated by the work done by the impeller 120 on the fluid can be approximated by Euler's equations for fluid mechanics as follows:

[0030] in The instantaneous volumetric flow rate through the impeller 120 of the fan, The aerodynamic coupling coefficient, which comprehensively reflects the airfoil, installation angle, and blade surface roughness, is physically represented by the shaft torque corresponding to unit angular velocity and unit flow rate. The one-dimensional transient motion of fluid in the industrial duct network 130, after considering the fluid's own mass inertia and the resistance of the duct wall and fittings, can be described by the following lumped model.

[0031] in This is the reduced fluid inertia term related to the pipeline length, cross-sectional area, and gas density. The comprehensive quadratic resistance coefficient of the pipeline network varies slowly or abruptly with changes in damper opening, filter bag clogging, and other factors. This refers to the pressure generation coefficient of the wind turbine itself. When scale, wear, or corrosion occurs on the blades, the surface roughness of the blades increases, the effective working airfoil changes, the shaft torque required to deliver a unit flow rate increases, and the pressure generation capacity of the turbine itself decreases, which manifests as the aerodynamic coupling coefficient. Rise and pressure coefficient Decline, the ratio of the two That is, proportional to the aerodynamic efficiency of the fan, its monotonous decrease constitutes the object to be tested in this embodiment.

[0032] When the industrial variable frequency fan unit responds to the command of the proportional-integral-derivative control loop and enters the acceleration phase, i.e., angular acceleration... At that time, affected by the converted fluid inertia term Constraints, traffic The increase lags behind the rotational speed in time phase. Growth; entering a slowdown phase, that is... At the same time, the natural decay of the flow rate also lags behind the decrease in rotational speed. Therefore, at the same given rotational speed, the transient flow rate of the acceleration phase is less than that of the deceleration phase. The electromagnetic torque required to maintain the same rotational speed is not equal in the two phases. If the running trajectories of the acceleration and deceleration phases are plotted with mechanical angular velocity as the abscissa and electromagnetic torque as the ordinate, the two trajectories will form a dynamic aerodynamic hysteresis annular region with a certain area. The shape of this region carries information about the aerodynamic coefficient of the fan body and the pipeline resistance, which constitutes the carrier for extracting and separating the above information in this embodiment.

[0033] See Figure 2 The detection method executed by the aerodynamic performance intelligent detection controller 300 includes steps S101 to S106, each step is executed sequentially and repeatedly in the background.

[0034] Step S101 is executed by the high-frequency data acquisition and preprocessing module 301. The high-frequency data acquisition and preprocessing module 301 uses a fixed sampling frequency. The real-time status register of the variable frequency drive controller 200 is read cyclically, with a sampling frequency ranging from 50Hz to 200Hz. In this embodiment, 100Hz is preferred, resulting in a timing period of 10ms to fully preserve transient characteristics while satisfying the sampling theorem. The timing parameters acquired in each sampling period include a high-precision system timestamp. Original mechanical angular velocity With the original electromagnetic torque feedback Simultaneously, the intake air temperature and ambient pressure are read for gas density conversion. Considering the presence of pulse width modulation inverter noise and electromagnetic interference in the industrial environment, the original sequence contains a large number of high-frequency spikes. The high-frequency data acquisition and preprocessing module 301 applies a moving average filter or a first-order hysteresis discrete low-pass filter to it, with a filter time constant of 0.05s, and outputs a smoothed state sequence. and The data is then written to a circular buffer. Simultaneously, the high-frequency data acquisition and preprocessing module 301 performs data validity checks, marking and removing records with communication timeouts, range out-of-bounds errors, or timestamp jumps to prevent anomalies from contaminating subsequent derivative and interpolation calculations.

[0035] Step S102 is executed by the motion phase recognition and segmentation module 302. The motion phase recognition and segmentation module 302 processes the smoothed rotational speed sequence. The first-order smoothed derivative is calculated using the five-point difference method or the Savitzky-Golay filter to obtain the real-time angular acceleration. Then, based on its sign and amplitude, the continuous data stream is dynamically segmented into different motion phases and stored in independently allocated memory queues. The phase determination threshold is not based on the acceleration / deceleration time setting of the frequency converter drive controller 200, because this setting can be modified by on-site personnel at any time and does not have physical inherentness; this embodiment uses the inherent no-load limit angular acceleration of the equipment. As a benchmark, among which The rated torque of drive motor 110, For the aforementioned calculated total moment of inertia, the effective acceleration threshold is taken as... proportionality coefficient The percentage is selected between 5% and 20%, and in this embodiment, it is 10%. For Approximately 15 rad / s 2 The unit, That is 1.5 rad / s 2 .when Furthermore, the duration of this state exceeds the minimum duration window. When it is determined to be the acceleration phase, when And lasting longer than When the time is determined to be a deceleration phase, the effective deceleration threshold is... The way of taking values ​​and same, Take 2 seconds. To ensure that the angular acceleration sign discrimination is not drowned out by measurement noise, the above threshold must also be greater than 3 times the noise standard deviation of the estimated angular acceleration value within this speed range. In practice, this standard deviation can be statistically analyzed during the constant speed operation period and used to verify the lower limit of the threshold. The remaining time periods, including small fluctuations in speed and long-term constant speed operation, are marked as steady-state phases or disordered fluctuation phases. In this embodiment, steady-state phase data does not participate in the extraction of hysteresis features, but is only used for communication link self-testing, noise standard deviation statistics, and rotational inertia verification. The acceleration and deceleration transition state data that are rejected by traditional methods become the main source of features.

[0036] Step S103 is executed by the dual phase state optimization matching module 303. The frequency conversion control commands issued by the industrial closed-loop control loop are typically random and asymmetrical. To construct a comparison interval with equivalent conditions, the dual phase state optimization matching module 303 pre-sets an effective analysis speed window that reflects the main working range of the fan. In this embodiment, the speed range is set to 40% to 80% of the rated full-load speed. This avoids the region with significant static friction and nonlinear effects at extremely low speeds while retaining the high-speed power delivery region. The dual phase optimization matching module 303 independently searches the historical phase cache queue for acceleration and deceleration phase timings within a set optimization time span, which can be 24 hours. When the speed ranges covered by both an acceleration phase timing sequence and a deceleration phase timing sequence are fully included... When the relative difference in the absolute value of the average angular acceleration between the two phases within a window is less than a preset threshold, they are paired and defined as a valid dual analysis group. Since the subsequent normalization process has eliminated the first-order influence of the excitation intensity, this matching threshold can be relaxed to 100%. The purpose of retaining this threshold is to ensure that both phase states are within the quasi-steady-state applicable range and have consistent rotational speed coverage. The dual phase state optimization matching module 303 also checks the sampling point density of each time series within the integration window. When there are fewer than 5 valid sampling points within every 10 rad / s rotational speed span, the phase state is determined to be insufficient to support the interpolation operation and is discarded.

[0037] If the process at a specific site is extremely stable, and there is a prolonged lack of natural deep acceleration and deceleration conditions, making it impossible to collect effective dual analysis data, then the aerodynamic performance intelligent detection controller 300 can superimpose an ultra-low frequency sinusoidal disturbance command onto the speed setpoint of the frequency converter drive controller 200 via its control interface. The disturbance form is as follows: ,in The current process speed, For the disturbance amplitude, The period is the disturbance period. The maximum angular acceleration that this disturbance can produce is... Therefore, the perturbation parameter must be obtained in conjunction with the phase determination threshold in step S102 to satisfy the following conditions: Margin coefficient The value should not be less than 1.5; otherwise, the transient state data generated by the active excitation will be judged as a steady-state phase by the phase state identification stage and discarded entirely. Taking a unit with a rated speed of 314 rad / s as an example, we take... This is 3% of the rated speed, which is 9.4 rad / s. If we take 20 seconds, then Approximately 2.96 rad / s 2 1.5 rad / s 2 1.97 times the threshold, and within a single half-cycle The duration exceeding the threshold was approximately 6.6 seconds, which is greater than the minimum duration window. The segmentation conditions are met. When the process only allows for smaller amplitudes or longer periods, the phase determination threshold under the excitation mode should be lowered accordingly. , and according to Retrieve the value while ensuring It is still greater than three times the standard deviation of angular acceleration noise. It should also be noted that the rotational speed range swept by the micro-amplitude disturbance is much smaller than the analysis window in passive capture mode. Therefore, in active excitation mode, the integration interval no longer uses 40% to 80% of the rated speed, but is reconstructed by the dual phase optimization matching module 303 based on the actual achieved rotational speed extreme value. Alternatively, the measured extreme value can be reduced by 5% within the specified interval, and the sampling point density can be checked within this interval. Systematic biases introduced by different integration intervals are eliminated by maintaining independent healthy baselines for both active and passive capture modes. The aforementioned amplitude levels are within the disturbance tolerance margin of most processes and do not affect production continuity. Active and passive capture modes can coexist and be switched according to priority; active excitation is only activated if a valid dual analysis group is not obtained for a set period of time.

[0038] Step S104 is executed by the dynamic hysteresis loop construction module 304. The original data in the effective duality analysis group are all in discrete time... To record the independent variable and perform torque calculations on the exact same speed scale, a coordinate transformation and resampling from the time domain to the speed domain must be performed. The dynamic hysteresis loop construction module 304 pairs the acceleration phase data set. With strictly monotonically increasing speed By applying cubic spline interpolation to the independent variable, a continuous and smooth acceleration phase torque function is constructed. With acceleration phase angle function For the deceleration phase data set, an equivalent processing method is performed with strictly monotonically decreasing rotational speed as the independent variable, and the coordinates are reversed to obtain the deceleration phase torque function. With deceleration phase angle acceleration function After interpolation, the two torque curves form a distinct envelope within the integration interval due to hydrodynamic hysteresis, i.e., the aforementioned dynamic aerodynamic hysteresis annular region. The dynamic hysteresis loop construction module 304 simultaneously constructs the mean curves of the two phase states. and The mean curve is used in subsequent steps to extract quasi-steady-state operating characteristics. The contour data and mean curve can be uploaded to the upper-level monitoring system 400 for graphical display, facilitating comparison of the ring shape at different times by maintenance personnel.

[0039] Step S105 is executed by the area integral decoupling module 305 and is the core step in this embodiment to achieve the separation of physical quantities. The shaft system dynamics equations are applied at arbitrary rotational speeds. Expanding by phase state, we can obtain the expression for the accelerating phase torque. With deceleration phase torque expression .remember , , After subtracting the two equations, the loss torque term is eliminated because it is a single-valued function of the rotational speed, resulting in:

[0040] Flow lag The difference is obtained from the lumped equation of the pipeline network at the same rotational speed. By writing out the equation in the acceleration and deceleration phases respectively and subtracting them, we can obtain... The pressure terms cancel each other out due to the same rotational speed. (Using the quasi-steady-state solution...) As a zeroth-order approximation, that is, and After substituting and organizing, we get

[0041] Its negative value is consistent with the physical fact that the flow rate in the accelerating phase lags behind that in the decelerating phase. Substituting it back into the torque difference expression, we obtain...

[0042] Therefore, the torque difference and the angular acceleration difference are proportional throughout the entire integration interval, and the second term in the proportionality coefficient represents the aerodynamic information contributed by the fluid inertial hysteresis. The area integration decoupling module 305 uses a numerical integration algorithm to calculate the area of ​​the hysteresis torque and the area of ​​the angular acceleration difference, respectively.

[0043] Numerical integration can be performed using the trapezoidal integration method, Simpson's rule, or equal-step Riemann summation, with an integration step size of 0.5 rad / s. The aerodynamic difference coefficient is obtained by normalizing the hysteresis torque area to the area of ​​the angular acceleration difference and subtracting the contribution of the rotor's rigid mechanical inertia.

[0044] This coefficient is independent of the severity of acceleration and deceleration; that is, although the areas of the hysteresis loops formed by different equipment operations vary, their normalization results remain consistent. To improve noise immunity, the area integral decoupling module 305 can also avoid direct quotient calculation and instead use the least squares method to... right Perform a linear regression and take the regression slope as the equivalent inertia. Press again In terms of calculation, the two are equivalent in a model sense, and the correlation coefficient of the regression can also be used as a criterion for data validity.

[0045] The aerodynamic differential coefficient is inversely proportional to the quadratic drag coefficient of the pipeline network, therefore it cannot be used alone as a health criterion for the wind turbine itself. To eliminate this dependence, the area integral decoupling module 305 extracts quasi-steady-state operating characteristics using the mean curve of the same set of dual data. Since the flow deviations between the acceleration and deceleration phases are first-order canceled out in the mean, the mean curve satisfies... , and in the integration interval A three-parameter least squares fit is performed using the form of the equation, where the first-order term and the constant term are used to absorb the variation of loss torque with speed. The coefficients of the resulting quadratic term are the quasi-steady-state torque coefficients.

[0046] The aerodynamic degradation index output in this embodiment is obtained by quoting the square of the quasi-steady-state torque coefficient with respect to the aerodynamic difference coefficient and correcting for the gas density ratio.

[0047] In the formula The current gas density is calculated from the measured intake air temperature and ambient pressure. The average gas density obtained during the baseline learning period serves as a reference state. The reduced fluid inertia term, pipeline resistance coefficient, pressure generation coefficient, and aerodynamic coupling coefficient are all proportional to the gas density. After synthesis, these indices are inversely proportional to the density; therefore, multiplying by the density ratio eliminates the influence of seasonal and operational fluctuations in medium properties. In this expression, the comprehensive quadratic pipeline resistance coefficient has been completely canceled out, and the reduced fluid inertia term is constant for the fixed-installation industrial duct network 130 after density correction. Therefore, the aerodynamic degradation index becomes a quantity determined solely by the aerodynamic coefficient of the fan impeller 120 itself. In other words, while adjustments to damper opening, terminal valve operation, or dust accumulation in the filter bag may cause significant changes in the pipeline resistance coefficient, they do not cause a systematic shift in this index. Only changes in the aerodynamic coefficient of the fan impeller 120 itself cause it to deviate from the healthy baseline. To facilitate causal explanation, the area integral decoupling module 305 simultaneously outputs auxiliary pipeline resistance indices. This index increases with the increase of pipeline resistance, mainly reflecting the resistance level on the pipeline side. However, it also includes the fan pressure coefficient, so it is only used as a working condition label and auxiliary explanation. The criterion for the degradation of the fan body is always based on the aerodynamic degradation index.

[0048] The above derivation is contingent upon several conditions, which must be guaranteed during implementation. First, within the timeframe occupied by a single dual analysis, the pipeline resistance coefficient, pressure generation coefficient, aerodynamic coupling coefficient, and equivalent fluid inertia term are all considered constant. Therefore, the optimization timeframe should avoid crossing known large-scale operating condition switching events. If necessary, the phase data can be labeled with operating conditions using the valve position signal from the 400-level monitoring system before pairing. Second, the loss torque is a single-valued function of the rotational speed. Third, the acceleration / deceleration timescale is significantly larger than the fluid inertia time constant. To ensure the effectiveness of the quasi-steady-state zeroth-order approximation, the amplitude of the angular acceleration involved in the analysis can be constrained to not exceed [a certain value]. And verify after interpolation is completed. and The linear correlation coefficient is used; if the correlation coefficient is below 0.9, the dual analysis group is considered not to meet the quasi-steady-state condition and is therefore discarded. The total moment of inertia is calculated. It can be obtained from the equipment nameplate, or identified through no-load speed-up and free-stop tests under conditions of fan impeller 120 disengagement or industrial air duct network 130 being fully open; because the aerodynamic difference coefficient is derived from... The difference between the inertia and the equivalent inertia is obtained. The error in the value will introduce a fixed bias to the aerodynamic degradation index, but this bias remains constant throughout the equipment's lifespan and can be completely absorbed by the health baseline, without affecting the determination of relative deviation and trend.

[0049] Step S106 is executed by the anomaly detection and early warning module 306. To adapt to the initial physical differences of wind turbines of different models and under different installation conditions, the anomaly detection and early warning module 306 adopts an adaptive unsupervised strategy to construct an alarm baseline. After the wind turbine is initially installed and put into operation, or after high-pressure water cleaning, sandblasting repair, or major overhaul of the blades, the anomaly detection and early warning module 306 enters baseline learning mode. The learning period is generally 3 to 7 days, during which time it automatically collects previous data. Calculated aerodynamic degradation index values ​​corresponding to each effective pair analysis group. Take 100 and calculate its expected mean using statistical methods. with standard deviation ,Will Record as a health benchmark Simultaneously, the average gas density during this period was recorded as a reference density. , This is used to eliminate gross errors and calculate the statistical confidence level of the alarm threshold. After learning is complete, the anomaly detection and early warning module 306 switches to long-term silent monitoring mode, and calculates the current period index each time a new effective dual analysis group is obtained. At that time, according to The deviation rate relative to the healthy baseline is calculated. Since blade dust accumulation, chemical corrosion, and surface wear are all extremely slow, gradual processes, the anomaly detection and early warning module 306 does not employ a single-out-of-bounds mechanism with poor noise immunity. Instead, it sends the deviation rate to an exponentially weighted moving average filter, according to... Recursion, weighting coefficients Take 0.05 as the initial value. To avoid underreporting due to initial value bias, when the number of valid samples is insufficient... That is, the filter output for the first 60 groups will not participate in alarm determination, but will only be used for trend recording. When the filter output exceeds the preset degradation sensitivity threshold... And recently When the linear regression slope of the group index is significantly non-zero and consistent with the direction of deviation, the anomaly judgment and early warning module 306 outputs an alarm signal for early degradation of the aerodynamic performance of the wind turbine blades to the upper-level monitoring system 400 via the industrial control protocol. At the same time, it reports the current index, health benchmark, pipeline resistance auxiliary indicators and trend curves. The alarm text can prompt the investigation of blade scaling, wear or corrosion. Take 50, A threshold of 0.12 is used, meaning an index deviation of 12% from the healthy baseline is used as the alarm condition. This threshold can be adjusted between 0.05 and 0.20 based on the fan capacity level, media dust concentration, and operation and maintenance strategy. The quantitative correspondence between the index deviation and the actual aerodynamic efficiency decline can be calibrated by comparing the index of the same equipment before and after the first blade cleaning and then solidified in the system parameters. The direction of the deviation is also determined in this calibration. Thereafter, only deviations consistent with the calibration direction will trigger an alarm. If the pipeline resistance auxiliary index changes significantly and synchronously while the aerodynamic degradation index does not deviate significantly, the anomaly judgment and early warning module 306 determines it as a change in pipeline resistance and only outputs a condition change prompt to the upper-level monitoring system 400 without triggering a fan degradation alarm. This further distinguishes between the two causes at the output level.

[0050] At the program implementation level, the aerodynamic performance intelligent detection controller 300 uses a timer interrupt to drive the main control event loop, with an interrupt period of 10ms, i.e., a loop frequency of 100Hz. During the system initialization phase, hyperparameters such as the converted total moment of inertia, rated torque, phase determination thresholds calculated from the two, integration interval boundaries, integration step size, weighting coefficients, alarm thresholds, and excitation parameters are loaded. The initial values ​​of the health baseline and filter output are set to empty or zero to trigger baseline learning. Within each loop cycle, the timestamp acquisition, register reading, low-pass filtering, derivative calculation, and phase determination are completed sequentially. When the angular acceleration is greater than the acceleration threshold, the timestamp, speed, torque, and angular acceleration quadruple is appended to the acceleration phase annular buffer; when it is less than the deceleration threshold (negative value), it is appended to the deceleration phase annular buffer. When the angular acceleration falls into the dead zone, i.e., the rotational speed tends to stabilize, it is considered that an acceleration / deceleration process has just ended. At this time, the duality validity check function is called. If the data in the acceleration phase annular buffer and the deceleration phase annular buffer meet the matching conditions of rotational speed coverage, sampling density, and excitation intensity, then four rotational speed domain spline interpolation objects and two mean curves are constructed sequentially. In the equal-step-size loop, the hysteresis torque area and the angular acceleration difference area are accumulated to check their linear correlation. The aerodynamic difference coefficient, quasi-steady-state torque coefficient, aerodynamic degradation index, and pipeline resistance auxiliary index are calculated. Then, the baseline or propulsion deviation rate and exponentially weighted moving average are updated, and an alarm judgment is executed. After completion, the processed phase buffer is released and cleared to avoid memory bloat and data reuse. If a valid duality analysis group is not obtained for more than a set time, the main control loop switches to the active excitation branch, issues a sinusoidal disturbance instruction according to the aforementioned constraints, and reconstructs the integral interval. The above logic can be compiled into embedded firmware in C or C++ and run on a microcontroller, or implemented in a scripting language on an edge gateway. Its computational load on a general-purpose industrial processor is far lower than the real-time constraints.

[0051] Without altering the core concept described above, this embodiment also allows for several variations and equivalent substitutions. The filtering stage used in the high-frequency data acquisition and preprocessing module 301 can be replaced by moving average filtering and first-order low-pass filtering with Kalman filtering or median filtering; the derivative estimation of the motion phase identification and segmentation module 302 can be replaced by slope estimation using least squares local fitting, in addition to five-point difference and Savitzky-Golay filtering; the cubic spline interpolation of the dynamic hysteresis loop construction module 304 can be replaced by piecewise linear interpolation, B-spline interpolation, or Gaussian process regression. In cases of dense sampling points, piecewise linear interpolation can meet the accuracy requirements; the numerical integration method of the area integral decoupling module 305 can be freely selected from trapezoidal method, Simpson method, and Riemann summation. In addition to fitting the mean curve, the extraction of the quasi-steady-state torque coefficient can also be obtained by fitting the torque and speed data of adjacent constant speed periods according to the same quadratic relationship. In this case, the consistency between the period and the dual phase in the working condition label should be checked; the trend extraction of the anomaly judgment and early warning module 306 can be replaced by cumulative sum control chart or sliding window linear regression, in addition to exponential weighted moving average. In cases where pipeline resistance remains stable over a long period and only rough trend monitoring is required, the step of using the quasi-steady-state torque coefficient as the quotient can be omitted. Instead, the inertia term can be directly subtracted from the hysteresis torque area before proceeding. The quantity obtained by normalization is used as a simplifying exponent, and its closed denominator is... This simplified form is easier to implement, but its ability to suppress changes in pipeline resistance is weaker than the aforementioned complete form. The effective analysis speed window boundary is not limited to 40% and 80% of the rated speed. It can be set separately according to the actual operating range of the fan and the resonant speed restriction zone. If necessary, multiple non-overlapping speed sub-windows can be set on the same equipment and baselines can be maintained separately to obtain the degradation characteristics along the speed distribution. When multiple fans are operating in parallel on the same industrial air duct network 130, steps S101 to S106 can be executed independently for the variable frequency drive controller 200 to which each fan belongs, thereby locating the degradation to a single fan. The function of the aerodynamic performance intelligent detection controller 300 can also be moved to a factory-level server or cloud platform. In this case, the high-frequency data acquisition and preprocessing module 301 is retained on the edge side to complete sampling and filtering, while the other modules are executed remotely. The two transmit phase fragment data through the network. In addition to industrial fans, the above method of normalizing the torque difference between acceleration and deceleration dual phases and decoupling resistance is also applicable to blowers, induced draft fans and other variable frequency driven impeller fluid machinery that also have fluid inertial hysteresis.

[0052] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for detecting abnormal aerodynamic performance of an energy-saving variable frequency fan, wherein the method is executed by a variable frequency drive controller or an intelligent aerodynamic performance detection controller communicatively connected to the variable frequency drive controller, characterized in that, Includes the following steps: Continuous time-series status data of industrial variable frequency fan units during operation is obtained, and the status data includes at least the mechanical angular velocity and electromagnetic torque of the drive motor; The angular acceleration is calculated based on the mechanical angular velocity, and the continuous time-series state data is segmented into motion phases according to a preset phase determination threshold. The acceleration phase time-series data and deceleration phase time-series data are identified and extracted. Within the set optimization time span, matching acceleration phase time series data and deceleration phase time series data covering the same effective analysis speed window are used to construct an effective dual analysis group; The time-domain data in the effective dual analysis group is converted to the speed domain, and the interpolated torque function and interpolated angular acceleration function of the acceleration phase and deceleration phase at the same speed scale are constructed respectively. The mean torque curve and mean angular acceleration curve of the two phase states are obtained. Within the effective analysis speed window, the integral of the torque difference between the interpolated torque functions of the two phases and the integral of the angular acceleration difference between the interpolated angular acceleration functions are calculated respectively. The aerodynamic difference coefficient is obtained by subtracting the quotient of the integral of the torque difference and the integral of the angular acceleration difference from the calculated total moment of inertia. The mean torque curve and the mean angular acceleration curve are fitted together to extract the quasi-steady-state torque coefficient; The aerodynamic degradation index is calculated based on the ratio of the aerodynamic difference coefficient to the square of the quasi-steady-state torque coefficient, and an aerodynamic performance abnormality alarm is triggered when the aerodynamic degradation index deviates systematically from the healthy benchmark.

2. The energy efficiency detection method for abnormal aerodynamic performance of an energy-saving variable frequency fan according to claim 1, characterized in that, The phase state determination threshold is determined by the inherent no-load limit angular acceleration of the industrial variable frequency fan unit combined with a proportional coefficient. The no-load limit angular acceleration is the ratio of the rated torque of the drive motor to the converted total moment of inertia. If the effective dual analysis group is not obtained for more than a set time, the speed setpoint of the variable frequency drive controller is superimposed with a low-frequency sinusoidal disturbance command for active excitation. The maximum angular acceleration generated by the sinusoidal disturbance command is greater than the phase determination threshold, and the effective analysis speed window is reconstructed from the actual speed extreme value reached.

3. The energy efficiency detection method for abnormal aerodynamic performance of an energy-saving variable frequency fan according to claim 1, characterized in that, The process of converting the time-domain data in the effective dual analysis group to the rotational speed domain includes: Spline interpolation is performed on the acceleration phase time series data with monotonically increasing mechanical angular velocity as the independent variable to generate smooth acceleration phase interpolated torque function and acceleration phase interpolated angular acceleration function; Using the monotonically decreasing mechanical angular velocity as the independent variable, equivalent interpolation is performed on the deceleration phase time series data, and coordinate inversion is performed to generate a smooth deceleration phase interpolated torque function and a deceleration phase interpolated angular acceleration function. The average torque curve is formed by adding the interpolated torque functions of the acceleration phase and the deceleration phase, and the average angular acceleration curve is formed by adding the interpolated angular acceleration functions of the acceleration phase and the deceleration phase.

4. The energy efficiency detection method for abnormal aerodynamic performance of an energy-saving variable frequency fan according to claim 1, characterized in that, The calculation process of the aerodynamic difference coefficient includes: The definite integral of the difference between the interpolated torque functions of the acceleration phase and the deceleration phase within the effective analysis speed window is calculated to obtain the hysteresis torque area; The definite integral of the difference between the interpolated angular acceleration functions of the acceleration phase and the deceleration phase within the effective analysis speed window is calculated to obtain the area of ​​the angular acceleration difference. The aerodynamic difference coefficient is obtained by subtracting the quotient of the hysteresis torque area and the angular acceleration difference area from the calculated total moment of inertia. The aerodynamic difference coefficient characterizes the aerodynamic information contributed by the flow hysteresis caused by fluid inertia.

5. The energy efficiency detection method for abnormal aerodynamic performance of an energy-saving variable frequency fan according to claim 1, characterized in that, The quasi-steady-state torque coefficient is determined by the difference between the product of the torque mean curve and the product of the reduced total moment of inertia and the angular acceleration mean curve, and the quadratic term coefficient obtained by performing a least-squares fitting of a quadratic polynomial within the effective analysis speed window. The linear term and constant term in the quadratic polynomial are used to absorb the changes in the inherent mechanical losses of the equipment with the rotational speed. The calculation and alarm triggering process of the aerodynamic degradation index includes: The aerodynamic degradation index for the current cycle is obtained by dividing the aerodynamic difference coefficient by the square of the quasi-steady-state torque coefficient and multiplying it by the ratio of the current measured gas density to the reference density. The deviation rate of the aerodynamic degradation index relative to the preset health benchmark is obtained, and the deviation rate is input into the exponential weighted moving average filter to extract the long-term evolution trend. When the long-term evolution trend after filtering exceeds the degradation sensitivity threshold, and the linear regression slope of the recent trend is not zero and the direction is consistent, an alarm signal for early degradation of blade aerodynamic performance is output.

6. The energy efficiency detection method for abnormal aerodynamic performance of an energy-saving variable frequency fan according to claim 1, characterized in that, After constructing the effective dual analysis group and before calculating the aerodynamic difference coefficient, the linear correlation coefficient between the torque difference and the angular acceleration difference within the effective analysis speed window is also checked. When the linear correlation coefficient is lower than the preset lower limit, the effective dual analysis group is determined to not meet the quasi-steady-state condition and is discarded. The pipeline resistance auxiliary index is also obtained based on the ratio of the quasi-steady-state torque coefficient to the aerodynamic differential coefficient. When the pipeline resistance auxiliary index deviates but the aerodynamic degradation index does not deviate systematically, a working condition change prompt is output without triggering an aerodynamic performance abnormality alarm.

7. An energy-saving variable frequency fan aerodynamic performance abnormality detection system, used to implement the energy-saving variable frequency fan aerodynamic performance abnormality detection method according to any one of claims 1 to 6, characterized in that, It includes a variable frequency drive controller and an intelligent aerodynamic performance detection controller that are connected in sequence; the intelligent aerodynamic performance detection controller is equipped with: The high-frequency data acquisition and preprocessing module is used to read the mechanical angular velocity and electromagnetic torque from the variable frequency drive controller at high frequency, and to perform filtering and validity verification. The motion phase identification and segmentation module is used to calculate angular acceleration and dynamically segment continuously maintained acceleration phase time series data and deceleration phase time series data according to the sign of angular acceleration and the phase determination threshold. The dual phase state optimization matching module is used to pair acceleration phases and deceleration phases that cover the same effective analysis speed window and whose average angular acceleration differences meet the conditions when independently searching historical phase states as effective dual analysis groups. The dynamic hysteresis loop construction module is used to map the effective dual analysis group to the speed domain and generate the interpolated torque function, the interpolated angular acceleration function, and the phase mean curve; The area integral decoupling module is used to calculate the integral of torque difference and angular acceleration difference between two phase states. Based on their ratio and the converted total moment of inertia, the aerodynamic difference coefficient is extracted, and the aerodynamic degradation index is calculated by combining the quasi-steady-state torque coefficient obtained by fitting the mean curve. The anomaly detection and early warning module is used to monitor the deviation trend of the aerodynamic degradation index relative to the healthy benchmark, trigger an alarm signal, and output it to the upper-level monitoring system.

8. The energy-saving variable frequency fan aerodynamic performance abnormality detection system according to claim 7, characterized in that, The phase state determination threshold used in the motion phase state identification and segmentation module is the unit's no-load limit angular acceleration multiplied by a proportional coefficient; the dual phase state optimization matching module is also configured to send a superimposed low-frequency sinusoidal speed disturbance command to the frequency converter drive controller to actively acquire transition state data when no effective dual analysis group is obtained for a long time under natural operating conditions, and reconstruct the effective analysis speed window based on the actual speed extreme value reached by the disturbance.

9. The energy-saving variable frequency fan aerodynamic performance abnormality detection system according to claim 7, characterized in that, The dynamic hysteresis loop construction module uses a cubic spline interpolation algorithm to process the original time domain data, so that under a unified speed scale, the acceleration phase torque curve and the deceleration phase torque curve form a dynamic aerodynamic hysteresis loop region with an enclosing area.

10. The energy-saving variable frequency fan aerodynamic performance abnormality detection system according to claim 7, characterized in that, The area integral decoupling module calculates the numerical integral ratio or linear regression slope of the torque difference relative to the angular acceleration difference, and subtracts the torque consumed by the rigid mechanical components in the form of equivalent inertia, thus separating the characteristic quantity jointly determined by the aerodynamic coefficient of the fan body and the fluid inertia.

11. The energy-saving variable frequency fan aerodynamic performance abnormality detection system according to claim 7, characterized in that, The anomaly detection and early warning module is pre-set with an adaptive unsupervised baseline learning logic, which is used to statistically analyze the expected mean of the aerodynamic degradation index corresponding to a set number of effective dual analysis groups after the initial commissioning of the equipment or after blade repair to establish the health benchmark, and to track the degradation deviation using an exponentially weighted moving average in silent monitoring mode.

Images