Method and system for energy consumption optimization of an electric actuator

Abstract

This application provides a method and system for optimizing the energy consumption of an electric actuator, relating to the field of intelligent control technology. The method includes acquiring valve body fluid property description information and actuator operation observation data, including stroke position, motor electrical parameters, and medium pressure difference; dividing the observation data into a preset baseline operating condition dataset and a preset pressurized operating condition dataset based on the medium pressure difference; establishing a correspondence between stroke position and electrical parameters from the baseline operating condition data to obtain mechanical baseline resistance characteristics; determining stroke load characteristics from the pressurized operating condition data and in combination with the mechanical baseline resistance characteristics; determining feedforward compensation control quantities and generating motor control commands according to preset energy consumption evaluation criteria based on the valve body fluid property description information and stroke load characteristics to achieve opening adjustment; and updating the above characteristics or parameters based on motion feedback data.

Description

Technical Field

[0001] This application relates to the field of intelligent control technology, and more specifically, to a method and system for optimizing the energy consumption of electric actuators. Background Technology

[0002] Electric actuators are widely used in the opening regulation and shut-off control of valves such as gate valves, globe valves, and control valves. Especially in multi-turn, high-torque scenarios, the actuator converts the motor output into valve stem displacement and valve core force through multi-stage reduction transmission. Its driving force is affected not only by mechanical factors such as transmission chain friction, hysteresis, and sealing packing preload, but also by the hydrodynamic torque caused by changes in medium density, viscosity, and pressure difference before and after the valve. In actual operation, the working conditions often exhibit characteristics such as pressure difference and flow state fluctuations over time, and significant nonlinear load when the valve is in different opening positions. At the same time, the worm gear pair, threaded pair, and bearings of the actuator will experience resistance drift due to wear, temperature, and changes in lubrication conditions, resulting in unstable motor electrical parameter response at the same opening position.

[0003] Existing control methods mostly adopt closed-loop regulation based on stroke or electrical parameters, and can be superimposed with empirical feedforward or fixed torque / current limiting strategies. However, the above methods usually cannot distinguish between "mechanical resistance background" and "medium load contribution" in the stroke dimension. The coupled effects of pressure difference change and mechanical wear often rely on manual tuning or conservative margin, which can easily lead to problems such as dynamic response lag, overshoot oscillation or drive redundancy, resulting in unnecessary energy consumption increase and accelerated component degradation.

[0004] The resulting technical challenge lies in how to achieve precise, on-demand allocation of the drive output of electric actuators in a complex environment where variable fluid conditions and mechanical wear coexist, in order to eliminate dynamic response lag and reduce system redundancy energy consumption. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this application provides a method and system for optimizing the energy consumption of electric actuators.

[0006] In a first aspect, this application provides a method for optimizing the energy consumption of an electric actuator, including: Obtain valve body fluid property description information and actuator operation observation data, including stroke position data, motor electrical parameter data and medium pressure difference data; Based on the medium pressure difference data, determine the observation dataset corresponding to the preset baseline operating condition and the observation dataset corresponding to the preset pressurized operating condition; Based on the observation dataset corresponding to the preset baseline working condition, a correspondence between the stroke position data and the motor electrical parameter data is established to obtain the mechanical baseline resistance characteristics associated with the stroke position. Based on the observation dataset corresponding to the preset pressurized working condition and the mechanical baseline resistance characteristics, the stroke load characteristics associated with the stroke position are determined; Based on the valve body fluid property description information and the stroke load characteristics, the feedforward compensation control quantity is determined according to the preset energy consumption evaluation criteria, and a motor control command is generated based on the feedforward compensation control quantity to control the actuator to change from the current opening degree to the target opening degree. Acquire motion feedback data, and update at least one of the generation parameters of the mechanical baseline resistance characteristic, the stroke load characteristic, and the feedforward compensation control quantity based on the motion feedback data.

[0007] Optionally, the preset baseline operating condition is an operating condition where the medium pressure difference is less than or equal to the baseline threshold.

[0008] Optionally, the preset pressurized operating condition is an operating condition where the medium pressure difference is greater than the baseline threshold, and the stroke load characteristic is determined by the deviation of the motor electrical parameter data under the preset pressurized operating condition from the mechanical baseline resistance characteristic.

[0009] Optionally, the preset energy consumption evaluation criterion evaluates the feedforward compensation control quantity based on the energy consumption index determined by the motor electrical parameter data within a preset action time window; wherein the feedforward compensation control quantity is determined from at least one candidate feedforward compensation control quantity.

[0010] Optionally, determining the observation dataset corresponding to the preset baseline operating condition and the observation dataset corresponding to the preset pressurized operating condition includes: Smoothing is performed on the differential pressure data of the medium, and the fluctuation index and rate of change index of the differential pressure of the medium are calculated within a preset time window to determine the stable differential pressure range that meets the stability criterion. In the pressure differential stability zone, a running segment with a medium pressure differential less than or equal to the baseline threshold is selected, and the stroke position data and motor electrical parameter data synchronously collected in the running segment are aggregated into an observation dataset corresponding to the preset baseline operating condition. In the pressure differential stability zone, a running segment with a medium pressure differential greater than the baseline threshold is selected, and the stroke position data and motor electrical parameter data synchronously collected within the running segment are aggregated into an observation dataset corresponding to the preset pressurized working condition. The differential pressure transition segment that crosses the baseline threshold is removed from the operating segment.

[0011] Optionally, before aggregating the observation datasets corresponding to the preset baseline operating condition and / or the observation datasets corresponding to the preset pressurized operating condition, the method further includes: The entire stroke is discretized into multiple opening interval blocks according to the preset stroke step size, and the opening interval blocks corresponding to the stroke positions in the opening and closing directions are combined to form a bidirectional stroke unit. Based on the travel end distance of the bidirectional stroke unit and the rate of change of the motor electrical parameters, the valve structure stage index to which each bidirectional stroke unit belongs is determined. The valve structure stage index includes a disengagement index, an adjustment index, and a re-engagement index. Only the stroke position data and motor electrical parameter data corresponding to the bidirectional stroke units with valve structure stage index as adjustment index are selected for the collection of the observation dataset, and the data corresponding to the bidirectional stroke units with valve structure stage index as disengagement index or in-seat index are removed.

[0012] Optionally, the step of establishing a correspondence between stroke position data and motor electrical parameter data based on the observation dataset corresponding to the preset baseline operating condition, and obtaining the mechanical baseline resistance characteristics associated with the stroke position, includes: The travel change direction identifier is determined based on the difference sign and / or change gradient sign of the travel position data, and the observation dataset corresponding to the preset baseline condition is split into an open baseline subset and a closed baseline subset according to the travel change direction identifier. The entire stroke is divided into multiple opening interval blocks according to a preset stroke resolution. For the opening baseline subset and the closing baseline subset, motor electrical parameter data samples are collected in each of the opening interval blocks respectively. Within each of the aforementioned opening interval blocks, a baseline representative value is determined for the motor electrical parameter data samples; Based on the travel position of each opening interval block and the corresponding baseline representative value, open-direction mechanical baseline resistance curves and closed-direction mechanical baseline resistance curves are generated respectively, and the open-direction mechanical baseline resistance curves and the closed-direction mechanical baseline resistance curves are combined to form a direction-sensitive mechanical baseline resistance feature.

[0013] Optionally, before determining the baseline representative value, the method further includes performing an equivalent conversion on the observed data using the structural parameters of the multi-turn drive train and removing backlash data, including: The actuator is a multi-turn actuator, and its transmission chain includes a worm gear transmission pair and a valve stem thread transmission pair; obtain the transmission parameter set corresponding to the worm gear transmission pair and the valve stem thread transmission pair, the transmission parameter set including reduction ratio parameter and thread lead parameter; Based on the transmission parameter set, the equivalent conversion of the motor electrical parameter data is performed to obtain the valve stem side equivalent resistance characterization sequence. The equivalent conversion includes converting the motor output represented by the motor electrical parameter data into the valve stem side equivalent torque or equivalent axial load. Identify the backlash section caused by the worm gear drive pair and / or the valve stem thread drive pair: when the equivalent resistance characterization sequence exceeds the backlash threshold and the stroke position data remains unchanged within the preset position dead zone, the corresponding data is determined to be backlash data and discarded; Outlier removal is performed on the motor electrical parameter data samples after removing idle data, and the baseline representative value is determined based on the removed samples.

[0014] Optionally, updating at least one of the generation parameters of the mechanical baseline resistance characteristic, the stroke load characteristic, and the feedforward compensation control quantity based on the motion feedback data includes: Based on the stroke position data in the motion feedback data and the valve stem side equivalent resistance characterization sequence, candidate empty stroke data pairs are identified. The candidate empty stroke data pairs satisfy that the stroke position data remains unchanged within the preset position dead zone and the valve stem side equivalent resistance characterization sequence exceeds the empty stroke threshold. In the motion feedback data of at least two motion cycles, the occurrence positions of the candidate idle data pairs are statistically analyzed to obtain the idle position distribution characteristics. When the empty-range position distribution characteristics meet the preset consistency condition, the preset position dead zone and / or the empty-range threshold are updated based on the empty-range position distribution characteristics. Based on the updated preset position dead zone and / or the idle travel threshold, idle travel data is removed again, and incremental updates are performed on at least one of the mechanical baseline resistance characteristics, the stroke load characteristics, and the generation parameters of the feedforward compensation control quantity using only the removed motion feedback data.

[0015] Secondly, this application provides an energy consumption optimization system for an electric actuator, comprising: The acquisition module is used to acquire valve body fluid property description information and actuator operation observation data, including stroke position data, motor electrical parameter data and medium pressure difference data. The processing module is used to determine the observation dataset corresponding to the preset baseline operating condition and the observation dataset corresponding to the preset pressurized operating condition based on the medium pressure difference data; establish the correspondence between the stroke position data and the motor electrical parameter data based on the observation dataset corresponding to the preset baseline operating condition, and obtain the mechanical baseline resistance characteristics associated with the stroke position; and determine the stroke load characteristics associated with the stroke position based on the observation dataset corresponding to the preset pressurized operating condition and the mechanical baseline resistance characteristics. The control module is used to determine the feedforward compensation control quantity according to the valve body fluid property description information and the stroke load characteristics, and generate motor control commands based on the feedforward compensation control quantity to control the actuator to change from the current opening degree to the target opening degree. The feedback module is used to acquire motion feedback data and update at least one of the generation parameters of the mechanical baseline resistance characteristic, the stroke load characteristic, and the feedforward compensation control quantity based on the motion feedback data.

[0016] Compared with existing technologies, this application introduces a data diversion mechanism driven by medium pressure difference. It uses observation data under low pressure difference conditions to characterize the mechanical baseline resistance features associated with the stroke position, and combines observation data under pressurized conditions with these mechanical baseline resistance features to extract stroke load features. This allows for a clearer, structured characterization of the actuator's output demand at different opening positions, considering both the "mechanical resistance background" and the "medium load contribution." Furthermore, valve body fluid property description information and stroke load features are used together to determine the feedforward compensation control quantity. An energy consumption evaluation criterion is introduced to constrain and select the compensation strategy, making the control commands more closely aligned with actual on-demand output.

[0017] Furthermore, by using motion feedback data to update the mechanical baseline resistance characteristics, stroke load characteristics, or compensation parameters, the system's adaptability to wear, temperature drift, and operating condition shifts can be enhanced, redundant drives caused by conservative torque limits and experience-based tuning can be reduced, and the response consistency and operational efficiency of the opening adjustment process can be improved. Attached Figure Description

[0018] Figure 1 A flowchart illustrating an energy consumption optimization method for an electric actuator provided in this application embodiment; Figure 2 This application provides a flowchart of a method for determining an observation dataset. Figure 3 A flowchart illustrating a method for obtaining mechanical baseline resistance characteristics provided in this application embodiment; Figure 4 This is a schematic diagram of an electric actuator energy consumption optimization system provided in an embodiment of this application. Detailed Implementation

[0019] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0020] See Figure 1 The diagram shown is a flowchart of an energy consumption optimization method for an electric actuator provided in an embodiment of this application, including steps S101 to S105, wherein: S101: Obtain valve body fluid property description information and actuator operation observation data, the operation observation data including stroke position data, motor electrical parameter data and medium pressure difference data; based on the medium pressure difference data, determine the observation dataset corresponding to the preset baseline operating condition and the observation dataset corresponding to the preset pressurized operating condition; S102: Based on the observation dataset corresponding to the preset baseline working condition, establish the correspondence between the stroke position data and the motor electrical parameter data to obtain the mechanical baseline resistance characteristics associated with the stroke position; S103: Based on the observation dataset corresponding to the preset pressurized working condition and the mechanical baseline resistance characteristics, determine the stroke load characteristics associated with the stroke position; S104: Based on the fluid property description information of the valve body and the stroke load characteristics, determine the feedforward compensation control quantity according to the preset energy consumption evaluation criteria, and generate a motor control command based on the feedforward compensation control quantity to control the actuator to change from the current opening degree to the target opening degree; S105: Acquire motion feedback data, and update at least one of the generation parameters of the mechanical baseline resistance characteristic, the stroke load characteristic, and the feedforward compensation control quantity based on the motion feedback data.

[0021] Regarding the above S101: In one embodiment, the valve body fluid property description information is used to provide a unified parameterized description of the valve structure and media conditions, so as to correlate it with operational observation data on the same time scale. The valve body fluid property description information can be composed of static configuration parameters and dynamic operating condition parameters: static configuration parameters include valve type identification (gate valve, globe valve, control valve, butterfly valve, ball valve, etc.), nominal diameter, pressure rating, valve core / plate characteristic type, flange standard, etc.; dynamic operating condition parameters include media density, media viscosity, media temperature, upstream pressure, downstream pressure, or their difference, etc.

[0022] The above information can be obtained in the following ways: from the parameter setting interface of the actuator controller (e.g., local setter, LCD menu, remote configuration interface), from the operating condition tag issued by the host computer / control system (DCS / SCADA), or from real-time acquisition by field instruments (temperature, pressure, differential pressure transmitters) and then forming a parameter vector within the controller. To ensure consistency, the valve body fluid attribute description information can be stored in the controller's non-volatile memory, such as EEPROM or the host computer's history library, in the form of "parameter name - parameter value - unit - timestamp / version number," facilitating the recall of the same set of definitions in different operating cycles.

[0023] In one embodiment, the operational observation data of the actuator includes stroke position data, motor electrical parameter data, and medium differential pressure data. Stroke position data can be acquired by position sensors, such as absolute encoders, incremental encoders, magnetic encoders, rotary transformers, or potentiometers. For multi-turn actuators, stroke position data can be characterized by a combination of "valve stem revolutions / angle—opening percentage—limit status," where the opening percentage can be calculated by the controller based on stroke calibration parameters. Motor electrical parameter data is used to characterize the electrical response on the motor drive side, and exemplarily includes stator current, bus voltage, phase voltage, phase current, active power, reactive power, power factor, drive frequency, PWM duty cycle, estimated electromagnetic torque, or equivalent load indication, etc. Its acquisition can be achieved through driver sampling circuits, Hall current sensors, shunt resistor sampling, isolation amplification, and ADC conversion, and timestamped and cached by the controller, such as a multi-core control unit based on a DSP. Medium differential pressure data is used to characterize the pressure difference between the valve and the upstream and downstream valves. It can be directly measured by a differential pressure transmitter, or it can be obtained by subtracting the pressure from the upstream and downstream pressure sensors in the controller.

[0024] Understandably, to ensure that the data can be used for segmentation, the observation data is preferably recorded with a unified sampling time base. When different channels have different sampling periods, the controller can add a unified timestamp to each channel and perform resampling and alignment to make the travel position, electrical parameters and pressure difference at the same time comparable.

[0025] In one embodiment, the observation dataset corresponding to the preset baseline operating condition and the observation dataset corresponding to the preset pressurized operating condition are determined based on the medium pressure difference data. In essence, the operating observation data is fragmented and split according to the "pressure difference level and its stability" to reduce the impact of pressure difference mutations on the consistency of subsequent modeling inputs.

[0026] The preset baseline operating condition can be defined as an operating condition where the medium differential pressure is no greater than the baseline threshold, and the preset pressurized operating condition can be defined as an operating condition where the medium differential pressure is greater than the baseline threshold. The baseline threshold can be set based on the accuracy and noise level of the differential pressure measurement link and the typical differential pressure range of the valve system. For example, the baseline threshold can be set to 1% to 5% of the full scale of the differential pressure transmitter, or set to a number of times (such as 3 times) the standard deviation of differential pressure measurement noise to avoid noise misjudgment. Alternatively, the baseline threshold can be set to an engineering threshold that can be considered as low load, such as a fixed threshold in MPa or a proportional threshold based on the rated differential pressure, in conjunction with the valve type and pipeline operation procedures.

[0027] In practice, the medium pressure difference data can be smoothed first to suppress the influence of transient spikes. The smoothing process can be a moving average, an exponential moving average, or a low-pass filter. Then, the fluctuation index and the rate of change index of the pressure difference are calculated within a preset time window to identify the relatively stable section of the pressure difference, so as to avoid mistakenly including transient pressure difference spikes induced by water hammer or cavitation, or transition sections of switching operating conditions into the dataset.

[0028] Subsequently, within the stable differential pressure zone, operating segments with a differential pressure less than or equal to the baseline threshold are selected. The synchronously collected travel position data and motor electrical parameter data within these segments are then aggregated into an observation dataset corresponding to the preset baseline operating condition. Similarly, within the stable differential pressure zone, operating segments with a differential pressure greater than the baseline threshold are selected, and the synchronously collected travel position data and motor electrical parameter data are aggregated into an observation dataset corresponding to the preset pressurized operating condition. For transitional segments where the differential pressure crosses the baseline threshold—for example, when the differential pressure rises from below the threshold to above the threshold, or falls from above the threshold to below the threshold—these segments can be removed according to preset transition rejection rules to avoid samples near the boundary being simultaneously affected by both types of operating condition factors.

[0029] For example, in a typical pipeline valve system, a multi-turn intelligent electric actuator (such as the VAD series) is used to drive a gate valve / stop valve. The controller obtains the valve stem revolutions through a position encoder and converts them into the opening percentage. It obtains the motor phase current and bus voltage through the driver sampling and calculates the medium pressure difference through pressure sensors before and after the valve. During the field commissioning phase, the controller can use the measurement noise level of the medium pressure difference as the basis for setting the baseline threshold: for example, when the valve is fully open and the pipeline is in a steady state, the pressure difference fluctuation range is statistically analyzed, and the baseline threshold is set to a multiple of the upper limit of this fluctuation range to ensure that the baseline operating condition data corresponds to the segment of "low pressure difference approximately no load". During the production operation phase, when the pressure difference is stable and exceeds the threshold, the controller will classify the corresponding segment into the pressurized operating condition dataset.

[0030] In one alternative implementation, the preset baseline operating condition is defined as an operating condition where the medium differential pressure is less than or equal to the baseline threshold. This is because the differential pressure signal in the field generally has measurement noise, zero drift and short-term spikes. If the boundary of "low differential pressure" is not defined in an engineering manner, it is easy to mistakenly classify segments that still have a significant medium load contribution into the baseline dataset, thereby contaminating the mechanical baseline resistance characteristics with the "pressure effect".

[0031] Therefore, the baseline threshold can be set and stored by the controller according to preset rules during the commissioning / operation phase. For example, the baseline threshold can be determined by selecting one or a combination of the following criteria: Firstly, based on the range and accuracy settings of the differential pressure measurement link, for example, by reading the upper limit of the range and resolution parameters of the differential pressure transmitter, the baseline threshold is set to the differential pressure value corresponding to a preset percentage range (e.g., 1% to 5%) of the upper limit of the range. Secondly, based on the static steady-state noise statistics setting, for example, when the valve is fully open and the upstream pump / compressor is running stably, a section of medium pressure difference data (such as 30s to 120s) is continuously collected, the pressure difference fluctuation range or noise bandwidth is statistically analyzed, and a preset margin is superimposed on the noise bandwidth as a baseline threshold so that the "low pressure difference" judgment is robust to noise. Third, based on pipeline operation procedures, for example, when there is a "low load / low differential pressure" label or a low differential pressure range allowed by the process in the DCS / SCADA, the baseline threshold can directly adopt the upper limit of that range or its magnification margin. To facilitate implementation on the equipment side, the baseline threshold can be saved in the actuator controller as a configuration item, for example, written to non-volatile memory with fields such as "dp_baseline_th, unit, version number, effective time", and can be modified by the local setter or remote configuration interface to take effect; at the same time, the controller can record the threshold source identifier (range ratio / noise statistics / process issuance).

[0032] In one optional implementation, the preset pressurized operating condition is defined as an operating condition where the medium pressure difference is greater than the baseline threshold, and the stroke load characteristics are determined by the deviation of the motor electrical parameters under the preset pressurized operating condition from the mechanical baseline resistance characteristics. This is because when operating under pressure, the motor electrical parameters simultaneously reflect the superposition effect of mechanical transmission resistance and medium load. If the pressurized electrical parameters are directly used as the load characterization, it will lead to a lack of comparability between different valve opening sections and different actuators. Therefore, a relative deviation caliber that does not introduce additional physical modeling but can be stably reused is needed, which can obtain more sensitive characteristics to medium load after stripping away the mechanical background terms.

[0033] To this end, stroke load characteristics can be generated in the controller or host computer in the following form: First, the observation dataset corresponding to the preset pressurized working condition is processed for stroke alignment, and the stroke position data in the pressurized segment is discretized into a stroke index system consistent with the mechanical baseline resistance characteristics; for example, the opening degree from 0% to 100% can be divided into several opening degree interval blocks according to the preset stroke resolution, and the motor electrical parameter samples whose timestamps fall into the interval are collected for each opening degree interval block.

[0034] Secondly, for each opening interval block, the baseline reference value of the same opening interval block is read from the mechanical baseline resistance characteristics. When the mechanical baseline resistance characteristics are stored in the form of a "stroke-electrical parameter" lookup table, they can be directly indexed. When stored in the form of an interpolation curve, interpolation can be performed on adjacent stroke nodes to obtain the reference value for that interval. Then, the deviation of the pressurized sample relative to the baseline reference value is determined, and the statistical characterization of the deviation is used as the stroke load characteristic. The specific form of the deviation can be any of the following: differential deviation, proportional deviation, or graded deviation. To avoid the influence of transient disturbances, a steady-state statistical caliber can be used for output, such as taking the median, upper quantile, or representative value after amplitude limiting for the deviation within the same interval block. Finally, the stroke load characteristics can be organized into a set of data structures, such as forming a record set of "position index - pressure differential segment identifier - load deviation representative value - sample quantity - time range" by opening interval block index.

[0035] For example, in a control system for a multi-turn actuator, differential pressure data is acquired by a differential pressure transmitter via 4–20mA and converted by the controller's ADC. Stroke position is obtained by an encoder, and motor electrical parameters are provided by the driver sampling module. The controller firmware can generate an observation record in each sampling cycle, containing fields such as "timestamp, stroke_pos, i_rms, u_bus, dp, dir_flag," and write it to a circular buffer. When the dp field is not greater than dp_baseline_th, the record is marked as a baseline candidate sample; when the dp field is greater than dp_baseline_th, it is marked as a live-line candidate sample. Further, in the live-line candidate samples, the controller aggregates i_rms or equivalent torque indications by opening interval blocks and looks up the baseline reference value for the corresponding interval by referring to the mechanical baseline resistance characteristics, generating interval-level load deviation representative values.

[0036] Regarding S102 above: In one embodiment, the "establishment of the correspondence between stroke position data and motor electrical parameter data" in this application can be understood as follows: From the observation dataset corresponding to the preset baseline operating condition, motor electrical parameters that characterize transmission chain friction and transmission resistance are extracted, and these parameters are aligned, aggregated, and robustned along the stroke position dimension. This results in a mapping structure of "stroke position to mechanical resistance characterization quantity" that can be queried or interpolated, serving as the mechanical baseline resistance feature. This mechanical baseline resistance feature is used to characterize the inherent resistance background of the actuator at different opening positions under conditions of no or low medium pressure differential. Its sources mainly include mechanical factors such as worm gear pairs, valve stem thread pairs, bearing supports, sealing packing preload, and reduction stage meshing.

[0037] In one embodiment, the observation dataset corresponding to the preset baseline operating condition is preferably a record sequence aligned by timestamps, and each record includes at least a travel position field and a motor electrical parameter field.

[0038] For example, a single record can be stored in the form of fields such as "timestamp, stroke_pos, I_rms, U_bus, P_in, dir_flag", where stroke_pos is the opening percentage or valve stem rotation conversion value, I_rms is the stator current RMS value, U_bus is the bus voltage or phase voltage, P_in is the input power, and dir_flag is the stroke change direction indicator.

[0039] The aforementioned records can be written to a circular buffer by the actuator controller (e.g., a DSP-based control board) within the sampling period, or generated as a time-series file by the host computer from the DCS / SCADA historical database. To facilitate embedded implementation, the observation dataset can be further organized into a segmented buffer structure, for example, written in fixed-length windows (e.g., 5s to 30s) to support online calculation and updates.

[0040] In one embodiment, the baseline characterization of the motor electrical parameters used to characterize mechanical resistance can be selected from one or more of current-related indicators, power-related indicators, or equivalent torque estimates provided by the driver.

[0041] For example, in multi-turn, high-torque actuators, the RMS current value is sensitive to load changes and the acquisition link is mature, so I_rms can be used as a baseline representation. When the field voltage fluctuates significantly, a normalization representation method can be further adopted, such as using I_rms combined with U_bus to generate electrical parameter indications, in order to reduce the interference of power supply fluctuations on baseline fitting. For scenarios where the drive supports torque estimation, the controller can also directly read the "estimated electromagnetic torque / load indication" as a baseline representation to reduce inconsistencies caused by differences in motor parameters.

[0042] In one embodiment, the stroke position can be discretized, and robust statistics can be performed on samples within the same interval. For example, the entire stroke can be divided into multiple opening interval blocks according to a preset stroke resolution. The stroke resolution can be determined based on the actuator calibration resolution, valve adjustment sensitivity, and sample density; for example, 0.5% to 2% opening can be used as an interval block. For multi-turn long-stroke structures, equivalent interval blocks can also be divided according to valve stem revolutions or pulse counts. Subsequently, for each opening interval block, motor electrical parameter samples falling into that interval block from the preset baseline operating condition observation data are collected to obtain an interval sample set.

[0043] It is important to note that to avoid baseline distortion caused by mechanical vibration, sampling spikes, or transient interference, outlier removal can be performed on the interval sample set. The removal rules can employ threshold criteria based on quantile intervals, robust criteria based on the absolute deviation of the median, or engineering criteria based on amplitude limiting. For example, this can be implemented as "removing samples above the upper quantile threshold and below the lower quantile threshold" or "removing samples that deviate from the interval median by more than a preset deviation threshold." The deviation threshold can be set based on the sampling noise bandwidth, the accuracy of the driver current measurement, and the current fluctuation range of the actuator under steady-state low differential pressure. For example, during the commissioning phase, the natural fluctuation range of I_rms within the same opening interval can be statistically analyzed, and the deviation threshold can be set as a multiple of the upper limit of this fluctuation range to balance robustness and retention of effective samples.

[0044] In one implementation, after outlier removal, a baseline representative value can be determined for the sample set of each open interval block. The baseline representative value can be selected as the median, truncated mean, or weighted mean, and the weights can be set according to the sample time continuity or electrical parameter stability. To improve data reliability, quality identification fields, such as sample size N, interval variance / volatility, and percentage of valid samples, can also be generated simultaneously and saved together with the baseline representative value.

[0045] This yields a discrete representation of the mechanical baseline resistance characteristics indexed by the opening interval blocks. For example, this discrete representation can be a lookup table structure: each entry includes fields such as "bin_id, stroke_range, baseline_value, N, stability_flag, and timestamp_range"; where baseline_value is the baseline representative value, and stability_flag indicates whether the interval meets a preset stability criterion, such as the interval fluctuation not exceeding a threshold and N not being less than the minimum sample size. The minimum sample size can be determined based on the sampling frequency and desired stability. For example, when the sampling period is 1ms to 10ms, each interval can be required to have at least 50 to 500 valid samples; in low-frequency sampling scenarios, each interval can also be required to cover at least several seconds of steady-state segments based on the time length.

[0046] In one implementation, to make the mechanical baseline resistance characteristics more continuous and easier to recall in the stroke dimension, a smoothing connection process can be performed on the discrete lookup table. The smoothing connection can be implemented by sliding window smoothing, piecewise linear connection, or spline interpolation. To avoid introducing "false transitions across intervals" through smoothing, the smoothing process can be limited to adjacent opening interval blocks, and only entries that meet the quality label conditions can participate in the smoothing.

[0047] For example, when certain aperture intervals lack representative baseline values ​​due to sparse samples or insufficient stability, these intervals can be marked as intervals to be supplemented and supplemented first when the baseline conditions reappear, rather than being directly filled by long-distance interpolation.

[0048] In an optional implementation, considering the potential for forward and reverse efficiency differences and frictional hysteresis in the transmission chain of multi-turn, high-torque actuators, to avoid generating physically inconsistent "intermediate baselines" after mixing samples at the same opening position, a stroke change direction identifier can be generated based on the difference sign or gradient sign of the stroke position data. Accordingly, the preset baseline operating condition observation dataset is split into an opening-direction baseline subset and a closing-direction baseline subset. Subsequently, two baseline curves or two sets of lookup tables are generated according to the aforementioned discretization, outlier removal, and representative value determination process, and these are combined as the mechanical baseline resistance feature. This implementation can make the baseline features more closely match the frictional differences between the worm gear and threaded pair under different force directions without introducing additional sensors, and is particularly suitable for multi-turn actuators with high torque and strong reduction ratios.

[0049] In one embodiment, the data processing in S102 described above can be performed locally by the actuator controller or offline / online by a host computer. For example, on the controller side, the firmware can implement bin aggregation and representative value calculation, and write the generated lookup table into non-volatile memory in the form of structured parameters for power-off retention; on the host computer side, a data processing program running on an industrial computer (e.g., a Python-based data cleaning script or a MATLAB-based signal processing module) can be used to batch process historical data, form a baseline table, and then send it to the actuator controller.

[0050] Regarding the above S103: In one embodiment, the stroke load feature is used to characterize the contribution of the additional load caused by the medium pressure difference and fluid action to the actuator output demand at different opening positions under a preset pressurized operating condition. This feature is preferably presented in the form of a "stroke position indexed load characterization quantity," so that it can be aligned and queried with the aforementioned mechanical baseline resistance feature using the same stroke indexing system.

[0051] It is understandable that the observation dataset corresponding to the preset pressurized working condition has been determined in step S101. Therefore, this step focuses on completing the stroke alignment, baseline referencing and load extraction within the observation dataset, and will not repeat the description of the filtering rules for the working condition segments.

[0052] In one embodiment, determining the stroke load characteristics based on the observation dataset corresponding to a preset pressurized operating condition may include: discretizing and mapping the stroke position data of each observation record in the pressurized observation dataset so that it falls into an opening interval block consistent with the mechanical baseline resistance characteristics; subsequently, collecting the corresponding motor electrical parameter data samples within each opening interval block to form an interval sample set. The discretization mapping may use an indexing method consistent with the mechanical baseline resistance characteristics, namely "bin_id, stroke_range", to ensure a one-to-one correspondence between the baseline characteristics and the pressurized samples for the same opening interval block.

[0053] For example, the pressurized observation record can still adopt the field structure of "timestamp, stroke_pos, I_rms, U_bus, dp, dir_flag", where dp is the medium pressure difference data; when the controller or host computer traverses the record, it calculates the bin_id to which the stroke_pos belongs and writes it into the interval buffer queue.

[0054] In one embodiment, to facilitate the separation of mechanical background items from piezoelectric parameters, the mechanical baseline resistance characteristics can be provided as a base reference value at the block level of the opening interval in the form of a lookup table or interpolation curve.

[0055] For example, when the mechanical baseline resistance characteristics are stored in the form of a lookup table, the baseline_value can be read directly through bin_id; when the mechanical baseline resistance characteristics include two directional channels (open / closed), the baseline reference value can be read from the corresponding channel based on dir_flag, avoiding mistaking directional friction differences for medium load differences. For certain interval blocks where the baseline is missing or the quality identifier does not meet the requirements, the interval block can be marked as an "unavailable interval," and the corresponding travel load characteristics will not be output in this step, or the calculation can wait for subsequent baseline data to be completed according to the preset supplementary sampling strategy before proceeding.

[0056] In one embodiment, the stroke load characteristics can be determined by the deviation of the live electrical parameters from the baseline reference value. Specifically, within each opening interval block, an interval deviation sample set is generated based on the motor electrical parameter samples in the interval sample set and the baseline reference value of that interval block; then, a load representative value is determined from the interval deviation sample set as the stroke load characteristic component of that opening interval block. The deviation can take any of the following forms: differential deviation, normalized deviation, or graded deviation: differential deviation is suitable for scenarios where the electrical parameters have stable dimensions and the power supply fluctuation is small; normalized deviation is suitable for scenarios where the bus voltage fluctuation is large or the individual motor constants of different actuators differ; graded deviation is suitable for scenarios where it is desirable to directly map the control strategy at discrete levels.

[0057] To enhance robustness, the median or truncated average is taken after removing peak samples. The determination of peak samples can follow the outlier removal logic consistent with the baseline construction, so that the load characteristics reflect the "typical load contribution of the stable pressure section" rather than the transient disturbances caused by water hammer, cavitation or operating condition switching.

[0058] In one implementation, to make the travel load characteristics interpretable for the pressure level and facilitate cross-cycle reuse, a hierarchical index of medium differential pressure can be introduced into the live-line observation dataset, adding a differential pressure level dimension in addition to the travel dimension. For example, the differential pressure sensor (dp) can be discretized into multiple differential pressure intervals (e.g., dp_bin_id) according to a preset differential pressure resolution, and samples can be further binned and aggregated within each interval according to dp_bin_id, thereby outputting a two-dimensional record set of "bin_id—dp_bin_id—load_value". The differential pressure resolution can be determined based on the differential pressure sensor resolution, the differential pressure fluctuation range of the process system, and the desired control granularity. For example, the minimum resolution can be a multiple of the differential pressure measurement noise bandwidth, or a preset proportion of the typical operating differential pressure range can be used as the grading step size. This implementation can distinguish the load contribution under different differential pressure levels within the live-line operating condition, avoiding the averaging of the load differences between the high and medium differential pressure sections, thus making subsequent feedforward compensation closer to the actual operating condition.

[0059] In one embodiment, the stroke load characteristics can be constructed into a load characteristic table, and each table entry includes fields such as "bin_id, stroke_range, dp_bin_id (optional), load_value, sample_count, stability_flag, timestamp_range"; wherein sample_count is used to indicate how many valid samples generated the table entry, and stability_flag is used to indicate whether the table entry meets the preset stability criteria (e.g., the deviation of sample volatility within the interval does not exceed the threshold and the proportion of valid samples is not lower than the threshold).

[0060] For example, in a typical pipeline gate valve scenario, a multi-turn intelligent electric actuator drives the valve stem. During pressurized operation, the controller continuously collects stroke_pos, I_rms, U_bus, and dp. The controller first writes the sample to the corresponding opening interval block cache based on stroke_pos, then reads the baseline_value of the same opening interval block (and the same direction channel) from the mechanical baseline resistance characteristic table to generate interval deviation samples. Subsequently, outlier removal is performed within the interval, and the median is calculated as the load_value. If dp exhibits multiple stable levels during this stage, the controller can further bin the interval samples according to dp_bin_id, outputting the load_value for different differential pressure levels. The resulting stroke load characteristic table can be written to the running memory with a version number cache, allowing subsequent steps to directly query the corresponding load contribution by "current stroke position / target stroke position and current differential pressure level" when calculating the feedforward compensation control quantity.

[0061] Regarding S104 above: In one embodiment, the "determining the feedforward compensation control quantity according to the preset energy consumption evaluation criteria" in this application is used to solve the following problem: when the valve is operating under pressure and the contribution of the medium load at different opening positions fluctuates with time, if only a fixed opening-control quantity mapping is used or only feedback adjustment is relied upon, there will often be response lag due to insufficient drive margin in the start-up stage, or redundant energy consumption and thermal stress accumulation due to the long-term use of excessively high drive margin to avoid lag.

[0062] Therefore, without introducing additional load sensors, this step uses the valve body fluid property description information to parameterize the load pattern on the medium side and uses the stroke load characteristics to quantify the load contribution at the current opening position, thereby generating a feedforward compensation control quantity that matches the target opening adjustment task, and forming a motor control command that can be directly executed by the driver.

[0063] In one embodiment, the valve body fluid property description information is preferably used to determine the "structural type coefficient" and "medium sensitivity coefficient" of the feedforward compensation.

[0064] For example, for gate valves and globe valves, the sensitivity of valve stem axial load and sealing surface friction to the opening range varies greatly, and the controller can select different compensation templates according to the valve type identification; for regulating valves or butterfly valves, the flow coefficient and pressure difference are more likely to cause load changes in the small and medium opening range, and the controller can increase the compensation weight of the corresponding opening range.

[0065] Furthermore, the medium density, viscosity, and temperature can be used as inputs to the medium sensitivity coefficient, which can be used to adjust the upper limit or slope of the compensation amplitude under the same stroke load characteristics. For example, under high viscosity or low temperature conditions, the compensation margin in the low speed range can be increased to avoid the amplification of minor jamming caused by changes in fluid resistance by the feedback loop.

[0066] In one embodiment, the travel load characteristics are preferably input to the feedforward calculation module in the form of "load representative value indexed by bin_id". After receiving the target opening, the controller can map the current opening and the target opening to the starting opening interval block and the target opening interval block, and determine the set of interval blocks to be traversed; then, it reads the corresponding load representative value from the interval block set to form a load sequence corresponding to the action path.

[0067] For example, the load sequence can be organized as a structured cache, such as a set of fields "stroke_bin_list, load_value_list, dir_flag, dp_level (optional)", for the subsequent candidate control quantity generation module to directly traverse.

[0068] In one embodiment, the feedforward compensation control quantity can be a control quantity increment or control quantity configuration item that can be directly consumed by the motor control loop. Exemplary examples include at least one of the following: the feedforward increment of the target electromagnetic torque command, the dynamic adjustment of the torque limit upper limit, the feedforward correction of the speed setpoint curve, the correction of the drive frequency setpoint, the correction of the PWM duty cycle, or the correction of the current setpoint. For drivers employing vector control or torque closed-loop control, it is preferable to generate a "feedforward increment of torque / current setpoint"; for drivers employing V / f or simplified control, it is preferable to generate a "feedforward correction of frequency / duty cycle". In this embodiment, the feedforward compensation control quantity can be described as "control quantity bias for improving drive output within the load interval block", and implemented using a lookup table or segmented rule method: for example, for each opening interval block, the corresponding compensation level code is selected based on the level interval into which its load representative value falls, and the compensation level code is then mapped to a control quantity configuration item recognizable by the driver.

[0069] In one implementation, "at least one candidate feedforward compensation control quantity" is used to make engineering trade-offs between energy consumption and response, thereby avoiding insufficient generalization of a single compensation strategy under different valve types, different media, and different differential pressure levels. The candidate feedforward compensation control quantity can be generated in one or a combination of the following ways: First, setting different sets of compensation gains based on the same load sequence to obtain multiple sets of candidate quantities with different compensation amplitudes; second, setting different speed curve templates based on the same compensation amplitude, such as uniform, two-stage, or three-stage speed inputs, to obtain multiple sets of candidate quantities; third, setting different limiting strategies based on the same speed template, such as different peak current limits or different torque limits, to obtain multiple sets of candidate quantities. For ease of embedded implementation, the candidate set can be stored in a fixed-length array and evaluated one by one by the controller according to the candidate number.

[0070] In one optional implementation, a preset energy consumption evaluation criterion is used to address the following problem: During the process of the actuator changing from the current opening degree to the target opening degree, different candidate feedforward compensation control quantities will lead to different current peak values, different steady-state consumption, and different action durations. If only the completion of the action is taken as the evaluation endpoint, there may be situations where the action is completed but the energy consumption is too high or the motor heat load accumulates too quickly. Therefore, it is necessary to perform quantifiable evaluation of the candidate quantities within the preset action time window in order to select the candidate quantities with better energy consumption while satisfying the achievability of the action.

[0071] In this optional implementation, the preset action time window can be defined as "the interval from the moment the motor control command is issued until the target opening degree is reached and the stable zone is entered", or as "a fixed duration window from the moment the command is issued". The time window can be set based on the rated action time of the actuator, the full stroke time of the valve, and the adjustment cycle allowed by the control system.

[0072] For example, in multi-turn gate valve scenarios, the time window can be set to 1.1 to 1.5 times the typical action time to cover the deceleration and stabilization phases near the operating position. For short-stroke, frequent adjustment scenarios of control valves, the time window can be set to a fixed window of 2 to 10 seconds to allow for energy consumption assessment using a fixed-length buffer when controller resources are limited. The time window parameter can be written as a configuration item to the controller's non-volatile memory, for example, saved as a field with the following parameters: "t_eval_win, unit ms, version number, effective time," and can be configured in the debugging interface.

[0073] In this optional implementation, the energy consumption index can be determined from the motor electrical parameter data within a preset action time window. For example, the energy consumption index may include at least one of the following: cumulative energy consumption indicator, average power indicator, peak current indicator, and over-limit duration indicator. The cumulative energy consumption indicator can be obtained by the controller accumulating the input power samples within the window; the average power indicator can be determined by the mean or quantile of the power samples within the window; the peak current indicator can be determined by the maximum value of the current samples within the window; and the over-limit duration indicator can be calculated from the cumulative number of sampling points where the estimated current or temperature rise exceeds a threshold. All of the above indicators can be implemented on the controller side through a circular buffer: the controller writes "timestamp, I_rms, U_bus, P_in" in each sampling cycle and synchronously updates the window statistics, outputting the energy consumption evaluation result of the candidate quantity at the end of the window.

[0074] In this optional implementation, the process of "determining the feedforward compensation control quantity from at least one candidate feedforward compensation control quantity" can employ a hierarchical screening strategy to ensure engineering availability: First, candidate quantities that do not meet the requirements for action reachability are eliminated. Action reachability can be determined by the position criterion, such as the travel position entering the target opening tolerance zone and the limit / position state meeting the requirements before the end of the time window; Second, among the candidate quantities that meet the requirements for action reachability, the target candidate quantity is selected based on energy consumption indicators, such as prioritizing the candidate quantity with the smallest cumulative energy consumption indication and peak current not exceeding the threshold; When the energy consumption indicators of multiple candidate quantities are similar, a stability auxiliary criterion can be further used for selection, such as selecting the candidate quantity with smaller current fluctuation or shorter over-limit duration.

[0075] To avoid process disturbances caused by multiple real-time online trial actions, the screening process can solidify parameters after offline evaluation of typical action tasks during the commissioning phase, or it can update the candidate optimization results online in a low-frequency triggering manner during the operation phase, such as performing an evaluation and updating the candidate number once during the maintenance window or a period of stable operating conditions.

[0076] In one embodiment, generating motor control commands based on feedforward compensation control quantities may include: writing the feedforward compensation control quantities into a drive control parameter register or control variable area, and forming a motor control command frame together with the basic control setpoint, which is then sent to the driver; the command frame may include fields such as target opening degree, running direction, speed template number, amplitude limiting template number, and feedforward compensation parameters. For example, in a scenario where the actuator controller is a DSP control board and the driver communicates via CAN or RS485, the controller can generate a "setpoint_frame" based on the target opening degree and direction, which includes fields such as "target_pos, dir_flag, speed_profile_id, torque_limit, and ff_bias_code," and enter a motion monitoring state after sending it, so that subsequent steps can continue to use motion feedback data for updates.

[0077] Optional, see Figure 2 The present application provides a flowchart of a method for determining an observation dataset, including steps S201 to S204, wherein: S201: Perform smoothing processing on the medium pressure difference data, and calculate the fluctuation index and change rate index of the medium pressure difference within a preset time window to determine the pressure difference stability section that meets the stability criterion; S202: In the pressure difference stable section, select the operating segment where the medium pressure difference is less than or equal to the baseline threshold, and collect the stroke position data and motor electrical parameter data synchronously collected in the operating segment into the observation dataset corresponding to the preset baseline operating condition; S203: In the pressure difference stable section, select the operating segment where the medium pressure difference is greater than the baseline threshold, and collect the stroke position data and motor electrical parameter data synchronously collected in the operating segment into the observation dataset corresponding to the preset pressurized working condition; S204: Remove the differential pressure transition section that crosses the baseline threshold from the operating segment.

[0078] In one alternative implementation, the field medium differential pressure data often includes zero drift, short-term spikes, rapid fluctuations caused by water hammer / cavitation, and transition sections during valve regulation. If the dataset is directly divided based on the instantaneous differential pressure, unstable segments may be mistakenly included in the baseline or pressurized dataset, which may lead to inconsistencies in the construction of subsequent mechanical baseline resistance characteristics and stroke load characteristics. This manifests as baseline curve jitter, load deviation being amplified by spikes, or non-reusability across cycles.

[0079] In response, the step of “determining the observation dataset corresponding to the preset baseline working condition and the observation dataset corresponding to the preset pressurized working condition” in S101 can be further refined into a section identification process oriented towards engineering noise and working condition switching disturbances.

[0080] In this optional implementation, the controller or host computer first performs smoothing processing on the medium differential pressure data to suppress the interference of transient spikes on the segment determination. Smoothing processing can be implemented using methods such as moving average, exponential moving average, or first-order low-pass filtering. For example, when the medium differential pressure is acquired by a differential pressure transmitter via 4–20mA and converted by the controller's ADC, the controller firmware can filter the raw differential pressure dp_raw in each sampling cycle, output the dp_filt field, and write it to the circular buffer; when the data is processed on the host computer side, it can smooth the historical curve and output the corresponding filtered sequence for subsequent stability analysis.

[0081] After smoothing, the fluctuation and rate of change of the medium pressure difference can be calculated within a preset time window to determine the stable pressure difference range. The preset time window is used in engineering to distinguish between "steady-state pressure / steady-state low pressure difference" and "disturbance / switching". The window length can be determined based on the sampling frequency, pipeline inertia, and the desired steady-state determination sensitivity.

[0082] For example, for actuator controllers with sampling periods of 10ms to 100ms, the time window can be set to 0.5s to 5s to cover a continuous steady state. For scenarios where samples can only be uploaded to DCS / SCADA at 1s intervals, the time window can be set to 5s to 30s to mitigate occasional fluctuations caused by discrete sampling. The volatility index can be characterized by the "amplitude of pressure difference fluctuation within the window" or the "dispersion of pressure difference within the window," for example, by the difference between the maximum and minimum values ​​within the window, or by the dispersion of samples around a representative value within the window. The rate of change index can be characterized by the "strength of the pressure difference change trend within the window," for example, by the absolute value of the difference between the start and end pressure differences within the window, or by the representative value of the difference between adjacent sampling points within the window. All of the above indicators can be updated online by the controller: the controller maintains the start and end values, extreme values, and statistics of adjacent differences within the circular buffer, and updates the dp_fluct and dp_rate fields as the window slides.

[0083] In this optional implementation, the determination of the differential pressure stability zone can be based on stability criteria, which at least constrain both the fluctuation index and the rate of change index to not exceed their corresponding thresholds. The thresholds can be set based on the noise level of the differential pressure measurement link, the allowable fluctuation range of the process system, and the requirements for the purity of the steady-state segment.

[0084] For example, during the commissioning phase, a dp_filt sequence can be collected under conditions of fully open valves and stable operating conditions. Its natural fluctuation range can be statistically analyzed, and the fluctuation threshold can be set as the upper limit of this range plus a preset margin. Simultaneously, the natural fluctuation band of adjacent sampling differences can be statistically analyzed, and the rate of change threshold can be set as a multiple of the upper limit of this fluctuation band to avoid misidentifying noise as a change in operating conditions. To enhance robustness, a minimum duration constraint can be introduced; that is, only when several consecutive time windows meet the stability criterion is the corresponding time period confirmed as a stable differential pressure segment, thereby avoiding misjudgments caused by the accidental fulfillment of a single window. The aforementioned threshold and duration parameters are also preferably stored as configuration items in the controller's non-volatile memory or the host computer's parameter library.

[0085] After determining the differential pressure stabilization zone, operating segments with a differential pressure less than or equal to the baseline threshold can be selected within the differential pressure stabilization zone. The stroke position data and motor electrical parameter data collected synchronously within the operating segments can be aggregated into an observation dataset corresponding to the preset baseline operating condition. Similarly, operating segments with a differential pressure greater than the baseline threshold can be selected within the differential pressure stabilization zone, and the stroke position data and motor electrical parameter data collected synchronously within the operating segments can be aggregated into an observation dataset corresponding to the preset pressurized operating condition.

[0086] In practical implementation, the controller or host computer can adopt a "segment labeling - record archiving" approach: when traversing the time series, a stable_flag and baseline_flag / pressurized_flag are generated for each observation record. The stable_flag indicates whether the time of the record falls within a stable differential pressure segment, the baseline_flag indicates whether dp_filt is less than or equal to the baseline threshold, and the pressedurized_flag indicates whether dp_filt is greater than the baseline threshold. Subsequently, only records where both stable_flag and baseline_flag are true are included in the baseline observation dataset, and only records where both stable_flag and pressedurized_flag are true are included in the pressurized observation dataset. The fields of the observation records can follow the aforementioned structure, such as "timestamp, stroke_pos, I_rms, U_bus, dp_filt, dir_flag, stable_flag, state_tag," etc., where state_tag can be "BASELINE / LOAD" so that subsequent steps can directly filter by label.

[0087] Furthermore, to avoid boundary contamination of the two datasets by the transition section where the differential pressure crosses the baseline threshold, the differential pressure transition section can be removed from the running segment. The differential pressure transition section can be understood as a short-term switching phase when the differential pressure crosses from one side of the threshold to the other. During this phase, the differential pressure and electrical parameters may be simultaneously affected by pipeline scheduling, valve opening changes, and fluid transients. Directly including it in either dataset would reduce the purity of the sample. The removal method can be implemented using a combination of "threshold hysteresis band + time dead zone": First, a hysteresis band is set near the baseline threshold. For example, "obvious baseline zone" is defined as differential pressure not greater than (baseline threshold minus a preset margin), and "obvious pressure zone" is defined as differential pressure greater than (baseline threshold plus a preset margin). Records falling into the hysteresis band are uniformly marked as transition candidates and are not included in either dataset. Second, when a differential pressure crossing event is detected, a preset duration of records is removed before and after the crossing moment as a transition section to cover the crossing instant and its short-term aftershocks. The preset margin and rejection time can be set based on the differential pressure noise bandwidth, water hammer decay time, and system response inertia. For example, in a pipeline network scenario where the differential pressure noise is on the order of 0.002 MPa and the residual water hammer vibration lasts for about 1 to 3 seconds, the hysteresis margin can be set to 0.005 MPa, and the 2 seconds before and after the crossing event can be recorded as the transition section to be rejected. When the controller is implemented, it can maintain three states in the state machine: "BASELINE_STABLE, PRESS_STABLE, and TRANSITION". When dp_filt enters the hysteresis band or a crossing occurs, it enters TRANSITION and starts the rejection timer. After the timer ends and dp_filt stably falls to one side, it switches back to the corresponding stable state.

[0088] For example, in a typical actuator system, the controller generates an observation record for each sampling cycle and writes it to a circular buffer. The buffer record includes "timestamp, stroke_pos, i_rms, u_bus, dp_raw". The controller filters dp_raw to obtain dp_filt and calculates dp_fluct and dp_rate using a 1-second sliding window. When dp_fluct does not exceed a preset fluctuation threshold and dp_rate does not exceed a preset rate of change threshold, and this condition is met continuously for at least 3 windows, the controller marks the corresponding time period as a pressure differential stable segment. Subsequently, within the stable segment, the controller labels the records based on the comparison result of dp_filt and the baseline threshold, and enters a transition state when it detects that dp_filt crosses the threshold, removing records 2 seconds before and after the crossing time. Finally, the controller aggregates the remaining records labeled BASELINE into an observation dataset corresponding to the preset baseline operating condition, and aggregates the records labeled LOAD into an observation dataset corresponding to the preset pressurized operating condition, and appends fields such as "segment start and end time, number of samples, threshold version number, and filter parameter version number" to the dataset.

[0089] In one alternative implementation, the valve is usually accompanied by structural effects such as static friction abrupt change, sealing pair compression, and limiting contact during the disengagement and re-seating stages. This causes the motor electrical parameters to show spikes or transitions with different mechanisms than the "regulation section load". If such samples are directly incorporated into the baseline dataset or the pressurized dataset, the subsequent baseline resistance characteristics and stroke load characteristics will be contaminated by the "end structure effect". This manifests as unstable representative values ​​of the same opening interval block, reduced cross-cycle reusability, and easy overcompensation or misjudgment of jamming at the end during feedforward compensation calculation.

[0090] Therefore, this application completes the engineering removal of end segments before data aggregation by means of process discretization, bidirectional unit construction, structural stage index determination, and retaining only adjustment segment samples.

[0091] In this optional implementation, the controller or host computer first discretizes the entire stroke into multiple opening interval blocks according to a preset stroke step size. The stroke step size is used to balance the sample density and interval granularity interpretability within the interval, and its setting can be determined based on the position sensor resolution, valve adjustment sensitivity, and typical operating speed.

[0092] For example, when the opening is represented as a percentage and the encoder's conversion resolution is better than 0.1%, the stroke step size can be set to 0.5% or 1%; when represented as valve stem revolutions / pulse count, the stroke step size can be set to a number of pulses or a number of division angles, so that a single interval block can cover no less than a preset number of sampling points during a typical operation. After discretization, each opening interval block can be represented by structured fields, such as "bin_id, stroke_range (upper and lower limits of the interval), stroke_center (center of the interval), bin_version".

[0093] Furthermore, to incorporate the directional friction differences between the opening and closing directions into the same criterion, the system combines the opening interval blocks corresponding to the travel positions in the opening and closing directions into bidirectional travel units. These bidirectional travel units can be understood as maintaining separate open-direction and closing-direction sample sets for the same bin_id, and treating them as a single unit for stage determination and sample selection.

[0094] For example, the system can construct a bidirectional cell table in the controller's circular buffer or in the host computer. Each entry includes "bin_id, stroke_range, open_sample_queue, close_sample_queue, open_count, close_count", where open_sample_queue stores the motor electrical parameter records that fall into the interval block when the direction is open, and close_sample_queue stores the motor electrical parameter records that fall into the interval block when the direction is closed.

[0095] For example, the queue record can use a subset of the aforementioned observation record fields, such as "timestamp, stroke_pos, I_rms (or equivalent torque indicator), U_bus, dir_flag, dp_tag, stable_flag", and dir_flag directly determines whether to write to the open or close queue.

[0096] In this optional implementation, the system determines the valve structure stage index to which each bidirectional stroke unit belongs based on the stroke end distance of the bidirectional stroke unit and the rate of change of motor electrical parameters. The stroke end distance is used to reflect the proximity of the interval block to the stroke end (e.g., the 0% end and the 100% end), which can be characterized by the distance between the number of intervals between bin_id and the end bin_id, or by the opening distance between the center opening of the interval and the end opening.

[0097] To avoid making overly strong assumptions about different valve types, this embodiment defines the stroke end distance as "distance to the nearest end" and sets the end influence bandwidth in the system parameters for coarse-grained marking of the end region.

[0098] For example, when the stroke step size is 1%, the end effect bandwidth can be set to 2% to 5%, meaning that bidirectional units no more than 2 to 5 blocks away from the 0% end or no more than 2 to 5 blocks away from the 100% end are considered as end candidate units. The end effect bandwidth setting can be based on the valve structure and sealing pair length, the duration of the low-speed segment caused by the actuator reduction ratio, and the opening span of the end peak in the field historical curve; this parameter can be written as a configuration item into the controller's non-volatile memory, for example, recorded as "stroke_end_band, unit = bin or %, version number, effective time".

[0099] The rate of change of motor electrical parameters is used to characterize the degree of abrupt change or the intensity of rise / fall within a given interval, in order to capture structural transitions caused by disengagement or repositioning. The rate of change can be calculated using load-sensitive and stable acquisition links, such as I_rms, estimated electromagnetic torque indication, and input power indication. The engineering implementation of the rate of change can employ adjacent sampling difference or short-window difference: the system takes several consecutive sampling points within the same interval, and the difference between the end sample value and the beginning sample value is used as a representative value of the change intensity. Short-window smoothing can be performed on the electrical parameter sequence before calculation to suppress sampling spikes.

[0100] A minimum sample size requirement can be set for the rate of change. When there are insufficient valid samples in a certain direction within the interval block, the rate of change in that direction can be marked as invalid, and the other direction can be used or the determination can be delayed until the samples are replenished during the stage determination. The setting of the rate of change threshold can be determined based on the steady-state fluctuation band during the debugging stage: for example, the natural fluctuation amplitude of I_rms is statistically analyzed in the middle adjustment interval, and the rate of change threshold is set to a certain multiple of this amplitude, so that normal adjustment fluctuations do not trigger the end stage determination, while the deseat / seat transitions can be reliably identified; the corresponding parameters can be recorded as "dE_param_th, parameter name, unit, version number".

[0101] Regarding the stage index determination rules, the system comprehensively considers the distance and rate of change at the end of the stroke, and combines directional information to generate disengagement index, adjustment index, and repositioning index. For example, when a bidirectional stroke unit is located within the end influence zone and the rate of change of electrical parameters in the open direction sample set exceeds a preset threshold, the unit is marked as a disengagement index to characterize the stage of detachment from the end structural constraint; when a bidirectional stroke unit is located within the end influence zone and the rate of change of electrical parameters in the closed direction sample set exceeds a preset threshold, the unit is marked as a repositioning index to characterize the stage of approaching the end structural constraint and experiencing compression / contact; when a bidirectional stroke unit is not located within the end influence zone, or although it is located within the end influence zone, the rates of change in both the open and closed directions do not exceed the threshold, the unit is marked as an adjustment index.

[0102] To avoid single-event triggers, consistency constraints can be introduced: for example, requiring that the determination of a rate of change exceeding a threshold occurs consecutively in several adjacent interval blocks, or requiring that at least a preset number of threshold-exceeding events occur in the same direction before the final detach / reposition index is confirmed, thereby suppressing missegmentation caused by accidental spikes. The final structural stage index can be a field in a bidirectional cell table, such as "phase_idx", whose value can be "UNSEAT / REGULATE / SEAT" or the corresponding encoded value, along with the determination source direction, trigger threshold version number, and number of valid samples.

[0103] After determining the structural stage index, only the stroke position data and motor electrical parameter data corresponding to the bidirectional stroke units with the valve structural stage index as the adjustment index are selected for the collection of observation datasets, and the data corresponding to the bidirectional stroke units with the valve structural stage index as the disengagement index or the in-seat index are removed. The "for collection" here can be directly reflected in the data archiving logic: when the controller or host computer is about to write an observation record into the preset baseline operating condition observation dataset or the preset pressurized operating condition observation dataset, the bin_id is first determined according to the stroke_pos of the record, and then the phase_idx of the bidirectional unit table is queried; only when phase_idx is the adjustment index is the record allowed to enter the subsequent baseline / pressurized collection process (including the differential pressure stable section determination, baseline threshold comparison, transition section removal, etc., which have been described in the aforementioned implementation), otherwise it is directly discarded or marked as "end removal sample".

[0104] For example, in a multi-turn gate valve scenario, the controller discretizes the opening interval blocks in 1% steps and maintains a set of opening and closing current samples for each bin_id. During the debugging phase, it was found that I_rms showed a significant rise peak when the valve started from near 0%, and a seating squeeze peak when the closing direction was close to 0%. The peaks typically covered an interval block of about 0% to 3%. Based on this, the end effect bandwidth was set to 3 interval blocks, and the rate of change threshold was set to several times the natural fluctuation amplitude of the current in the middle section regulation interval. During operation, the system marks the cells with bin_id near 0% and opening rate of change exceeding the threshold as dissemination indexes, and the cells with bin_id near 0% and closing rate of change exceeding the threshold as seating indexes. The remaining middle section interval blocks are marked as regulation indexes. When collecting baseline or pressurized observation datasets, only the records in the regulation index interval blocks are retained, so that the subsequent mechanical baseline resistance characteristics and stroke load characteristics more centrally reflect the load law of the controllable regulation section, reducing the interference of end structure effects on energy consumption evaluation and feedforward compensation.

[0105] Optional, see Figure 3 The flowchart of a method for obtaining mechanical baseline resistance characteristics provided in this application embodiment includes steps S301 to S304, wherein: S301: Determine the travel change direction identifier based on the difference sign and / or change gradient sign of the travel position data, and split the observation dataset corresponding to the preset baseline condition into an open baseline subset and a closed baseline subset according to the travel change direction identifier; S302: Divide the entire stroke into multiple opening interval blocks according to a preset stroke resolution, and collect motor electrical parameter data samples in each of the opening baseline subset and the closing baseline subset respectively within each opening interval block; S303: Determine the baseline representative value for the motor electrical parameter data sample within each of the aforementioned opening interval blocks; S304: Generate an opening mechanical baseline resistance curve and a closing mechanical baseline resistance curve based on the travel position of each opening interval block and the corresponding baseline representative value, and combine the opening mechanical baseline resistance curve and the closing mechanical baseline resistance curve to form a direction-sensitive mechanical baseline resistance feature.

[0106] In one optional implementation, for multi-turn electric actuators, the worm gear and valve stem thread pair often exhibit efficiency differences, frictional hysteresis, and force direction differences in the opening and closing directions. This results in inconsistent motor electrical parameters corresponding to the same opening position in different motion directions. If a single "stroke-electrical parameter" relationship is still used as the mechanical baseline resistance feature, directional differences are easily mistakenly treated as random noise and averaged out, thereby weakening the discriminative power of subsequent load deviation extraction and feedforward compensation. Based on this, this implementation first splits the preset baseline operating condition observation dataset according to the direction of stroke change, then constructs baseline resistance curves for the opening and closing directions respectively, and combines them into a direction-sensitive mechanical baseline resistance feature.

[0107] In this optional implementation, the stroke change direction identifier can be determined based on the differential sign and / or gradient sign of the stroke position data. For example, the controller calculates the difference component for the stroke_pos of two adjacent observation records within the sampling period and compares it with a preset position deadband threshold: when the difference component is greater than the position deadband threshold, it is marked as open; when the difference component is less than a negative position deadband threshold, it is marked as closed; when the difference component falls within the position deadband threshold range, it is marked as held / not included in the direction statistics. The position deadband threshold is used to suppress direction misjudgment caused by encoder quantization jitter or slight rebound. Its setting can be determined based on the position sensor resolution and the ambient noise level, for example, taking a multiple of the encoder resolution corresponding to the opening conversion value, or taking the upper limit of the natural fluctuation amplitude of stroke_pos in the "stopped" state during the debugging phase; and can be permanently stored in the controller as a configuration item, for example, recorded as "pos_deadband, unit, version number, effective time". In another example, to enhance noise immunity, the controller can smooth the stroke_pos within a short window before calculating the sign of the changing gradient, and introduce a minimum number of continuous points constraint when switching directions. For example, the direction is confirmed only when several consecutive sampling points satisfy the same sign, thereby avoiding frequent switching triggered by instantaneous spikes.

[0108] After obtaining the stroke change direction identifier, the observation dataset corresponding to the preset baseline condition is split into an open-direction baseline subset and a closed-direction baseline subset based on the identifier. For example, the observation records can still be stored using structured fields, such as "timestamp, stroke_pos, I_rms (or equivalent torque indication), U_bus, P_in (optional), dir_flag," etc.; where dir_flag is obtained from the above determination. During splitting, the system can write records with dir_flag indicating an open direction to the open-direction baseline cache queue and records with dir_flag indicating a closed direction to the closed-direction baseline cache queue; records with dir_flag indicating a hold can be removed according to preset rules, or incorporated into the adjacent direction cache when it is confirmed that they are in a stable uniform speed adjustment segment.

[0109] After the splitting is completed, the entire stroke is divided into multiple opening interval blocks according to a preset stroke resolution. Motor electrical parameter data samples are then collected within each opening interval block for both the opening and closing baseline subsets. The stroke resolution can be consistent with the aforementioned S102, and its setting can be based on valve adjustment sensitivity, sample density, and control granularity requirements. For example, an interval block can be defined as 0.5% to 2% opening, and the interval can be defined in the form of "bin_id, stroke_range, stroke_center". During the collection process, the controller or host computer can maintain an opening sample container and a closing sample container for each bin_id. The records in the containers can retain only a subset of fields used for baseline construction, such as "timestamp, I_rms (or equivalent torque indication), U_bus, quality_flag", and simultaneously maintain counters and fluctuation statistics to support subsequent quality judgment.

[0110] When determining the baseline representative value for motor electrical parameter data samples within each opening interval block, a robust statistical approach consistent with S102 can be used, but the opening and closing directions are calculated independently to obtain the baseline representative values ​​for the opening and closing directions. For example, the system can first perform outlier removal within the interval block, using quantile limiting, robust thresholds based on median deviation, or engineering limiting criteria as removal rules; then, the median or truncated mean of the remaining samples is taken as the baseline representative value for that interval block. To improve usability, a quality identification field, such as "sample_count, valid_ratio, stability_flag," can be generated for each interval block, and the baseline representative value is output only for interval blocks that meet the minimum sample size and stability threshold; insufficient interval blocks can be marked as intervals to be supplemented.

[0111] After obtaining the stroke position and corresponding baseline representative value of each opening interval block, the system generates the opening-direction mechanical baseline resistance curve and the closing-direction mechanical baseline resistance curve, respectively. To facilitate rapid querying on the embedded side, the curves are preferably stored in the form of a lookup table or a piecewise linear interpolation table, rather than expressed by a complex fitting model. For example, a "baseline_table" structure can be constructed, with each entry including "bin_id, stroke_range, baseline_open_value, baseline_close_value, sample_count_open, sample_count_close, stability_flag_open, stability_flag_close, version_tag", where baseline_open_value and baseline_close_value correspond to the opening-direction and closing-direction baseline representative values, respectively. When the controller needs to query the baseline reference value at any stroke_pos, it can first locate the bin_id, then select the corresponding channel value based on dir_flag, and perform linear interpolation at the bin boundary according to the stroke_center of the adjacent interval block to obtain a more continuous baseline reference. Finally, the opening-direction mechanical baseline resistance curve and the closing-direction mechanical baseline resistance curve are combined to form a direction-sensitive mechanical baseline resistance feature.

[0112] Optionally, multi-turn transmission chains (worm gear pairs and valve stem thread pairs) commonly exhibit backlash and idle travel, especially when the direction reverses or the end force changes. The motor side may have already output a significant torque / current response, while the stroke position remains unchanged within the "preset dead zone" for a short period. If such idle travel segments are included in the baseline sample set, the representative values ​​of electrical parameters within the interval block will be abnormally inflated, thus distorting the mechanical baseline resistance curve. Therefore, this embodiment first performs an equivalent conversion on the observed data using the transmission chain structural parameters before determining the baseline representative value, identifies and removes backlash and idle travel data, and then performs outlier removal and representative value determination on the purified sample.

[0113] In this optional embodiment, the actuator is a multi-turn actuator, and its transmission chain includes a worm gear drive pair and a valve stem thread drive pair. The system acquires a set of transmission parameters corresponding to the worm gear drive pair and the valve stem thread drive pair. The set of transmission parameters includes at least the reduction ratio parameter and the thread lead parameter. The parameters can be acquired by reading from the actuator nameplate / factory configuration file, reading from the controller parameter setting interface, or being written and sent to the controller by the host computer during the debugging phase; and preferably stored in a structured form, such as "gear_ratio, screw_lead, efficiency_tag (optional), backlash_deadband (optional), version_tag". The backlash_deadband can be calculated from the manufacturer's nominal backlash angle / pulse count, or obtained by statistically analyzing the duration of the stationary position interval during a "direction reversal test".

[0114] After obtaining the set of transmission parameters, the system performs equivalent conversion on the motor electrical parameter data based on this set to obtain the equivalent resistance characterization sequence on the valve stem side. Here, the "equivalent conversion" preferably uses coefficient conversion or table lookup conversion that the controller can directly implement, without requiring complex physical modeling. For example, when the driver provides an "estimated electromagnetic torque / load indication" register, the controller can directly read this indication as the motor output and, combined with the reduction ratio parameter and a preset efficiency coefficient (configurable), convert it into an equivalent torque indication on the valve stem side. When the driver does not provide torque estimation, the controller can generate a "motor-side torque indication" based on the current RMS value I_rms and a pre-stored motor torque constant calibration coefficient, and then convert it to the valve stem side according to the reduction ratio parameter. Furthermore, if it is necessary to characterize the resistance using the axial load diameter, the thread lead parameter can be combined, and the equivalent torque on the valve stem side can be mapped to an equivalent axial load indication through pre-stored conversion coefficients or table lookup. Ultimately, the equivalent resistance characterization sequence on the valve stem side can be represented using a unified data structure. For example, each record includes "timestamp, stroke_pos, dir_flag, equivalent_torque (or equivalent_axial_load), dp_tag, bin_id", and is aligned with the original electrical parameter records in terms of timestamps, which facilitates subsequent filtering and statistics.

[0115] When identifying the idle section caused by hysteresis, this implementation uses a combined criterion of equivalent resistance exceeding a threshold and stroke position remaining unchanged. Specifically, the system sets an idle threshold and a preset position dead zone: the idle threshold is used to avoid misjudging slight current fluctuations as idle force, and can be determined based on the upper limit of the natural fluctuation of equivalent resistance in the mid-section adjustment range under baseline operating conditions, and can be set as several times that upper limit or based on engineering experience; the preset position dead zone can be consistent with or more stringent than the aforementioned position dead zone threshold, and is used to determine "stroke position remains unchanged". When the equivalent resistance characterization sequence on the valve stem side exceeds the idle threshold and the stroke position data remains unchanged within the preset position dead zone, the corresponding data is determined to be idle data and discarded. To enhance robustness, "remain unchanged" can require continuous satisfaction of a certain number of sampling points or continuous satisfaction of a preset time length. For example, the stroke_pos change must fall within the position dead zone for M consecutive sampling points, and the equivalent_torque must continuously exceed the idle distance threshold. The parameters M and time length can be set according to the sampling period and the typical hysteresis idle distance duration, and can be stored as configuration items, such as "idle_hold_count" and "idle_hold_time". In addition, for direction switching scenarios, the first segment of data after the dir_flag flips can be used as an idle distance candidate segment: when the dir_flag is detected to switch from open to closed (or vice versa), several subsequent sampling points are monitored first. If the criterion of "position unchanged but equivalent drag increased" is met, the entire segment is removed to reduce the effect of hysteresis on the baseline representative value.

[0116] After removing idle data, outlier removal is performed on the remaining motor electrical parameter data samples, and a baseline representative value is determined based on the removed samples. This outlier removal can follow the robustness logic in S102 and above, but it is preferred to perform it under the "valve stem side equivalent resistance caliber" to improve cross-device comparability: for example, within the same bin_id, the equivalent_torque samples are first amplitude-limited and then the median is taken as the baseline representative value, with the sample count and stability indicator output simultaneously. This process can be performed separately on the open-direction baseline subset and the closed-direction baseline subset, thus ensuring that the representative value of each directional channel has undergone hysteresis idle cleanup when generating the direction-sensitive baseline curve.

[0117] For example, in a multi-turn gate valve actuator, the controller reads the load indication and RMS current value from the driver via CAN, and reads the reduction ratio parameter and thread lead parameter from the configuration area. The controller converts the load indication into an equivalent resistance characterization sequence on the valve stem side. After detecting a reversal of direction, if the position changes of several consecutive sampling points are all within the dead zone and the equivalent resistance has exceeded the idle threshold, the record is marked as idle_flag and discarded. Subsequently, the median value of the purified equivalent resistance samples is taken in each opening interval block to form the baseline representative values ​​for opening and closing directions, respectively. Finally, a dual-channel baseline table of "baseline_open_value / baseline_close_value" is output, providing a more stable mechanical background reference for subsequent load deviation extraction and feedforward compensation calculation.

[0118] Regarding the above S105: In one embodiment, the motion feedback data is used to characterize the entire response of the actuator within one actual motion cycle, from the issuance of the target opening command to the entry into the target opening stable zone. This data is used to perform adaptive correction on the aforementioned mechanical baseline resistance characteristics, stroke load characteristics, and the generation parameters of the feedforward compensation control quantity. The motion feedback data may include stroke position data, motor electrical parameter data, and medium differential pressure data of the same caliber as S101. For example, the controller can generate a motion cycle number (cycle_id) at the start of the motion and write it to the motion log record with a unified sampling time base during the motion's duration. Each record can be stored in the form of fields such as "timestamp, cycle_id, stroke_pos, dir_flag, I_rms (or equivalent torque indication), U_bus, P_in (optional), dp, limit_state (optional)"; where limit_state is used to mark whether the open / close limit or the position is reached, facilitating the identification of abnormal motions.

[0119] In one embodiment, the action feedback data may further include an action evaluation result field calculated online by the controller, which supports the engineering closed loop of "updating generated parameters". For example, the controller can statistically analyze the arrival time, arrival error (whether it enters the target opening tolerance zone), peak current, over-limit duration, cumulative energy consumption indication, and current fluctuation in the stable phase after the action within a preset action time window, and write them as "cycle_summary" to the log header or as an independent record entry. The above statistical caliber can be directly implemented by the circular buffer: the controller updates the peak value and cumulative amount during the action, and solidifies the result once at the end of the action for subsequent incremental update modules to call.

[0120] In one implementation, updates can be distributed based on data availability and operating condition consistency to avoid mistakenly including pressurized load contributions in the mechanical baseline resistance characteristics or using abnormal motion data for parameter updates, which could lead to drift. For example, the update target can be determined based on the relationship between dp and the baseline threshold, and whether it is in a segment where the valve structure stage index is the adjustment index (this filtering logic can follow the aforementioned valve structure stage determination approach): when dp is in the baseline operating condition and the motion segment meets the adjustment index, the "stroke_pos—electrical parameter sample" of the corresponding segment is used as the incremental sample of the mechanical baseline resistance characteristics; when dp is in the pressurized operating condition and the motion segment meets the adjustment index, the corresponding segment sample is used as the incremental sample of the stroke load characteristics; and regardless of the dp level, the motion evaluation results can be used for adaptive correction of feedforward compensation control quantity generation parameters (e.g., compensation gain, amplitude limiting template selection weight, candidate optimal threshold, etc.).

[0121] In one implementation, for action segments that meet the baseline operating condition criteria, the controller first maps stroke_pos to bin_id and assigns them to the open or closed cache according to dir_flag. Then, it removes outliers from the incremental samples within each bin_id and calculates the incremental baseline representative value for that bin. Subsequently, it merges and updates the incremental baseline representative value with the currently stored baseline_open_value or baseline_close_value according to a preset weight. The weight can be determined based on the number of incremental samples, the sample stability indicator, and the "aging degree" of the historical baseline version: for example, when the number of effective samples in a bin reaches the minimum sample size and the stability meets the requirements, the update weight of that bin is increased; when the samples are sparse or the volatility is high, the update weight is reduced and only the quality indicator field is updated without updating the numerical field. To avoid control risks caused by abrupt updates, the controller can also set a limit constraint on the magnitude of a single update, such as limiting the change of the baseline representative value in a single update to no more than a preset proportion or no more than a preset absolute step size, and recording an alarm flag when the limit is triggered to prompt maintenance checks (such as lubrication status, packing preload, or mechanical wear).

[0122] In one implementation, the incremental update of the stroke load characteristics can adopt a stroke indexing system consistent with the mechanical baseline resistance characteristics, and under pressurized conditions, the "deviation from the baseline" is used as the update object to ensure cross-cycle comparability. For example, the controller collects electrical parameter samples by bin_id for action segments that meet the pressurized condition criteria, and reads the baseline reference values ​​of the same bin and the same dir_flag channel from the direction-sensitive baseline table to form a deviation sample set; robust statistics are performed on the deviation sample set to obtain the incremental load representative value of that bin, and it is then merged and updated with the load_value in the existing load characteristic table according to a preset weight. If the system enables differential pressure classification dp_bin_id, further bucket updates can be performed within bin_id according to dp_bin_id, thus forming incremental updates under a two-dimensional index; in resource-constrained scenarios, only the "currently operating main differential pressure level" can be updated, while other levels remain unchanged and await subsequent accumulated samples for further updates.

[0123] In one implementation, the update of the generation parameters for the feedforward compensation control quantity can be performed based on the deviation between the action evaluation result and the desired target, enabling the control strategy to adaptively trade off between "response lag" and "redundant energy consumption". For example, the controller can compare reach_time_ms with the upper limit of the target action time, I_peak with the peak current threshold, and energy_idx with the energy consumption benchmark, and accordingly perform incremental corrections on at least one of the following parameters: the compensation gain level boundary in the candidate feedforward compensation control quantity set, the speed template selection weight, the upper limit or switching threshold of the limiting template, and the energy consumption weight / response weight when candidates are preferred. This correction is preferably implemented in the form of "level code / template number / threshold configuration item," for example, when the action is achievable but energy consumption is high, the priority of a certain compensation level is lowered or the peak limiting upper limit is reduced; when the action experiences a timeout or start-up lag, the compensation level in the start-up phase is raised or the low-speed compensation margin is increased.

[0124] For example, the actuator controller is a DSP control board. The driver provides the effective value of the current and the load indication via CAN communication. The pressure before and after the valve is collected via 4–20mA to form dp, and the position encoder provides stroke_pos. The controller writes to the action log during each action cycle and automatically determines whether to update the baseline table or the load table based on dp and the baseline threshold after the action is completed. At the same time, it uses the energy consumption index and arrival time index in cycle_summary to fine-tune the boundary of the candidate feedforward compensation level and solidifies the update result in the format of "parameter version number + effective time".

[0125] In an optional implementation, adaptive updates to the hysteresis backlash identification parameters may also be included. This is because the location and equivalent resistance level of the hysteresis backlash may gradually drift during the operating cycle due to wear, changes in lubrication conditions, or variations in assembly clearances between the worm gear pair and the valve stem thread pair. If the preset dead zone and backlash threshold fixed during the commissioning phase are still used, backlash rejection may result in over-rejection or under-rejection, thus contaminating the incremental samples of subsequent baseline / load / feedforward parameters. Therefore, this implementation first identifies candidate backlash data pairs in the motion feedback data and statistically analyzes their positional distribution. Then, it updates the threshold parameters when the consistency condition is met. Finally, it performs incremental updates to the baseline / load / feedforward parameters under the updated rejection rules.

[0126] In this optional implementation, the controller identifies candidate idle data pairs based on the stroke position data and the valve stem-side equivalent resistance characterization sequence in the motion feedback data. A candidate idle data pair can be understood as "a pair of observations under the same time index," satisfying the condition that the stroke position data remains unchanged within a preset dead zone and the valve stem-side equivalent resistance characterization sequence exceeds an idle threshold. The controller can record candidate idle events as structured entries, such as "cycle_id, timestamp, stroke_pos, bin_id, dir_flag, equivalent_res_value, dp, idle_flag," where idle_flag indicates whether the entry meets the candidate idle criterion; and can additionally record "number of points / duration of unchanged state" when the criterion is met to enhance reliability screening.

[0127] In the action feedback data of at least two action cycles, the controller statistically analyzes the occurrence positions of candidate idle data pairs to obtain idle position distribution characteristics. The "occurrence position" can use the bin_id or stroke_pos interval after stroke binning as the statistical caliber, thus aligning with the aforementioned baseline / load table indexing system. For example, the controller can maintain an idle counter `idle_count` for each bin_id, and maintain `idle_count_open` and `idle_count_close` for open / closed channels respectively; simultaneously, it maintains the action cycle number `cycle_total` and the number of valid sampling points per cycle to normalize the idle count to a comparable scale. The idle position distribution characteristics can include the peak position (the bin_id with the largest count), the peak coverage bandwidth (the cumulative proportion of several bins near the peak), and directional consistency (whether the open and closed peaks overlap or are adjacent), used to characterize whether idles are concentrated in certain stable position segments, thereby supporting threshold adaptation.

[0128] When the idle position distribution characteristics meet the preset consistency conditions, the controller updates the preset position dead zone and / or idle threshold based on the distribution characteristics. The consistency conditions can be set as an engineering-achievable combination of criteria, such as: in at least the most recent two action cycles, the main peak position falls within the same bin_id or adjacent bin_id range; the cumulative proportion near the main peak exceeds the preset proportion threshold; and the total number of candidate idle events reaches the minimum number threshold to exclude occasional noise triggering. The threshold update can be performed in a "from observation statistics to parameter configuration items" manner: for the preset position dead zone, it can be adjusted based on the upper bound of the maximum fluctuation amplitude of stroke_pos in the candidate idle events plus a margin, so that the "remain unchanged" criterion covers the position micro-jitter under the actual hysteresis idle; for the idle threshold, it can be adjusted based on the statistical distribution of equivalent_res_value in the candidate idle events (e.g., taking the representative high quantile level and adding a margin), so that it can better distinguish between "actual force driving position change" and "hysteresis force but no position movement".

[0129] After updating the preset dead zone and / or idle threshold, the controller re-removes idle data based on the updated removal rules, and incrementally updates at least one of the generation parameters of the mechanical baseline resistance characteristics, stroke load characteristics, and feedforward compensation control variables using only the removed motion feedback data. This "re-removal" can be implemented on the controller side via replay: the recorded motion logs are re-marked with idle_flag according to the new threshold, and the corresponding samples are removed; then the remaining samples are added to the baseline / load binning and statistics module. Alternatively, it can be done offline on the host computer side: the host computer reads the motion log file (e.g., CSV / binary log), recalculates idle_flag according to the new threshold, outputs the cleaned sample set, and then sends the updated baseline / load table to the controller.

[0130] For example, in a long-term operation scenario of a multi-turn gate valve, the controller detects that candidate backlash events in the last three operating cycles are concentrated near a certain fixed opening range and are in the same direction, and the total number of events exceeds the minimum number threshold. Based on this, the controller slightly relaxes the position dead zone and raises the backlash threshold, making the hysteresis backlash elimination more closely reflect the current mechanical state. Subsequently, the controller purifies the action samples with the updated elimination rules, and only performs a weighted fusion update of the baseline representative value and the load representative value for bin_id samples that meet the quality requirements. At the same time, it makes a small adjustment to the feedforward compensation candidate optimization threshold, thereby maintaining a stable trade-off between response and energy consumption in long-term operation where mechanical wear and fluid conditions coexist.

[0131] Based on the same inventive concept, this application also provides an electric actuator energy consumption optimization system corresponding to an electric actuator energy consumption optimization method. Since the principle of the system in this application is similar to the electric actuator energy consumption optimization method described above in this application, the implementation of the system can refer to the implementation of the method, and the repeated parts will not be described again.

[0132] Reference Figure 4 The diagram shown is a schematic of an energy consumption optimization system for an electric actuator provided in an embodiment of this application. The system includes: The acquisition module 10 is used to acquire valve body fluid property description information and actuator operation observation data, including stroke position data, motor electrical parameter data and medium pressure difference data. Processing module 20 is used to determine, based on the medium pressure difference data, the observation dataset corresponding to the preset baseline operating condition and the observation dataset corresponding to the preset pressurized operating condition; based on the observation dataset corresponding to the preset baseline operating condition, establish the correspondence between the stroke position data and the motor electrical parameter data to obtain the mechanical baseline resistance characteristics associated with the stroke position; and based on the observation dataset corresponding to the preset pressurized operating condition and the mechanical baseline resistance characteristics, determine the stroke load characteristics associated with the stroke position. The control module 30 is used to determine the feedforward compensation control quantity according to the valve body fluid property description information and the stroke load characteristics, and generate a motor control command based on the feedforward compensation control quantity to control the actuator to change from the current opening degree to the target opening degree. Feedback module 40 is used to acquire motion feedback data and update at least one of the generation parameters of the mechanical baseline resistance characteristic, the stroke load characteristic, and the feedforward compensation control quantity based on the motion feedback data.

[0133] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for optimizing the energy consumption of an electric actuator, characterized in that, include: Obtain valve body fluid property description information and actuator operation observation data, including stroke position data, motor electrical parameter data and medium pressure difference data; Based on the medium pressure difference data, determine the observation dataset corresponding to the preset baseline operating condition and the observation dataset corresponding to the preset pressurized operating condition; Based on the observation dataset corresponding to the preset baseline working condition, a correspondence between the stroke position data and the motor electrical parameter data is established to obtain the mechanical baseline resistance characteristics associated with the stroke position. Based on the observation dataset corresponding to the preset pressurized working condition and the mechanical baseline resistance characteristics, the stroke load characteristics associated with the stroke position are determined; Based on the valve body fluid property description information and the stroke load characteristics, the feedforward compensation control quantity is determined according to the preset energy consumption evaluation criteria, and a motor control command is generated based on the feedforward compensation control quantity to control the actuator to change from the current opening degree to the target opening degree. Acquire motion feedback data, and update at least one of the generation parameters of the mechanical baseline resistance characteristic, the stroke load characteristic, and the feedforward compensation control quantity based on the motion feedback data.

2. The method for optimizing energy consumption of an electric actuator according to claim 1, characterized in that, The preset baseline operating condition is an operating condition where the medium pressure difference is less than or equal to the baseline threshold.

3. The method for optimizing energy consumption of an electric actuator according to claim 1, characterized in that, The preset pressurized operating condition is an operating condition in which the medium pressure difference is greater than the baseline threshold, and the stroke load characteristics are determined by the deviation of the motor electrical parameter data under the preset pressurized operating condition from the mechanical baseline resistance characteristics.

4. The method for optimizing energy consumption of an electric actuator according to claim 1, characterized in that, The preset energy consumption evaluation criterion evaluates the feedforward compensation control quantity based on the energy consumption index determined by the motor electrical parameter data within a preset action time window; wherein, the feedforward compensation control quantity is determined from at least one candidate feedforward compensation control quantity.

5. The method for optimizing energy consumption of an electric actuator according to claim 2, characterized in that, The determination of the observation dataset corresponding to the preset baseline operating condition and the observation dataset corresponding to the preset pressurized operating condition includes: Smoothing is performed on the differential pressure data of the medium, and the fluctuation index and rate of change index of the differential pressure of the medium are calculated within a preset time window to determine the stable differential pressure range that meets the stability criterion. In the pressure differential stability zone, a running segment with a medium pressure differential less than or equal to the baseline threshold is selected, and the stroke position data and motor electrical parameter data synchronously collected in the running segment are aggregated into an observation dataset corresponding to the preset baseline operating condition. In the pressure differential stability zone, a running segment with a medium pressure differential greater than the baseline threshold is selected, and the stroke position data and motor electrical parameter data synchronously collected within the running segment are aggregated into an observation dataset corresponding to the preset pressurized working condition. The differential pressure transition segment that crosses the baseline threshold is removed from the operating segment.

6. The method for optimizing energy consumption of an electric actuator according to claim 5, characterized in that, Before aggregating the observation datasets corresponding to the preset baseline operating conditions and / or the preset pressurized operating conditions, the method further includes: The entire stroke is discretized into multiple opening interval blocks according to the preset stroke step size, and the opening interval blocks corresponding to the stroke positions in the opening and closing directions are combined to form a bidirectional stroke unit. Based on the travel end distance of the bidirectional stroke unit and the rate of change of the motor electrical parameters, the valve structure stage index to which each bidirectional stroke unit belongs is determined. The valve structure stage index includes a disengagement index, an adjustment index, and a re-engagement index. Only the stroke position data and motor electrical parameter data corresponding to the bidirectional stroke units with valve structure stage index as adjustment index are selected for the collection of the observation dataset, and the data corresponding to the bidirectional stroke units with valve structure stage index as disengagement index or in-seat index are removed.

7. The method for optimizing energy consumption of an electric actuator according to claim 1, characterized in that, The method of establishing a correspondence between stroke position data and motor electrical parameter data based on the observation dataset corresponding to the preset baseline operating condition, and obtaining the mechanical baseline resistance characteristics associated with the stroke position, includes: The travel change direction identifier is determined based on the difference sign and / or change gradient sign of the travel position data, and the observation dataset corresponding to the preset baseline condition is split into an open baseline subset and a closed baseline subset according to the travel change direction identifier. The entire stroke is divided into multiple opening interval blocks according to a preset stroke resolution. For the opening baseline subset and the closing baseline subset, motor electrical parameter data samples are collected in each of the opening interval blocks respectively. Within each of the aforementioned opening interval blocks, a baseline representative value is determined for the motor electrical parameter data samples; Based on the travel position of each opening interval block and the corresponding baseline representative value, open-direction mechanical baseline resistance curves and closed-direction mechanical baseline resistance curves are generated respectively, and the open-direction mechanical baseline resistance curves and the closed-direction mechanical baseline resistance curves are combined to form a direction-sensitive mechanical baseline resistance feature.

8. The method for optimizing energy consumption of an electric actuator according to claim 7, characterized in that, Before determining the baseline representative value, the process also includes performing an equivalent conversion on the observed data using the structural parameters of the multi-turn drive train and removing backlash data, including: The actuator is a multi-turn actuator, and its transmission chain includes a worm gear transmission pair and a valve stem thread transmission pair; obtain the transmission parameter set corresponding to the worm gear transmission pair and the valve stem thread transmission pair, the transmission parameter set including reduction ratio parameter and thread lead parameter; Based on the transmission parameter set, the equivalent conversion of the motor electrical parameter data is performed to obtain the valve stem side equivalent resistance characterization sequence. The equivalent conversion includes converting the motor output represented by the motor electrical parameter data into the valve stem side equivalent torque or equivalent axial load. Identify the backlash section caused by the worm gear drive pair and / or the valve stem thread drive pair: when the equivalent resistance characterization sequence exceeds the backlash threshold and the stroke position data remains unchanged within the preset position dead zone, the corresponding data is determined to be backlash data and discarded; Outlier removal is performed on the motor electrical parameter data samples after removing idle data, and the baseline representative value is determined based on the removed samples.

9. The method for optimizing energy consumption of an electric actuator according to claim 8, characterized in that, The step of updating at least one of the generation parameters of the mechanical baseline resistance characteristic, the stroke load characteristic, and the feedforward compensation control quantity based on the motion feedback data includes: Based on the stroke position data in the motion feedback data and the valve stem side equivalent resistance characterization sequence, candidate empty stroke data pairs are identified. The candidate empty stroke data pairs satisfy that the stroke position data remains unchanged within the preset position dead zone and the valve stem side equivalent resistance characterization sequence exceeds the empty stroke threshold. In the motion feedback data of at least two motion cycles, the occurrence positions of the candidate idle data pairs are statistically analyzed to obtain the idle position distribution characteristics. When the empty-range position distribution characteristics meet the preset consistency condition, the preset position dead zone and / or the empty-range threshold are updated based on the empty-range position distribution characteristics. Based on the updated preset position dead zone and / or the idle travel threshold, idle travel data is removed again, and incremental updates are performed on at least one of the mechanical baseline resistance characteristics, the stroke load characteristics, and the generation parameters of the feedforward compensation control quantity using only the removed motion feedback data.

10. An energy consumption optimization system for an electric actuator, characterized in that, include: The acquisition module is used to acquire valve body fluid property description information and actuator operation observation data, including stroke position data, motor electrical parameter data and medium pressure difference data. The processing module is used to determine the observation dataset corresponding to the preset baseline operating condition and the observation dataset corresponding to the preset pressurized operating condition based on the medium differential pressure data. Based on the observation dataset corresponding to the preset baseline working condition, a correspondence between the stroke position data and the motor electrical parameter data is established to obtain the mechanical baseline resistance characteristics associated with the stroke position. Based on the observation dataset corresponding to the preset pressurized working condition and the mechanical baseline resistance characteristics, the stroke load characteristics associated with the stroke position are determined; The control module is used to determine the feedforward compensation control quantity according to the valve body fluid property description information and the stroke load characteristics, and generate motor control commands based on the feedforward compensation control quantity to control the actuator to change from the current opening degree to the target opening degree. The feedback module is used to acquire motion feedback data and update at least one of the generation parameters of the mechanical baseline resistance characteristic, the stroke load characteristic, and the feedforward compensation control quantity based on the motion feedback data.

Images