Self-correcting hybrid power adjusting method based on working condition compensation

Abstract

The invention relates to a self-correction hybrid power regulation method based on working condition compensation. The method comprises the following steps: monitoring and collecting a group of system real-time data streams; accessing a three-dimensional mapping model and indexing a group of load segmentation thresholds matched with the real-time system working conditions; generating a power reference value and a real-time power calculation value respectively by using the real-time data and the load segmentation threshold value; judging whether interval offset occurs or not based on the change trend of the real-time system load, and resetting the real-time power calculation value by using the power reference value when the interval offset occurs; monitoring a correction offset generated during resetting, and performing three-dimensional mapping on the model based on the correction offset; and outputting the reset real-time power calculation value as a final power regulation instruction. The method can utilize the high resolution characteristic of real-time parameters, and also can periodically correct accumulative errors through a reference model capable of being adaptively updated, so that high-precision and long-period stability can be realized at the same time under complex and changeable working conditions.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power regulation of servo hydraulic systems, and in particular to a self-correcting hybrid power regulation method based on working condition compensation. BACKGROUND

[0002] In the field of power regulation of servo hydraulic systems, how to accurately estimate and output power matching the real-time load is the key to improving system performance and energy efficiency. In order to achieve accurate power matching, the current technical path mainly faces a technical contradiction between response accuracy and long-term stability.

[0003] One technical path is to use real-time parameter calculation. This method calculates a high-resolution power value by collecting instantaneous parameters such as servo motor drive current and speed in real time. The advantage of this method is that the response is very sensitive and can capture subtle changes in the load to provide a continuous and smooth power output. However, it is highly dependent on the accuracy of the sensor. In a long time, the sensor's inevitable zero drift, temperature drift, or slight signal noise will be amplified by continuous calculation. This causes the calculated real-time power value to gradually deviate from the system's true power baseline over time, resulting in cumulative errors that affect the system's accuracy over a long period.

[0004] Another technical path is to use a pre-set static model. This method usually establishes a "working condition-load-power" mapping model through calibration before the device is shipped. During operation, the system queries the model to obtain a reference power value. The advantage of this method is that it provides a stable power anchor point that is not affected by instantaneous noise. However, the main problem is that the model is static. The servo hydraulic system is a dynamic physical system, and as the device runs, the viscosity of the hydraulic oil will change with temperature, the pump and valve will leak due to wear, and the system efficiency will decrease. When the device ages, the pre-set static model can no longer reflect the true power demand of the system, making the reference value itself inaccurate.

[0005] In summary, the existing technical paths have their own limitations. Pure real-time calculation is accurate in the short term, but lacks long-term stability; while a pure static model is stable, but lacks long-term accuracy and adaptability. SUMMARY

[0006] In order to take advantage of the high-resolution characteristics of real-time parameters, and at the same time correct cumulative errors through a self-adaptable reference model periodically, thereby achieving high accuracy and long-term stability in complex and variable working conditions, the present application provides a self-correcting hybrid power regulation method based on working condition compensation and related devices.

[0007] In a first aspect, the application provides a self-correcting hybrid power regulation method based on working condition compensation, which adopts the following technical solution: A self-correcting hybrid power regulation method based on working condition compensation, comprising the following steps: S1. Monitoring and collecting a set of system real-time data streams, wherein the system real-time data streams include real-time system working condition, real-time system load, real-time servo motor current and real-time servo motor speed; S2. Accessing a working condition-load-power mapping model, and indexing a set of load segment thresholds matching the real-time system working condition based on the real-time system working condition in the system real-time data streams; S3. Running a parallel power calculation program, which uses the real-time data in the system real-time data streams and the load segment thresholds to generate a power reference value and a real-time power calculation value respectively; S4. Judging whether an interval deviation occurs based on the trend of change of the real-time system load, and resetting the real-time power calculation value using the power reference value when the interval deviation occurs; S5. Monitoring the correction deviation generated when the resetting, and updating the working condition-load-power mapping model based on the correction deviation; S6. Taking the reset real-time power calculation value as the final power regulation instruction output.

[0008] By adopting the above technical solution, the method first monitors and collects system real-time data streams containing real-time working condition, load, motor current and speed, then indexes corresponding load segment thresholds from a preset working condition-load-power mapping model using the real-time working condition in the data streams. Then, a stable power reference value is generated using the indexed thresholds and collected load data, and a high-resolution real-time power calculation value is generated using motor current and speed data. This dual-track parallel calculation architecture enables the system to obtain both reference data for long-period anchoring and real-time data for high-frequency response.

[0009] In order to solve the cumulative error of real-time power calculation value caused by sensor noise or drift, the difference between real-time load and current threshold in a preset period is accumulated, and then a variable decision threshold growing with the period is calculated. By comparing the cumulative sum and the variable threshold, the interval deviation is confirmed. The resetting action of S4 is only triggered when the real-time load has a persistent, non-transient real deviation, so as to effectively filter out the short-term signal fluctuations caused by load transient impact, ensuring the stability of the calibration action.

[0010] To solve the problem of long-term invalidation of the mapping model due to device aging or wear, a correction offset generated at the time of reset is calculated, and then the offset is monitored and accumulated. By judging whether the offset is greater than the aging threshold for a plurality of working cycles, it is identified that there is a permanent deviation between the mapping model and the actual efficiency of the system. Then the mapping model is updated based on the offset. This step constitutes a slow adaptive closed loop, ensuring the accuracy of the power reference value throughout the life cycle of the device.

[0011] Optionally, the sub-step of S3 comprises: S31. Compare the real-time system load with the load segmentation threshold to determine a power reference interval; and obtain the power reference value corresponding to the power reference interval; S32. Calculate and continuously update the real-time power calculation value based on the real-time servo motor current and the real-time servo motor speed.

[0012] By adopting the above technical scheme, by comparing the real-time system load with the load segmentation threshold, a power reference interval is determined, and the power reference value corresponding to the interval is obtained. This method uses the system load as a stable macroscopic representation, so that the power reference value generated by it is not easily disturbed by the instantaneous electrical noise of the motor, providing a reliable anchor point for subsequent fusion calibration. Based on the real-time current and speed of the servo motor, the real-time power calculation value is calculated and continuously updated through the physical model of the motor. This method uses the high dynamic characteristics of motor parameters, so that the generated real-time power calculation value has high resolution and can sensitively track the subtle fluctuations of the load in a short time. Finally, the calculation results of S31 and S32 are combined, and the method simultaneously obtains two power data with different characteristics: one is the stable but low-resolution reference value from S31, and the other is the high-resolution but possibly cumulative error real-time value from S32. The two data complement each other and provide the necessary input for the subsequent fusion calibration of S4.

[0013] Optionally, the sub-step of S4 comprises: S41. Calculate the difference accumulation sum of the real-time system load being lower or higher than the load threshold corresponding to the current power reference interval within a continuously updated preset period; S42. Calculate a variable decision threshold according to the duration or total sampling number of the preset period; S43. Compare the difference accumulation sum with the variable decision threshold, and when the difference accumulation sum is greater than the variable decision threshold, it is confirmed that the interval offset occurs; S44. When the interval offset is confirmed, the reset is performed.

[0014] By adopting the technical solution, the logic first calculates the accumulated sum of the difference between the real-time system load and the current load threshold in a preset time period which is continuously updated. This step integrates the load deviation in time, and the result can reflect the amplitude and duration of the deviation. Meanwhile, the method calculates a variable decision threshold according to the duration of the preset time period or the total number of samples. The threshold dynamically changes with the growth of the observation time, providing a non-fixed reference for judgment.

[0015] By comparing the accumulated sum of the difference with the variable decision threshold, the system can effectively distinguish between real load changes and transient signal interference. The deviation caused by instantaneous load impact or sensor noise is short in duration, and the accumulated sum of the difference is not enough to exceed the variable decision threshold, so it is effectively filtered. Only when the real-time system load has a sustained deviation, causing the accumulated sum of the difference to grow and eventually exceed the variable decision threshold, the system will confirm that an interval deviation has occurred and trigger the subsequent reset action, ensuring the reliability of the calibration.

[0016] Optionally, the sub-step S5 comprises: S51. When the reset is performed, calculate the correction offset between the real-time power calculation value before the reset and the power reference value; S52. Monitor and accumulate the correction offset, and when the correction offset is greater than a preset aging threshold for multiple working cycles, it is determined that the mapping model has failed; S53. When it is determined that the mapping model has failed, perform the update on the mapping model based on the correction offset.

[0017] By adopting the technical solution, when the reset is performed, the correction offset between the real-time power calculation value before the reset and the power reference value is calculated. The correction offset is continuously monitored and accumulated, and the system determines whether the correction offset is greater than a preset aging threshold for multiple working cycles. The purpose of this long-period monitoring is to identify whether the working condition-load-power mapping model itself has a systematic deviation due to device aging or wear, and to determine that the mapping model has failed when the judgment condition is met.

[0018] Once it is determined that the mapping model has failed, the system will perform an update on the mapping model based on the accumulated correction offset. This adds a slow model adaptive closed loop to the original fast fusion calibration closed loop, ensuring that the power reference value as the calibration reference can resist long-term decay, and the entire power regulation method remains accurate throughout the life cycle of the device.

[0019] Optionally, the step S43 further comprises a parallel judgment condition: When the absolute difference between the real-time power calculation value and the power reference value continues to exceed a preset maximum allowed drift for more than a preset time, it is also determined that the interval deviation occurs.

[0020] By using the above technical solution, the comparison of the difference accumulation and the variable determination threshold is mainly used to filter transient fluctuations. However, in the case of a very slow load rise, the growth rate of the difference accumulation may be slower than the growth rate of the variable determination threshold, resulting in a delay in the deviation confirmation. At this time, the real-time power calculation value may continue to deviate from the power reference value, and cannot trigger a reset, resulting in the accumulated error being unable to be eliminated.

[0021] The newly added parallel judgment condition makes up for this deficiency. It monitors the absolute difference between the real-time power calculation value and the power reference value, and determines whether the difference continues to exceed the preset maximum allowed drift. This provides a long-period, absolute amplitude-based trigger for fusion calibration, ensuring that even in slowly changing conditions, when the deviation between the two calculation tracks accumulates to a certain extent, the system can still forcibly trigger the interval deviation and perform a reset, thereby preventing long-term drift of the micro track.

[0022] Optionally, the sub-step of S4 further comprises: After performing the reset once, starting a calibration cooling timer; Before the calibration cooling timer is zeroed, suppressing the next interval deviation judgment to achieve a hysteresis comparison.

[0023] By using the above technical solution, after performing a reset once, the system starts a calibration cooling timer. Before the timer is zeroed, it suppresses the next interval deviation judgment. This introduces a time hysteresis comparison into the system. In the case of a high-frequency oscillation of the system load between two load segment thresholds, the conventional judgment logic may cause the fusion calibration action to be triggered too frequently, causing the power output to rapidly jump between the two reference values.

[0024] The newly added suppression mechanism can effectively avoid this problem. It ensures that after a reset occurs, the system must be stable in the newly determined power reference interval for a period of time before the next deviation judgment is performed. This prevents the final power regulation instruction from being dithered due to consecutive resets, improving the stability of the entire regulation process.

[0025] Optionally, after S1, it further comprises: Performing validity verification on the data in the system real-time data stream, including range checking or rate of change checking; When the verification fails, discarding the current data or entering a failure safety mode.

[0026] By adopting the technical scheme, whether the collected data exceeds a reasonable physical range or whether a change rate thereof exceeds a physical limit is judged through range checking or change rate checking. When verification fails, the system discards the current set of invalid data or enters a failure safety mode when consecutive failures occur. This ensures that abnormal data caused by transient sensor failure or strong signal interference will not be sent to a subsequent parallel power calculation program, thereby preventing the entire power regulation method from producing an incorrect calculation result due to a damaged input data source, and ensuring the robustness of the system.

[0027] To sum up, the present application includes at least one of the following beneficial technical effects: 1. By running a parallel power calculation program, a high-resolution real-time power calculation value and a stable power reference value are generated simultaneously, and a fusion calibration step is used to periodically reset the real-time power calculation value using the power reference value. This dual-track combination and reset mechanism solves the problem of cumulative errors in long-period operation of real-time parameter calculation, enabling the final power regulation instruction to have high dynamic response characteristics while also achieving long-period stability.

[0028] 2. By setting a model adaptive update step, a correction offset is monitored and calculated when the reset is performed. This method uses the correction offset to inversely assess the accuracy of the working condition-load-power mapping model itself, and updates the mapping model based on the offset when the model fails due to device aging or wear. This ensures that the power reference value used for calibration can adapt to long-term changes in device efficiency, keeping the method effective throughout the device's life cycle.

[0029] 3. By using a difference accumulation combined with a variable decision threshold in the fusion calibration step, and supplemented by a hysteresis comparison mechanism implemented by a calibration cooling timer. This design enables the system to effectively filter signal interference caused by transient load shocks or high-frequency oscillations, triggering calibration only for real and sustained load changes. At the same time, the added parallel judgment conditions also ensure the calibration effectiveness of the method in slow load climbing conditions, improving the overall stability and robustness of the system in complex conditions. BRIEF DESCRIPTION OF DRAWINGS

[0030] Figure 1 A flowchart of a self-correcting hybrid power regulation method based on working condition compensation in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0031] The present application will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely intended to explain the present application and are not intended to limit the present application.

[0032] In the following description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the inventive concept. As part of the description, some of the diagrams in the present disclosure are represented in block diagram form to avoid obscuring the disclosed principles. Not all features of a practical implementation are necessary in order to describe the disclosed principles. Further, the language used in the present disclosure has been principally selected for readability and instructional purposes and can not have been selected to delineate or circumscribe the inventive subject matter, resort to the claims being necessary to determine such inventive subject matter. Reference in the specification to "one implementation" or "an implementation" means that a particular feature, structure, or characteristic described is included in at least one implementation and multiple references to "one implementation" or "an implementation" do not necessarily all refer to the same implementation.

[0033] The terms "a," "an," and "the" are not intended to refer to singular entities, but include the general class of which a specific example can be used for illustration. Thus, the use of the term "a" or "an" can mean any number of including "one," "one or more," "at least one," and "one or more than one." The term "or" means any one of the alternatives, as well as any combination of the alternatives, including all of the alternatives, unless the alternatives are expressly indicated to be mutually exclusive. The phrase "at least one of' in combination with a list of items means that a single item from the list can be used, or any combination of the items from the list can be used, unless expressly indicated to the contrary.

[0034] Servo hydraulic system is a common power actuator in industrial equipment, which combines the precision control of servo motor and the power of hydraulic transmission. In injection molding machines, presses and other equipment, servo hydraulic system needs to provide varying pressure and flow in different working stages. The purpose of power regulation is to match the energy output of the servo motor with the actual load demand during system operation. If effective power regulation is not performed, the system tends to run in a fixed power mode, i.e. the motor maintains a high power output even in light load or standby stage.

[0035] Currently, the technical path to achieve power regulation mainly faces the technical contradiction between response accuracy and long-term stability. One technical path is to use real-time parameter calculation, by collecting the instantaneous parameters of the servo motor such as drive current and speed at high frequency, to calculate a high-resolution power value in real time. This method is responsive, but the zero drift of the sensor or signal noise will be amplified over time, causing the calculated real-time power value to gradually deviate from the system's true power reference line, resulting in cumulative error.

[0036] Another technical path is to adopt a pre-set static model. This method establishes a working condition-load-power mapping model through calibration, and queries the model to obtain the reference power value at runtime. This method provides a stable power anchor point, but the model is static. Hydraulic system is a dynamic physical system, and as the equipment runs, component wear will cause system efficiency to decay. When the equipment ages, the pre-set static model can no longer reflect the real power demand of the system, causing the reference value itself to become inaccurate.

[0037] Therefore, the embodiment of the present application discloses a self-correcting hybrid power regulation method based on working condition compensation, comprising the following steps S1 to S6. This method solves the limitations of the above two technical paths by using S3 double-track calculation, S4 fusion calibration and S5 model adaptive update, while obtaining high resolution using real-time parameters, and periodically correcting cumulative errors through an adaptive reference model.

[0038] Reference Figure 1 In particular, the self-correcting hybrid power regulation method based on working condition compensation comprises the following steps S1-S6.

[0039] S1. Monitor and collect a set of system real-time data streams, wherein the system real-time data streams include: real-time system working condition, real-time system load, real-time servo motor current and real-time servo motor speed.

[0040] Each component in the system real-time data stream jointly determines the power output and actual power consumption of the servo hydraulic system. The real-time system working condition, such as the injection, holding or mold opening phase of the injection molding machine, determines the current operating mode of the system. Different operating modes correspond to completely different power demand references and load characteristics, so the real-time system working condition is a key basis for subsequent index mapping model and selection of load segment threshold. The real-time system load, usually represented by the pressure of the hydraulic system or the torque of the servo motor, directly reflects the external resistance that the system needs to overcome to complete the current working condition, and is the main input for calculating the power reference value. Real-time servo motor current and real-time servo motor speed are parameters representing the instantaneous electrical state of the motor, and their product can directly calculate the instantaneous power consumption value of the motor.

[0041] In specific implementation, the collection frequency of the system real-time data stream can be set at a high level, for example, matched with the control period of the servo driver or PLC, such as collecting once every 10 milliseconds, to ensure that the rapid changes of the load and motor state can be captured. The collected data is usually combined into a synchronous data frame for use by the subsequent parallel power calculation program in the same calculation period.

[0042] In some embodiments, the data collection frequency of each data in the data stream can not be consistent. For example, the real-time system condition is event triggered or low frequency change, while the load, current and speed are high frequency sampling. In this case, data alignment processing is needed. A feasible way is to keep the slow changing condition data as the current state value, and at each faster sampling period point, combine it with the latest collected load, current and speed data to form a unified, time-aligned data frame for subsequent calculation.

[0043] S2. Access the condition-load-power mapping model, and based on the real-time system condition in the real-time system data stream, index a set of load segment thresholds matching the real-time system condition.

[0044] The condition-load-power mapping model is a multi-dimensional database pre-calibrated and stored in the controller. It describes the nonlinear correspondence between the system load and the required power reference of the servo hydraulic system under different operating conditions.

[0045] A three-dimensional model is pre-constructed because the relationship between system load and power is not single and fixed, but changes significantly with the change of real-time system condition. For example, in the injection molding machine application, the power required under the injection condition is completely different from that under the holding pressure condition, although the load pressure is 10 MPa. The model relates the data of the three dimensions of real-time system condition, system load and power reference value. In runtime, the method takes the real-time system condition collected in S1 as input, queries this three-dimensional model, and thus indexes or calculates a set of load segment thresholds dedicated to the current condition.

[0046] The advantage of the three-dimensional model is that it can provide a stable and reliable reference. Compared with simply relying on instantaneous current calculation, it is not disturbed by short-term sensor noise, providing a reliable anchor point for subsequent fusion calibration. However, the disadvantage of this pre-set model is that its data is static. It is generated during device calibration and cannot automatically adapt to the system efficiency decay caused by component wear and oil change during long-term operation of the device. When the device ages, the model data may no longer be accurate.

[0047] The method overcomes this disadvantage through subsequent model adaptive updating steps. In the method, the correction offset between the real-time calculation value and the model reference value is continuously monitored. When this offset is identified as a long-term and persistent deviation, the system will update the mapping model in reverse, so that its data can keep up with the actual performance changes of the device, thus ensuring the long-term effectiveness of the reference. Optionally, S2 includes the following sub-steps S21-S23.

[0048] S21. Assign the real-time system condition in the system real-time data stream as a query parameter.

[0049] S22. Retrieve one or more associated base data models from the condition-load-power mapping model using the query parameter.

[0050] S23. Interpolate or fuse the retrieved base data models to resolve an effective set of load segment thresholds under the current condition.

[0051] The real-time system condition in the system real-time data stream refers to the specific working phase that the device is currently in. Taking an injection molding machine as an example, a working cycle includes injection, holding pressure, cooling, mold opening, ejection, and other conditions. Different system conditions correspond to completely different system load and system power requirements. For example, the injection condition requires high load and high power to drive the screw forward; while the holding pressure condition requires high load to maintain pressure, but only requires very low power to compensate for leakage; the mold opening condition is a medium-low load and medium power. If the system power is not adjusted according to the condition and load, for example, maintaining the high power output of the injection condition in the holding pressure condition, it will result in a large amount of power waste and system heating.

[0052] The condition-load-power mapping model is a collection of multiple base data models. The base data model is a load-power correspondence relationship established for a specific, calibrated operating condition. For example, the mapping model stores three base data models corresponding to high-speed injection, standard injection, and low-speed injection, respectively.

[0053] During runtime, the real-time system condition collected by S1 may be a state between two calibrated conditions, such as a medium-speed injection condition. When S22 retrieves from the mapping model using this medium-speed injection condition as a query parameter, it finds that it does not have an exactly matching base data model, but is associated with the high-speed injection and standard injection base data models.

[0054] Therefore, S23 needs to interpolate or fuse the two associated base data models retrieved. The system will weight and average the load threshold data in the two base data models according to the proximity of the current medium-speed condition to the high-speed and standard calibrated conditions. The purpose of this is to dynamically generate a new set of load segment thresholds specifically for the current medium-speed condition, rather than mechanically applying any one calibrated base model, thereby improving the accuracy of subsequent power reference value calculation.

[0055] Optionally, S23 includes the following sub-steps S231-S233.

[0056] S231. Divide the whole load working range of the system into multiple non-equal-width sub-intervals according to the variation of its power response slope.

[0057] S232. In the stable working interval where the power response slope is small, i.e. the load variation has little effect on the power, set a larger number of threshold points to increase the partitioning accuracy of this interval.

[0058] S233. In the interval where the power response slope is large, set a smaller number of threshold points accordingly.

[0059] The division of the load working range is based on the variation of the power response slope. The power response slope represents the relationship between the load variation and the required power variation. The whole range is divided into non-equal-width sub-intervals in order to concentrate the limited computing resources and threshold points in the intervals that need accurate identification.

[0060] In a servo hydraulic system, the power response slope is different in different load intervals. For example, in the pressure maintaining condition of an injection molding machine or the pressure maintaining condition of a press, the system load is in a high but flat platform period. In this stable working interval, the power demand corresponding to the small load variation is very small, i.e. the power response slope is small. On the contrary, in the injection starting or load rapidly rising interval, the small load variation may cause rapid growth of the power demand, i.e. the power response slope is large.

[0061] Since the power variation is not obvious in the stable working interval with small slope, the identification difficulty is increased. Therefore, setting a larger number of threshold points in this interval, i.e. increasing the partitioning accuracy of this interval, can improve the identification ability and estimation accuracy of the system in this core working area. In the interval with large power response slope, the power variation itself is already very obvious, and a smaller number of threshold points can meet the judgment requirements, thereby realizing the optimal allocation of threshold resources.

[0062] Optionally, the steps of the method for constructing the working condition-load-power mapping model include S221-S222.

[0063] S221. In the device calibration stage, test and obtain the corresponding basic load-power characteristic data for multiple representative operating intensities.

[0064] S222. Synthesize the obtained multiple sets of basic characteristic data through a preset fusion algorithm to generate a composite load-power relationship that can represent a specific calibration working condition, and store it in the mapping model.

[0065] The plurality of representative operating intensities are preset modes that simulate different workloads of the device during the calibration process. Taking an injection molding machine as an example, these operating intensities can be defined as a high-speed operating mode, a regular operating mode, and an energy-saving operating mode. These three modes correspond to different power consumption characteristics of the device when pursuing the fastest production rhythm, the standard rhythm, and the lowest energy consumption, respectively.

[0066] Device calibration is to make the device run according to each of the above-mentioned representative operating intensities respectively in a controlled environment. During the running process, the system records the real-time system load and the corresponding actual power consumption of the servo motor at a high frequency, thereby forming a set of basic load-power characteristic data. For example, in the regular operating mode, a load of 10 MPa may correspond to a power of 40 kW; while in the high-speed operating mode, the same load of 10 MPa may need to consume a power of 45 kW due to higher requirements for dynamic response. These data sets are the basic load-power characteristic data.

[0067] After obtaining a plurality of sets of basic characteristic data, the system will synthesize them through a preset fusion algorithm. For example, when constructing a calibration working condition model for standard production, the fusion algorithm may assign a weight of 0.3 to the data of the high-speed operating mode and a weight of 0.7 to the data of the regular operating mode, and then perform weighted averaging. This is done to generate a composite load-power relationship. Because in actual production, the working condition of the device is rarely a single case or always a certain standard case, it is usually an intermediate state of different working conditions or changes. Through this fusion processing, the composite relationship stored in the mapping model can more truly reflect the average or comprehensive performance of the device under a certain calibration working condition, rather than relying on single and rigid characteristic data, thereby improving the universality of the model in actual application.

[0068] Optionally, the sub-steps of S221 include S2211-S2212.

[0069] S2211. Set the representative operating intensities for calibration to simulate an alternating cycle mode that simulates actual application, and this cycle mode includes different stages such as high load, medium load, and low load.

[0070] S2212. During calibration, make the system run according to these alternating cycle modes, and collect the dynamic response data of load and power during this process, to ensure that the basic data used to construct the model can reflect the true characteristics of the system when the load is switched.

[0071] This calibration method defines the operating intensity as an alternating cycle pattern that simulates the actual application, rather than a series of static, constant load points. The purpose of step S2211 is to make the calibration process closer to the real workflow of the device. Take an injection molding machine as an example, a representative alternating cycle pattern can be set as: first perform a high-load injection phase for 3 seconds, then a medium-load holding phase for 5 seconds, and finally a low-load cooling standby phase for 10 seconds.

[0072] Step S2212 is to collect dynamic response data of load and power during the cycle pattern defined in step S2211. Continuing with the injection molding machine example, the system will record how the power responds and stabilizes at the moment of switching from high-load injection to medium-load holding. This is done to ensure that the basic data used to build the model not only contains the load-power relationship of the device under stable working conditions, but also contains the transient response characteristics when the working conditions are switched, such as when the load is suddenly applied or removed. If only static data points are used for calibration, the model built will not accurately reflect the real power demand of the system during these dynamic switching processes.

[0073] S3. Run a parallel power calculation program that uses real-time data from the system real-time data stream and the load segment threshold to generate a power reference value and a real-time power calculation value, respectively.

[0074] The parallel power calculation program refers to an algorithm process designed to perform two independent calculation tasks simultaneously to obtain two different characteristics but related power data. The purpose of running this program is to solve the limitations of a single technical path in estimating power, i.e., high-precision estimation lacks long-term stability, and stable models lack real-time high-resolution.

[0075] The power reference value is generated through the first calculation track of this program. It is usually based on the system load and the pre-set mapping model, representing a relatively stable power anchor point that the system should theoretically have under the current working condition and load. It is not easily affected by motor electrical noise or sensor transient drift.

[0076] The real-time power calculation value is generated through the second calculation track of this program. It is usually calculated based on the real-time current and speed of the servo motor, representing the actual power consumption of the motor at the current moment. It has high resolution and high dynamic response characteristics, and can sensitively reflect the subtle fluctuations of power. The power reference value is used as a reliable, non-accumulative error calibration standard to reset the real-time power calculation value, which has high resolution but may have long-term drift.

[0077] Optionally, in some embodiments, the sub-steps of S3 include S31-S32.

[0078] S31. Comparing the real-time system load with the load segment threshold to determine a power reference interval; and obtaining the power reference value corresponding to the power reference interval.

[0079] S32. Calculating and continuously updating the real-time power calculation value based on the real-time servo motor current and the real-time servo motor speed.

[0080] The two sub-steps define two independent calculation tracks of the parallel power calculation program in detail. S31 uses the system load as a macro physical quantity to compare it with the load segment threshold indexed from the mapping model. The power reference interval and the corresponding power reference value determined in this way have high stability and are not easily affected by motor electrical noise or transient signal fluctuations. It provides a reliable anchor point for the subsequent calibration step.

[0081] The principle of S32 is to use micro motor operating parameters. It calculates the instantaneous actual power consumption, i.e. the real-time power calculation value, by collecting real-time servo motor current and speed according to the physical model of the motor. This value has high resolution and fast dynamic response characteristics, and can sensitively track the subtle changes in power, but it may also accumulate measurement errors of the sensor over a long period of time.

[0082] Taking the holding pressure working condition of an injection molding machine as an example, S31 detects that the real-time system load is 9 MPa at this time, and compares it with the load segment threshold in the holding pressure working condition to determine that it is currently in the "holding pressure-medium" interval, and obtains the corresponding power reference value as 15 kW. At the same time, the second sub-step S32 measures the motor current as 20 A and the speed as 700 rpm, and calculates the real-time power calculation value as 15.2 kW. If the current sensor drifts slightly after a period of operation, the real-time value calculated by S32 may become 15.9 kW, but the judgment result of S31 based on the load is still 15 kW.

[0083] S4. Judging whether an interval shift occurs based on the trend of the real-time system load, and resetting the real-time power calculation value using the power reference value when the interval shift occurs.

[0084] Interval shift refers to the persistent migration of real-time system load due to working condition advancement, which crosses the preset load segment threshold. The real-time power calculation value will continuously accumulate small measurement deviations of the servo motor current or speed sensor over a long period of time, causing its calculation result to gradually deviate from the true power level, i.e. calculation drift occurs.

[0085] The purpose of resetting the real-time power calculation value with the power reference value is to eliminate the accumulated error. The power reference value is derived from the comparison between the load and the mapping model, which does not have the accumulated error characteristic and represents a stable and reliable anchor point in a long period. When the interval shift is determined, the system uses this anchor point as a reference to force the real-time power calculation value that has drifted to return to this accurate reference. This eliminates the accumulated error of the real-time calculation value periodically without sacrificing the high resolution characteristic of the real-time calculation value, and ensures the long-period accuracy of the final output result.

[0086] Optionally, in some embodiments, the sub-steps of S4 include S41-S44.

[0087] S41. Calculate the accumulated sum of the difference between the real-time system load and the load threshold corresponding to the current power reference interval in a continuously updated preset period.

[0088] S42. Calculate a variable determination threshold according to the duration of the preset period or the total number of samples.

[0089] S43. Compare the accumulated sum of the difference with the variable determination threshold, and when the accumulated sum of the difference is greater than the variable determination threshold, determine that the interval shift has occurred.

[0090] S44. When the interval shift is determined, perform the resetting.

[0091] The purpose of S41-S44 is to provide a reliable dynamic filter for the determination of the interval shift to distinguish between real load changes and transient signal disturbances. The principle is that the accumulated sum of the difference calculated in the first step is an integral value that measures both the load deviation amplitude and the duration. The variable determination threshold calculated in the second step is a dynamic reference that increases with time or the number of samples.

[0092] By comparing the two values, the system can effectively filter out transient impacts. A short-term load fluctuation may produce a large instantaneous deviation, but its duration is short, so the accumulated sum of the difference is not enough to exceed the variable determination threshold calculated in the same short period of time, and the system will not determine the shift. Only when the real-time system load has a sustained change, for example, entering the next working condition, the growth rate of the accumulated sum of the difference will be faster than that of the variable determination threshold, and eventually exceed the threshold in comparison. At this time, the system determines that an effective interval shift has occurred and triggers the resetting action.

[0093] Continuing with the example of an injection molding machine, assume that the system is currently in a holding pressure condition with a power reference value of 15 kW. At this time, a transient disturbance occurs in the load, causing the real-time system load to temporarily deviate from the load threshold for 0.1 seconds. The system calculates a difference sum of 50. However, according to the 0.1 second duration, the variable decision threshold is calculated to be 60. Since 50 is less than 60, the system does not recognize the deviation, thus filtering the disturbance. Subsequently, the holding pressure condition ends and the system enters a lower load condition of cooling. The real-time system load continues to be below the load threshold of the holding pressure condition. After 1.5 seconds, the difference sum is calculated to be 800. At this time, according to the 1.5 second duration, the variable decision threshold is calculated to be 300. Since 800 is greater than 300, the system recognizes a true interval deviation and triggers a reset action.

[0094] Optionally, in some embodiments, the step S43 further comprises a parallel judging condition: When the absolute difference between the real-time power calculation value and the power reference value continuously exceeds a preset maximum allowed drift, the interval deviation is also recognized.

[0095] The purpose of this parallel judging condition is to provide a supplementary triggering mechanism for fusion calibration to deal with specific working conditions where the main judging logic may fail. The principle is that the judging logic of S41 to S43 relies on the change of real-time system load to calculate the difference sum. This logic may not be able to trigger effectively when the load slowly rises or the real-time power calculation value itself slowly deviates due to sensor drift. Because the difference sum on the load side may grow slowly and always fail to exceed the variable decision threshold.

[0096] The newly added parallel condition solves this problem. Instead of monitoring the change of load, it directly compares the results of the two power calculation tracks, i.e. the absolute difference between the real-time power calculation value and the power reference value. If this absolute difference, which represents the magnitude of accumulated error, continuously exceeds a preset maximum allowed drift, the system will also recognize an interval deviation.

[0097] Continuing with the example of the holding pressure condition with a power reference value of 15 kW. Assume that the load does not change, but the current sensor drifts due to temperature drift, causing the real-time power calculation value of S32 to slowly drift from 15.2 kW to 17.5 kW in 60 seconds. At this time, the difference sum of S41 (based on load) is zero and cannot trigger a deviation. However, this parallel condition will calculate an absolute difference of 2.5 kW (17.5 kW - 15 kW). If the maximum allowed drift is set to 2 kW, the system will recognize a deviation according to the parallel condition and forcibly reset 17.5 kW back to 15 kW, thus preventing unlimited drift of the calculation value.

[0098] Optionally, in some embodiments, the sub-step of S4 further comprises: After performing the reset once, start a calibration cooling timer; Before the calibration cooling timer expires, suppress the next determination of the interval deviation to achieve hysteresis comparison.

[0099] The purpose of this step is to prevent the power regulation command from being unstable due to high-frequency oscillation of the load. The principle is that if the real-time system load fluctuates slightly around a load segment threshold, the main determination logic may be continuously triggered at high frequency, causing the reset action to frequently switch between two adjacent power reference values, resulting in system output jitter.

[0100] The newly added step introduces a time hysteresis. After the system performs a reset once, a calibration cooling timer, for example 1.5 seconds, is started. Before the timer expires, even if the real-time system load crosses back to the original threshold, the determination of the interval deviation will be suppressed, so that the next reset will not be triggered.

[0101] Continuing with the example of an injection molding machine, assume that the power reference value is 15kW. When the load temporarily exceeds the threshold, the system triggers a reset, the power reference value is updated to 20kW, and a 1.5 second cooling timer is started. If the load fluctuates back below the original threshold after 0.1 seconds, the determination logic is suppressed because the timer has not expired, and the system will maintain the reference value of 20kW. This forces the system to run stably in the new reference interval for at least 1.5 seconds, effectively filtering out high-frequency load disturbances, achieving hysteresis comparison, and ensuring the stability of the power output.

[0102] S5. Monitor the correction offset generated when the reset is performed, and update the working condition-load-power mapping model based on the correction offset.

[0103] The correction offset is the difference between the real-time power calculation value and the power reference value when the reset action is performed. This value will change as the device state changes. For example, a new device in the holding pressure working condition, the power reference value is 15kW, the real-time power calculation value before the reset may be 15.2kW, and the correction offset is only 0.2kW. After two years of operation, due to the wear of the hydraulic pump, the system needs to consume more energy to maintain the same holding load. At this time, the power reference value (based on load determination) is still 15kW, but the real-time power calculation value (based on actual motor parameters) may have drifted to 17.0kW before the reset, and the correction offset becomes 2.0kW.

[0104] The working condition-load-power mapping model is updated based on this correction offset, because this offset, especially when it persists, objectively reflects the deviation between the mapping model and the actual characteristics of the device. The mapping model is static, while the device performance is dynamically changing. By this step, the correction offset is fed back to the mapping model and updates it, which constitutes a slow adaptive closed loop. The purpose of this is to ensure that the power reference value, which is the basis for calibration, itself can be corrected to keep up with the performance degradation of the device due to aging and wear, thereby ensuring the accuracy of the entire method throughout the life cycle of the device.

[0105] Optionally, in some embodiments, the sub-steps of S5 include S51-S53.

[0106] S51. At the reset, calculate the correction offset between the real-time power calculation value before the reset and the power reference value.

[0107] S52. Monitor and accumulate the correction offset, and when the correction offset persists to be greater than a preset aging threshold in multiple working cycles, it is determined that the mapping model has failed.

[0108] S53. When it is determined that the mapping model has failed, update the mapping model based on the correction offset.

[0109] The purpose of S51-S53 is to provide a specific and reliable judgment logic for model adaptive update, to distinguish between transient, random offsets and permanent offsets caused by device aging. The principle is that S51 precisely defines the calculation timing and method of the correction offset. S52 is a long-period filter. It will not update the model immediately because of a single large offset, but will monitor and accumulate this offset. It confirms that the deviation is permanent by judging whether the offset is greater than a preset aging threshold (e.g. 1.5kW) in multiple working cycles (e.g. 1000 consecutive cycles). Only when this strict condition is met, it is determined that the mapping model has failed. S53 is the update execution action, which only updates the mapping model based on the long-term accumulated correction offset after S52 confirms the failure.

[0110] Continuing with the example of device aging, S51 calculates the current correction offset as 2.0 kW. S52 starts monitoring, and in the subsequent multiple cycles, the calculated offsets are 1.9 kW, 2.1 kW, 2.0 kW, respectively. Assuming the aging threshold is 1.5 kW, and the monitoring period is 1000 cycles. After 1000 cycles, the system confirms that the average offset 2.0 kW is consistently greater than the aging threshold 1.5 kW, and thus concludes that the mapping model has failed. S53 immediately updates the power reference value 15.0 kW in the mapping model for the holding mode condition by adding the average correction offset 2.0 kW, to 17.0 kW. In the subsequent operation, the new power reference value becomes 17.0 kW, and the correction offset will regress back to near zero.

[0111] S6. Output the reset real-time power calculation value as the final power regulation command.

[0112] This step is the final execution link of the entire power regulation method. Its purpose is to send the power value that has both high resolution and long-term accuracy, obtained through the S3 double-track calculation and the S4 fusion calibration reset, as a command to the controller of the servo hydraulic system. The principle is that the output value is the real-time power calculation value. It maintains the characteristics of high resolution, enabling smooth and sensitive power output. But it has been forcibly reset by the power reference value in S4, so it has eliminated the drift accumulated by long-term operation. The value output by S6 ensures that the command sent to the hardware is both real-time and accurate.

[0113] Continuing with the example of the holding mode condition, when the real-time power calculation value calculated by S32 reaches 17.5 kW due to drift, the judgment logic of S4 is triggered, and it performs a reset on it using the power reference value of 15.0 kW. At this time, the real-time power calculation value is updated to 15.0 kW. S6 step immediately takes this 15.0 kW value as the final power regulation command and outputs it to the servo drive module, making the actual power of the motor return from the wrong 17.5 kW to the accurate 15.0 kW level.

[0114] Optionally, in some embodiments, after S1, it further includes: Performing validity verification on the data in the real-time data stream of the system, including range check or rate of change check; When the verification fails, discard the current data or enter a failure safety mode.

[0115] The purpose of this step is to guarantee the data quality for the following calculation, and prevent abnormal data caused by sensor instantaneous failure or strong signal interference from entering the parallel power calculation program. The principle is that the data collected by S1 is pre-processed before being used for calculation. The range check will determine whether the collected value exceeds the preset physical range. For example, if a load sensor designed for 25 MPa has a reading of 50 MPa or -1 MPa, the data is considered invalid. The rate of change check determines whether the change between the current data point and the previous data point exceeds the physical limit. For example, if the servo motor current suddenly changes from 20A to 200A in 10ms, such a rate of change is considered impossible, and the data is also considered invalid.

[0116] When the verification fails, the system will take action according to the severity of the error. If it is only a single data point instantaneous failure, the system will discard the current set of invalid data, and use valid data in the next sampling period to prevent a single disturbance from contaminating the calculation. If the verification fails continuously, for example, 100 consecutive samplings have failed, it indicates that the sensor may have been permanently damaged. At this time, the system will enter the failure safety mode, stop executing the power regulation method, and switch to a preset fixed safety power output to prevent the device from losing control due to erroneous data.

[0117] It should be understood that the size of the serial number of each step in the above embodiments does not mean the order of execution, and the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0118] The above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A self-calibrating hybrid power regulation method based on operating condition compensation, characterized in that, Includes the following steps: S1. Monitor and collect a set of real-time system data streams, wherein the real-time system data streams include: real-time system operating conditions, real-time system load, real-time servo motor current, and real-time servo motor speed; S2. Access the operating condition-load-power mapping model, and based on the real-time system operating condition in the real-time system data stream, index a set of load segmentation thresholds that match the real-time system operating condition; S3. Run a parallel power calculation program, which uses real-time data in the system's real-time data stream and the load segmentation threshold to generate a power reference value and a real-time power calculation value, respectively. S4. Determine whether an interval offset has occurred based on the changing trend of the real-time system load, and when the interval offset occurs, reset the real-time power calculation value using the power reference value; S5. Monitor the correction offset generated during the reset, and update the operating condition-load-power mapping model based on the correction offset; S6. Use the reset real-time power calculation value as the final power adjustment command output.

2. The self-calibrating hybrid power regulation method based on operating condition compensation according to claim 1, characterized in that, The sub-step of S3 includes: S31. Compare the real-time system load with the load segmentation threshold to determine a power reference interval; and obtain the power reference value corresponding to the power reference interval; S32. Calculate and continuously update the real-time power calculation value based on the real-time servo motor current and the real-time servo motor speed.

3. The self-calibrating hybrid power regulation method based on operating condition compensation according to claim 2, characterized in that, The sub-step of S4 includes: S41. Within a continuously updated preset time period, calculate the cumulative sum of the differences between the load thresholds corresponding to the current power reference range when the real-time system load is lower or higher than the load threshold. S42. Calculate a variable judgment threshold based on the duration of the preset time period or the total number of samples; S43. Compare the sum of the differences with the variable decision threshold. If the sum of the differences is greater than the variable decision threshold, then confirm that the interval shift has occurred. S44. When the interval offset is confirmed, the reset is performed.

4. The self-calibrating hybrid power regulation method based on operating condition compensation according to claim 1, characterized in that, The sub-step of S5 includes: S51. During the reset, calculate the correction offset between the real-time power calculation value before the reset and the power reference value; S52. Monitor and accumulate the correction offset. When the correction offset is continuously greater than a preset aging threshold in multiple working cycles, the mapping model is deemed to have failed. S53. When it is determined that the mapping model has failed, the update is performed on the mapping model based on the correction offset.

5. The self-calibrating hybrid power regulation method based on operating condition compensation according to claim 3, characterized in that, Step S43 further includes a parallel judgment condition: When the absolute difference between the real-time power calculation value and the power reference value continuously exceeds a preset maximum allowable drift amount, the interval offset is also confirmed to have occurred.

6. The self-calibrating hybrid power regulation method based on operating condition compensation according to claim 3, characterized in that, The sub-step of S4 further includes: After performing the reset once, a calibration cooling timer is started; Before the calibration cooling timer returns to zero, the next determination of the interval offset is suppressed to achieve hysteresis comparison.

7. The self-calibrating hybrid power regulation method based on operating condition compensation according to claim 1, characterized in that, Following S1, the following is also included: Perform validity verification on the data in the real-time data stream of the system, including range checks or rate of change checks; If verification fails, discard the current data or enter fail-safe mode.

Images