An EC adaptive disturbance compensation predictive control method and system for an integrated water and fertilizer machine
By adopting an adaptive disturbance compensation predictive control method, the problem of disturbance observation mismatch in EC predictive control of the integrated water and fertilizer machine is solved, achieving higher EC control accuracy and response speed, and adapting to disturbance compensation under complex working conditions.
Patent Information
- Application Number
- CN202611599015.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-15
- Publication Date
- 2026-11-17
AI Technical Summary
Existing EC predictive control systems for integrated water and fertilizer systems struggle to distinguish between residual noise and actual disturbances under complex operating conditions. The intensity of disturbance observations does not match the sensitivity of mixing, and the degree of disturbance retention in the prediction time domain cannot adapt to changes, resulting in insufficient EC control accuracy and response speed.
An adaptive disturbance compensation predictive control method is adopted. By estimating the residual noise level, residual persistence reliability and mixing sensitivity online, the disturbance observation gain and the predicted time-domain disturbance attenuation coefficient are dynamically adjusted to optimize the fertilizer injection pump opening and achieve EC closed-loop regulation.
It improves the accuracy and response speed of EC control, reduces the probability of random noise compensation, enhances the compensation speed for continuous disturbances, ensures that the disturbance compensation intensity matches the current mixed state, and avoids control commands from exceeding the range of the actuator.
Smart Images

Figure CN123113399A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of integrated water and fertilizer control technology, specifically to an EC adaptive disturbance compensation predictive control method and system for an integrated water and fertilizer machine. Background Technology
[0002] Integrated water and fertilizer management technology is an agricultural production method that combines irrigation water supply with the delivery of soluble fertilizers. Integrated water and fertilizer machines typically provide basic irrigation flow through a clean water branch and inject a high-concentration mother liquor into the clean water at a specific ratio via a fertilizer injection pump, forming a nutrient solution of the target concentration in the mixing unit. Electrical conductivity (EC) comprehensively reflects the overall concentration of soluble ions in the nutrient solution and is a crucial parameter for online regulation of fertilizer concentration in integrated water and fertilizer machines. The accuracy of EC regulation directly affects the stability of fertilizer concentration and the consistency of fertilization across different irrigation sections.
[0003] In existing integrated water and fertilizer systems, EC control mostly employs feedback control methods such as PID or fuzzy PID. This type of method adjusts the fertilizer pump drive based on the deviation between the target EC and the measured EC. It has a relatively simple structure and can achieve basic closed-loop regulation when the clean water flow rate, mother liquor concentration, and equipment parameters are relatively stable. However, PID control inherently relies on feedback correction after an error occurs. When the clean water flow rate changes abruptly or the mother liquor concentration changes, the controller needs to wait for the measured EC value to deviate from the target before compensating, which easily leads to significant dynamic lag and persistent deviation.
[0004] In actual water and fertilizer regulation, the factors affecting EC (Electrical Control Unit) are not singular disturbances. The flow rate of clean water fluctuates due to pump pressure, valve switching, and changes in irrigation zones; the concentration of mother liquor changes with fertilizer mixing errors, storage conditions, and mother liquor consumption; the flow gain and dynamic time constant of the fertilizer injection pump have calibration errors; there is a certain mixing lag after the clean water and mother liquor enter the mixing unit; and the EC sensor itself has measurement delay, random noise, and zero-point bias. When all these factors are present simultaneously, a fixed-parameter feedback controller struggles to simultaneously achieve rapid response, steady-state accuracy, and noise immunity.
[0005] Model predictive control (MDI) can predict the EC (excessive energy) over several future sampling periods using a model of the controlled object, and explicitly handles constraints on the opening degree and rate of change of the fertilizer injection pump through rolling optimization. Therefore, it is suitable for dynamic EC control of integrated water and fertilizer systems. However, conventional MDI is highly dependent on the accuracy of the control-side model. When the fertilizer injection pump flow gain, pump time constant, and mixing time constant used by the controller deviate from the actual object, or when the real-time EC of the mother liquor cannot be accurately measured, a continuous deviation will form between the nominal EC prediction value and the actual EC, thereby reducing the regulation effect of MDI.
[0006] To mitigate the impact of model mismatch and unmeasured disturbances on predictive control, the total disturbance can be estimated using the output residual between the measured EC and the nominal EC, and then compensated for in the predictive model. However, traditional disturbance observation compensation typically employs a fixed disturbance observation gain, a fixed residual dead zone, and a fixed prediction time-domain disturbance preservation coefficient. When the fixed observation gain is large, EC sensor noise can easily be mistaken for real disturbances, leading to overcompensation; conversely, a small fixed observation gain reduces the tracking speed for continuous disturbances such as changes in mother liquor concentration. Furthermore, when the fertilizer injection ratio changes, the sensitivity of the mother liquor-side disturbance to the actual impact on the mixed EC also changes, and a fixed compensation strength cannot reflect the physical sensitivity differences in the water-mother liquor mixing process.
[0007] Therefore, existing EC predictive control for integrated water and fertilizer systems still suffers from problems under complex operating conditions, such as difficulty in distinguishing residual noise from actual continuous disturbances, mismatch between the observed disturbance intensity and the current mixing sensitivity, and the inability of the disturbance persistence level in the prediction time domain to adaptively change according to the disturbance persistence. An EC predictive control scheme is needed that can adjust the disturbance estimation and prediction compensation intensity online based on the residual quality and the physical state of water and fertilizer mixing. Summary of the Invention
[0008] To address the shortcomings of existing technologies, this invention provides an EC adaptive disturbance compensation predictive control method and system for integrated water and fertilizer machines, which solves the technical problems existing in the current EC closed-loop control of integrated water and fertilizer machines, such as model mismatch, insufficient continuous disturbance compensation, and easy miscompensation under sensor noise conditions.
[0009] In a first aspect, the EC adaptive disturbance compensation predictive control method for integrated water and fertilizer machines provided by the present invention includes: Obtain the target EC setpoint, the measured EC value of the mixture, the clear water flow rate, and the fertilizer pump control volume of the previous sampling cycle; The nominal fertilizer injection flow rate and nominal EC value are updated based on the target EC setpoint, clean water flow rate, and the execution drive quantity of the previous sampling period; The output residual is constructed based on the measured EC value and the nominal EC value of the mixture, and the residual noise level is estimated online based on the output residual increment. An adaptive dead zone is generated based on the residual noise level to obtain the effective residual. The reliability of residual persistence is determined based on the co-directional persistence of the current effective residual and the effective residual of the previous sampling period, and the residual amplitude factor is determined based on the effective residual amplitude. The mixing sensitivity is determined based on the nominal fertilizer injection flow rate and the clean water flow rate, and a sensitivity gating coefficient is generated based on the mixing sensitivity. The adaptive perturbation observation gain is determined based on the residual persistence confidence, residual magnitude factor, and sensitivity gating coefficient, and the total perturbation estimate is updated. The prediction time-domain disturbance attenuation coefficient is determined based on the residual persistence reliability, a total disturbance prediction sequence in the prediction time domain is generated, and the total disturbance prediction sequence is superimposed on the nominal EC prediction sequence. The objective function is constructed using the tracking error between the target EC and the compensated EC prediction value, as well as the increment of the driving force, and the optimal control sequence is solved under constraints. Adjust the opening degree or equivalent driving quantity of the fertilizer injection pump according to the first control quantity of the optimal control sequence, so that the mother liquor is injected into the clean water-mother liquor mixing unit at the corresponding flow rate, thereby realizing the EC closed-loop regulation of the mixed liquor.
[0010] Optionally, the nominal EC value is updated using a first-order hybrid dynamic update, specifically as follows: , The sampling period is The nominal mixing time constant on the control side. For nominal hybrid EC, For instantaneous mixing EC, For clean water flow, For Qingshui EC, To control the nominal fertilizer injection flow rate on the control side, The nominal mother liquor EC used in the controller; The instantaneous mixing EC is calculated according to the following formula: , The nominal fertilizer injection flow rate on the control side is updated dynamically using the first-order method of the fertilizer injection pump. , To control the nominal fertilizer injection pump time constant on the control side, To control the nominal flow gain of the fertilizer injection pump on the control side, This refers to the opening degree of the fertilizer injection pump or its equivalent driving force.
[0011] Optionally, the steps for obtaining the effective residual include: Construct the output residual: , The measured EC value of the mixture is shown. Estimate residual noise level: , The weighting coefficient for the residual noise index; Generate adaptive dead zone: , To achieve the minimum adaptive dead zone, This is the noise amplification factor; Then when Less than At that time, the effective residual Set to 0; when Not less than At that time, let the effective residual = .
[0012] Optionally, the residual persistence confidence is updated according to the following formula: ; in, The reliability memory coefficient; when the effective residual and When all values are non-zero and have the same sign, the state s(k) continues in the same direction, s(k) = 1; otherwise, s(k) = 0. The residual magnitude factor is: ; Sat(·) is the normalization scale for residual magnitudes, which means restricting the variable to a given upper and lower limit.
[0013] Optionally, the mixed sensitivity is: , Then the sensitivity gating coefficient , For reference sensitivity, This is the lower limit of the sensitivity gate.
[0014] Optionally, the steps for obtaining the total disturbance estimate include: The confidence-amplitude composite weight is calculated based on the residual persistence confidence and the residual amplitude factor: , Basic weights; Calculate the adaptive perturbation observation gain: , This is the lower limit of the gain limit. This is the upper limit of the perturbation observation gain; Update the total perturbation estimate using adaptive perturbation observation gain: , This represents the upper limit of the estimated amplitude of the total disturbance.
[0015] Optionally, the steps for obtaining the nominal EC prediction sequence include: The prediction time-domain perturbation attenuation coefficient is adaptively determined based on the residual persistence reliability: , , These are the upper and lower limits for predicting the attenuation coefficient of the time-domain disturbance; In the i-th prediction step in the prediction time domain, the total perturbation prediction is propagated according to the following formula: , The total disturbance prediction is added to the nominal EC prediction: .
[0016] Optionally, the objective function is: , in, For EC tracking error weights, To control the incremental weight, Representing the prediction time domain, the terminal tracking error term is based on... Sure, This refers to the terminal weighting coefficient; The constraints are , ; At each sampling time k, the optimal control increment sequence is obtained by minimizing the objective function J to solve for the control increment sequence in the predictive control time domain. ; Take the optimal control increment sequence China's first control increment and in accordance with Calculate the current driving quantity of the fertilizer injection pump.
[0017] Secondly, the EC adaptive disturbance compensation and prediction control system for the integrated water and fertilizer machine provided by the present invention includes: The information acquisition unit is used to acquire the target EC setpoint, the measured EC value of the mixed liquor, the clear water flow rate, and the fertilizer injection pump control volume of the previous sampling cycle. The central control unit includes a nominal model state update module, a residual noise estimation and adaptive dead zone module, a residual persistence reliability calculation module, a hybrid sensitivity gating module, a total disturbance observation module, a prediction compensation module, and a model prediction optimization module. The nominal model state update module is used to update the nominal fertilizer injection flow rate and nominal EC value based on the target EC setpoint, clean water flow rate, and the execution driving quantity of the previous sampling period; The residual noise estimation and adaptive dead zone module is used to construct the output residual based on the measured EC value and the nominal EC value of the mixture, estimate the residual noise level online based on the output residual increment, generate an adaptive dead zone based on the residual noise level, and obtain the effective residual. The residual persistence reliability calculation module is used to determine the residual persistence reliability based on the same-direction persistence of the current effective residual and the effective residual of the previous sampling period, and to determine the residual amplitude factor based on the effective residual amplitude. A mixing sensitivity gating module is used to determine the mixing sensitivity based on the nominal fertilizer injection flow rate and the clean water flow rate, and to generate a sensitivity gating coefficient based on the mixing sensitivity. The total disturbance observation module is used to determine the adaptive disturbance observation gain based on the residual persistence confidence, residual magnitude factor and sensitivity gating coefficient, and update the total disturbance estimate. The prediction compensation module is used to determine the prediction time-domain disturbance attenuation coefficient based on the residual persistence reliability, generate the total disturbance prediction sequence in the prediction time domain, and superimpose the total disturbance prediction sequence onto the nominal EC prediction sequence. The model prediction optimization module is used to construct the optimal control sequence under constraints by using the tracking error between the target EC and the compensated EC prediction value and the increment of the execution driving quantity as the objective function. The execution drive unit is used to adjust the opening degree or equivalent drive quantity of the fertilizer injection pump according to the first control quantity of the optimal control sequence, so that the mother liquor is injected into the clean water-mother liquor mixing unit at the corresponding flow rate, thereby realizing the EC closed-loop regulation of the mixed liquor.
[0018] By adopting the above technical solution, this application has the following beneficial effects: (1) The present invention dynamically adjusts the residual dead zone according to the residual high-frequency noise level, so that weak residuals are automatically filtered when the EC sensor noise is large, which can reduce the probability that random noise is compensated as a real process disturbance. (2) The present invention characterizes the persistence of residuals by residual persistence confidence and makes both the disturbance observation gain and the prediction time domain disturbance retention degree change online with confidence. It can improve the compensation speed when persistent disturbances occur and reduce the risk of overcompensation when low confidence residuals occur. (3) The present invention introduces the proportion of nominal fertilizer injection flow to total mixed flow into the perturbation observation update, and uses the physical sensitivity of the water-mother liquor mixing process to gate the observation gain so that the perturbation compensation intensity matches the current fertilizer injection ratio. (4) This invention combines adaptive total disturbance compensation with constrained model predictive control, which improves the accuracy of EC regulation while explicitly limiting the driving amount of the fertilizer injection pump and its single-cycle change, thus avoiding obtaining control commands that exceed the allowable range of the actuator. Attached Figure Description
[0019] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.
[0020] Figure 1 The flowchart of an EC adaptive disturbance compensation predictive control method for an integrated water and fertilizer machine provided in an embodiment of the present invention is shown. Figure 2 A schematic diagram of an EC adaptive disturbance compensation predictive control system for an integrated water and fertilizer machine provided in an embodiment of the present invention is shown. Figure 3 A schematic diagram of the prediction compensation module and the model prediction optimization module provided in an embodiment of the present invention is shown; Figure 4 A schematic diagram of the clean water-mother liquor mixing EC object and feedback relationship provided in an embodiment of the present invention is shown; Figure 5 A structural block diagram of the CSAD-MPC controller provided in an embodiment of the present invention is shown; Figure 6 The EC response comparison diagram of MPC, DOB-MPC and CSAD-MPC under the mother liquor concentration disturbance condition in the embodiments of the present invention is shown. Figure 7 A comparison chart of the EC responses of MPC, DOB-MPC, and CSAD-MPC under the integrated disturbance condition in an embodiment of the present invention is shown. Detailed Implementation
[0021] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are only used to more clearly illustrate the technical solution of the present invention and are therefore merely examples, and should not be construed as limiting the scope of protection of the present invention.
[0022] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application should have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.
[0023] In one embodiment, such as Figure 1 As shown, an EC adaptive disturbance compensation predictive control method for an integrated water and fertilizer machine is provided, including: S1. Obtain the target EC setpoint, the measured EC value of the mixture, the clear water flow rate, and the fertilizer pump control volume of the previous sampling cycle.
[0024] S2. Update the nominal fertilizer injection flow rate and nominal EC value based on the target EC setpoint, clean water flow rate, and the execution drive quantity of the previous sampling period.
[0025] The nominal EC value is updated using a first-order hybrid dynamic update, specifically: , The sampling period is The nominal mixing time constant on the control side. For nominal hybrid EC, For instantaneous mixing EC, For clean water flow, For Qingshui EC, To control the nominal fertilizer injection flow rate on the control side, The nominal mother liquor EC used in the controller; The instantaneous mixing EC is calculated according to the following formula: , The nominal fertilizer injection flow rate on the control side is updated dynamically using the first-order method of the fertilizer injection pump. , To control the nominal fertilizer injection pump time constant on the control side, To control the nominal flow gain of the fertilizer injection pump on the control side, This refers to the opening degree of the fertilizer injection pump or its equivalent driving force.
[0026] S3. Construct the output residual based on the measured EC value and the nominal EC value of the mixture, estimate the residual noise level online based on the output residual increment, generate an adaptive dead zone based on the residual noise level, and obtain the effective residual. S3 includes: The output residual is constructed based on the measured EC value and the nominal EC value: , The measured EC value of the mixture; Furthermore, to avoid compensating for random noise from the EC sensor as a real disturbance, the high-frequency noise level of the residual is estimated using the residual increment between adjacent sampling periods: , in, These are the weighting coefficients for the residual noise exponent. An adaptive dead zone is dynamically generated based on the residual noise estimation. , To achieve the minimum adaptive dead zone, This is the noise amplification factor; Then when Less than At that time, the effective residual Set to 0; when Not less than At that time, let the effective residual = This allows the residual update threshold to automatically change with the EC measurement noise level.
[0027] S4. Determine the reliability of residual persistence based on the co-directional persistence of the current effective residual and the effective residual of the previous sampling period, and determine the residual amplitude factor based on the effective residual amplitude.
[0028] The likelihood of the current residual originating from a persistent process disturbance or random noise is determined by whether the effective residuals are continuously in the same direction. A continuous state s(k) in the same direction is defined as: when the effective residuals... and When all values are non-zero and have the same sign, the continuation state s(k) is 1; otherwise, s(k) is 0. The reliability of residual persistence is updated according to the following formula: ; in, This is the reliability memory coefficient. Further, the residual magnitude factor is calculated based on the magnitude of the effective residual exceeding the adaptive dead zone: ; in, Sat(·) is the normalization scale for residual magnitudes, which means restricting the variable to a given upper and lower limit.
[0029] S5. Determine the mixing sensitivity based on the nominal fertilizer injection flow rate and the clean water flow rate, and generate the sensitivity gating coefficient based on the mixing sensitivity.
[0030] The sensitivity of fertilizer-side disturbances to the mixed EC is represented by the proportion of the current nominal fertilizer injection flow rate to the total mixed flow rate. , Then the sensitivity gating coefficient , For reference sensitivity, This is the lower limit of the sensitivity gate.
[0031] When the fertilizer injection ratio is low This decreases, thereby suppressing over-observation updates caused by small residuals in low-sensitivity regions; when the fertilizer injection ratio increases... Gradually approaching 1, so that the actual fertilizer-side disturbance can be fully compensated.
[0032] S6. Determine the adaptive perturbation observation gain based on the residual persistence confidence, residual magnitude factor, and sensitivity gating coefficient, and update the total perturbation estimate. S5 includes: The confidence-amplitude composite weight is calculated based on the residual persistence confidence and the residual amplitude factor: , Basic weights; The adaptive perturbation observation gain is calculated according to the following formula: , This is the lower limit of the gain limit. This is the upper limit of the gain limit; Update the total perturbation estimate using adaptive perturbation observation gain: , L(k) represents the upper limit of the total disturbance estimation amplitude. When the residual confidence is high, the residual amplitude is large, and the mixing process is in a highly sensitive region, L(k) increases accordingly, enabling the controller to quickly track continuous disturbances; when the residual mainly manifests as low-amplitude noise or the mixing sensitivity is low, L(k) decreases accordingly.
[0033] S7. Determine the prediction time-domain disturbance attenuation coefficient based on the residual persistence reliability, generate the total disturbance prediction sequence in the prediction time domain, and superimpose the total disturbance prediction sequence onto the nominal EC prediction sequence.
[0034] The prediction time-domain perturbation attenuation coefficient is adaptively determined based on the residual persistence reliability: , , These are the upper and lower limits for predicting the attenuation coefficient of the time-domain disturbance; In the i-th prediction step in the prediction time domain, the total perturbation prediction is propagated according to the following formula: , The total disturbance prediction is added to the nominal EC prediction: .
[0035] When the reliability of residual persistence is high Approaching its upper limit indicates that the current disturbance is more likely to persist in the future prediction time domain; when the residual confidence is low, This reduces the likelihood of random residuals or sensor noise propagating to subsequent prediction steps over a long period.
[0036] S8. Using the tracking error between the target EC and the compensated EC prediction value, and the increment of the driving quantity, as the objective function, solve for the optimal control sequence under constraints.
[0037] The objective function is: , in, For EC tracking error weights, To control the incremental weight, Representing the prediction time domain, the terminal tracking error term is based on... Sure, This refers to the terminal weighting coefficient; The constraints are , ; These are the upper and lower limits of the execution drive quantity, respectively; At each sampling time k, the optimal control increment sequence is obtained by minimizing the objective function J to solve for the control increment sequence in the predictive control time domain. ; Using the optimal control increment sequence The first optimal control increment and in accordance with Calculate the current driving quantity of the fertilizer injection pump.
[0038] S9. Adjust the opening degree or equivalent driving quantity of the fertilizer injection pump according to the first control quantity of the optimal control sequence, so that the mother liquor is injected into the clean water-mother liquor mixing unit at the corresponding flow rate, thereby realizing the EC closed-loop regulation of the mixed liquor.
[0039] The current fertilizer injection pump's driving quantity u(k) is output to the execution drive unit after the solution is completed at sampling time k, and is maintained within the current sampling period until the next sampling time. The remaining predictive control increments are not executed directly. At the next sampling time, the measured EC and clear water flow are reacquired, the total disturbance is updated, and the optimization problem is solved again, thus forming a rolling closed-loop control.
[0040] In one embodiment, such as Figure 2 As shown, an EC adaptive disturbance compensation predictive control system for an integrated water and fertilizer machine is provided, comprising: The information acquisition unit is used to acquire the target EC setpoint, the measured EC value of the mixed liquor, the clear water flow rate, and the fertilizer injection pump control volume of the previous sampling cycle. The central control unit includes a nominal model state update module, a residual noise estimation and adaptive dead zone module, a residual persistence reliability calculation module, a hybrid sensitivity gating module, a total disturbance observation module, a prediction compensation module, and a model prediction optimization module. The nominal model state update module is used to update the nominal fertilizer injection flow rate and nominal EC value based on the target EC setpoint, clean water flow rate, and the execution driving quantity of the previous sampling period; The residual noise estimation and adaptive dead zone module is used to construct the output residual based on the measured EC value and the nominal EC value of the mixture, estimate the residual noise level online based on the output residual increment, generate an adaptive dead zone based on the residual noise level, and obtain the effective residual. The residual persistence reliability calculation module is used to determine the residual persistence reliability based on the same-direction persistence of the current effective residual and the effective residual of the previous sampling period, and to determine the residual amplitude factor based on the effective residual amplitude. A mixing sensitivity gating module is used to determine the mixing sensitivity based on the nominal fertilizer injection flow rate and the clean water flow rate, and to generate a sensitivity gating coefficient based on the mixing sensitivity. The total disturbance observation module is used to determine the adaptive disturbance observation gain based on the residual persistence confidence, residual magnitude factor and sensitivity gating coefficient, and update the total disturbance estimate. The prediction compensation module is used to determine the prediction time-domain disturbance attenuation coefficient based on the residual persistence reliability, generate the total disturbance prediction sequence in the prediction time domain, and superimpose the total disturbance prediction sequence onto the nominal EC prediction sequence. The model prediction optimization module is used to construct the optimal control sequence under constraints by using the tracking error between the target EC and the compensated EC prediction value and the increment of the execution driving quantity as the objective function. The execution drive unit is used to adjust the opening degree or equivalent drive quantity of the fertilizer injection pump according to the first control quantity of the optimal control sequence, so that the mother liquor is injected into the clean water-mother liquor mixing unit at the corresponding flow rate, thereby realizing the EC closed-loop regulation of the mixed liquor.
[0041] Among them, the prediction compensation module and the model prediction optimization module are as follows: Figure 3 As shown.
[0042] The EC adaptive disturbance compensation predictive control system for fertigation machines provided in this application adopts the same inventive concept as the EC adaptive disturbance compensation predictive control method for fertigation machines described above, and can achieve the same beneficial effects, so it will not be described again here.
[0043] like Figure 4 As shown, the implementation of the above control system will be explained below.
[0044] The clear water branch provides the basic irrigation flow to the mixing unit, and the clear water flow measurement module outputs the flow rate in real time. In this embodiment, the nominal flow rate of clean water is 25 L / min, and the EC of clean water is 0.30 mS / cm.
[0045] The mother liquor branch provides an equivalent high-concentration fertilizer solution. The fertilizer injection pump receives the opening degree or equivalent driving quantity u(k) output by the central control unit and dynamically generates the actual fertilizer injection flow rate based on the pump's own dynamics. In this embodiment, the nominal EC of the equivalent mother liquor is 18.0 mS / cm.
[0046] After the clean water and mother liquor enter the mixing unit, an instantaneous mixture (EC) is formed, and after a certain mixing time constant, a true downstream mixture (EC) is gradually formed. The flow gain Kf of the fertilizer injection pump in the actual object is set to 0.060 L / (min·deg), and the time constant τ of the fertilizer injection pump is... p The time constant is set to 8.0 s, and the mixing time constant τm is set to 22.0 s.
[0047] The EC measurement module is located downstream of the mixing unit, measuring the EC of the mixed nutrient solution and feeding it back to the central control unit. To reflect actual sensor dynamics, the basic measurement delay in this embodiment is 4 s, and the basic measurement noise standard deviation is 0.003 mS / cm; in the sensor uncertainty and comprehensive disturbance test, the delay and noise are further increased and a bias is added.
[0048] The central control unit reads the target EC, clean water flow rate and measured EC in each sampling period, runs the nominal model state update, residual noise estimation, adaptive dead zone, residual persistence confidence calculation, mixed sensitivity gating, total disturbance observation, prediction compensation and model prediction optimization program, and outputs the fertilizer pump control quantity for the next sampling period.
[0049] To simulate pump parameter calibration errors and mixing dynamic estimation errors in the engineering process, the nominal fertilizer injection pump flow gain on the control side was set to 0.054 L / (min·deg), the fertilizer injection pump time constant was set to 9.5 s, and the mixing time constant was set to 27.0 s, which showed a clear mismatch with the actual parameters. The real-time mother liquor EC was not considered a known quantity for the controller; the controller used the nominal mother liquor EC for nominal prediction, while the actual mother liquor concentration change was uniformly estimated and compensated by the total disturbance observation module.
[0050] The parameters used in this embodiment are explained below.
[0051] The control period Ts is set to 1 s. During system initialization, the nominal fertilizer injection flow rate is set to 0, the nominal mixed EC state is set to 0.50 mS / cm, and the total disturbance estimate is set to... The initial values are set to 0, and the residual noise estimation, residual persistence confidence, and adaptive prediction decay coefficient are initialized.
[0052] In each sampling period, the model is first updated according to the nominal pump model and the nominal hybrid model. and , then calculate In this embodiment, the initial residual noise estimate is taken as 0.002 mS / cm, and the residual noise exponential weighting coefficient is... Set the value to 0.025, the minimum adaptive dead zone. Take 0.002 mS / cm, noise amplification factor Take 1.00.
[0053] Residual persistence reliability memory coefficient Take 0.89 as the normalized scale for residual amplitude. Take 0.065 mS / cm. When the current effective residual is in the same direction as the previous effective residual, the continuous state s(k) is set to 1; otherwise, it is set to 0, and c(k) is updated using exponential memory.
[0054] Mixed sensitivity reference value Set the sensitivity threshold to 0.090. Take 0.66. The nominal lower limit of the perturbation observation gain. Take 0.16, upper limit Set the value to 0.29, which is the lower limit of the gain limit. The confidence-amplitude composite weighting is set to 0.10, the basic weighting β is set to 0.49, and the total disturbance estimation amplitude is limited. Take 0.45 mS / cm.
[0055] The range of the predicted time-domain perturbation attenuation coefficient is set to =0.966、 =0.999. As c(k) increases, ρ(k) gradually approaches 0.999; when c(k) is low, ρ(k) approaches 0.966, thus changing the future disturbance retention strength according to the residual persistence.
[0056] To ensure a fair comparison with traditional DOB-MPC, CSAD-MPC uses the same MPC subject prediction time domain, control time domain, and objective function weights as DOB-MPC. CSAD-MPC, for example... Figure 5 As shown. In this embodiment, the prediction time domain... Choose 32 to control the time domain. Set the weight to 5 for EC tracking error. Take 360 and control the incremental weight. The value is set to 0.32; the fertilizer injection pump drive range is 0–90 degrees, and the change in a single sampling period does not exceed 5 degrees. The terminal weighting coefficients for CSAD-MPC and DOB-MPC are also considered. All values are 0.20.
[0057] To verify the effectiveness of the method of this invention, a digital closed-loop verification model for the EC mixing of clean water and mother liquor in an integrated water and fertilizer machine was established. The total verification time was 1200 s, the sampling period was 1 s, and a fixed random seed was used for random noise to ensure that each comparison algorithm ran under the same perturbation conditions. To reduce the impact of sensor delay and discrete state initialization on performance statistics, index statistics started from 30 s.
[0058] The comparative control methods include PID, MPC, PSO-MPC, DOB-MPC, and the CSAD-MPC of this invention. DOB-MPC uses a fixed disturbance observation gain of 0.18, a fixed residual dead zone of 0.002 mS / cm, and a fixed prediction time-domain disturbance attenuation coefficient of 0.985; CSAD-MPC uses the aforementioned adaptive parameters and employs the same parameters as DOB-MPC. , , and Both use the same terminal weighting coefficient of 0.20.
[0059] Six typical operating conditions are set: Condition 1 is target EC step tracking, with the target EC set to 1.20, 1.80, 2.20 and 1.50 mS / cm at 0 s, 300 s, 600 s and 900 s respectively; Condition 2 involves continuous switching of the target EC, with the target EC set to 1.50, 2.00, 2.50, 1.80, and 2.20 mS / cm at 0 s, 240 s, 480 s, 720 s, and 960 s, respectively.
[0060] Condition 3 involves a disturbance in the flow rate of clean water. The target EC is maintained at 2.00 mS / cm. The flow rate of clean water changes to 80%, 120%, 90%, and 110% of the nominal flow rate after 250–450 s, 450–650 s, 650–850 s, and 850 s, respectively, with an additional 2.5% periodic flow fluctuation.
[0061] Condition 4 involves a disturbance in the concentration of the mother liquor. The target EC is maintained at 2.00 mS / cm. After 250–500 s, 500–750 s, 750–950 s, and 950 s, the EC of the mother liquor becomes 80%, 120%, 90%, and 110% of the nominal EC of the mother liquor, respectively.
[0062] Condition 5 represents sensor uncertainty. The target EC is set to 1.60, 2.10, and 1.70 mS / cm at 0 s, 350 s, and 700 s, respectively. The sensor delay is increased to 10 s, the noise standard deviation is increased to 0.015 mS / cm, and a bias of 0.03 mS / cm is added.
[0063] Condition 6 is a comprehensive disturbance. The target EC is set to 1.60, 2.20, 1.80, and 2.40 mS / cm at 0 s, 300 s, 600 s, and 900 s, respectively. The clean water flow rate increases to 115% of the nominal value from 250 to 550 s and decreases to 85% of the nominal value from 700 to 1000 s, with an additional 4% periodic fluctuation. The mother liquor EC decreases to 85% of the nominal value from 500 to 850 s and increases to 112% of the nominal value after 850 s. At the same time, an 8-s measurement delay, a noise standard deviation of 0.012 mS / cm, and a measurement bias of -0.02 mS / cm are set.
[0064] The average performance indicators for the six operating conditions are shown in Table 1.
[0065] Table 1 Average performance indicators for six types of operating conditions
[0066] As shown in Table 1, the CSAD-MPC of this invention has an average RMSE of 0.0876 mS / cm, an average MAE of 0.0414 mS / cm, an average ITAE of 19.743, and an average time proportion within ±0.05 mS / cm of EC error of 81.20%. Compared with DOB-MPC, the average RMSE is reduced by 10.02%, the average MAE by 38.62%, the average ITAE by 47.61%, and the in-band time is improved by 22.46 percentage points. Compared with PSO-MPC, the average RMSE is reduced by 29.92%, the average ITAE by 61.81%, and the average ControlTV is lower than that of PSO-MPC.
[0067] The performance of DOB-MPC and CSAD-MPC under various typical operating conditions was further compared, and the results are shown in Table 2.
[0068] Table 2 Comparison of Typical Operating Condition Performance of DOB-MPC and CSAD-MPC
[0069] Under the clear water flow disturbance condition, the RMSE of DOB-MPC is 0.0587 mS / cm, the ITAE is 26.056, and the in-band time is 72.67%; while the RMSE of CSAD-MPC decreases to 0.0406 mS / cm, the ITAE decreases to 7.308, and the in-band time increases to 90.61%. At the same time, the ControlTV of CSAD-MPC decreases from 253.88 of DOB-MPC to 242.47, indicating that the performance improvement under this condition is not obtained by increasing the intensity of the control action.
[0070] Under conditions of mother liquor concentration disturbance, DOB-MPC had an RMSE of 0.1092 mS / cm, an ITAE of 51.251, and an in-band time of 26.30%; CSAD-MPC had an RMSE that decreased to 0.0906 mS / cm, an ITAE that decreased to 23.000, an in-band time that increased to 79.25%, and a steady-state absolute error that decreased from 0.0870 mS / cm to 0.0045 mS / cm. Figure 6 The EC time-series response of MPC, DOB-MPC and CSAD-MPC under this operating condition is given. It can be seen that after the mother liquor concentration changes, CSAD-MPC can eliminate the persistent deviation and return to the vicinity of the target EC more quickly.
[0071] In sensor delay noise conditions, the ITAE of CSAD-MPC decreased from 32.312 of DOB-MPC to 18.379, and the in-band time increased from 52.86% to 77.11%, indicating that residual noise estimation, adaptive dead zone and residual persistence reliability can maintain good long-term tracking performance even in the presence of strong measurement noise.
[0072] Under combined disturbance conditions, CSAD-MPC reduced RMSE by 2.73% and ITAE by 28.43% compared to DOB-MPC, while increasing in-band time from 46.46% to 67.89%. Figure 7 As shown, the main improvement under the combined disturbance is reflected in the long-term cumulative error and the steady-state tracking ability after the disturbance is recovered, rather than simply pursuing the minimum instantaneous peak value.
[0073] The above results show that, without changing the prediction time domain, control time domain, and main objective function weights of the DOB-MPC main body, this invention improves the adaptability of total disturbance estimation and prediction compensation to different operating conditions through residual noise adaptive dead zone, residual persistence reliability, hybrid sensitivity gating, and reliability-driven prediction time domain disturbance attenuation.
[0074] The main variables in this embodiment are described in Table 3.
[0075] Table 3 Explanation of Key Variables
[0076] Based on the same inventive concept, this invention can also be applied to nutrient solution mixing equipment with similar mixing and execution dynamics. When the actuator uses pump speed, pulse width modulation duty cycle, or equivalent fertilizer injection flow rate as the control quantity, it is only necessary to replace the mapping relationship between the control quantity and fertilizer injection flow rate in the nominal pump model with the calibration relationship of the corresponding actuator. The residual noise estimation, persistence reliability, mixing sensitivity gating, disturbance prediction compensation, and rolling optimization process remain unchanged.
[0077] Based on the same inventive concept, the water flow rate of this invention can be provided in real time by the flow measurement module or by the upper control system of the water and fertilizer integrated machine. When the real-time EC of the mother liquor can be reliably obtained, the real-time EC of the mother liquor can be used as the nominal mixing model with known disturbance input. When the real-time EC of the mother liquor is unavailable or the measurement reliability is insufficient, the nominal EC of the mother liquor is used, and the output deviation caused by the change in the concentration of the mother liquor is compensated by the total disturbance observation module.
[0078] The above embodiments are only used to provide a detailed description of the technical solutions of this application. However, the descriptions of the above embodiments are only for the purpose of helping to understand the methods of the embodiments of the present invention and should not be construed as limiting the embodiments of the present invention. Any variations or substitutions that can be easily conceived by those skilled in the art should be covered within the protection scope of the embodiments of the present invention.
Claims
1. A method for EC adaptive disturbance compensation predictive control of an integrated water and fertilizer machine, characterized in that, include: Obtain the target EC setpoint, the measured EC value of the mixture, the clear water flow rate, and the fertilizer pump control volume of the previous sampling cycle; The nominal fertilizer injection flow rate and nominal EC value are updated based on the target EC setpoint, clean water flow rate, and the execution drive quantity of the previous sampling period; The output residual is constructed based on the measured EC value and the nominal EC value of the mixture, and the residual noise level is estimated online based on the output residual increment. An adaptive dead zone is generated based on the residual noise level to obtain the effective residual. The reliability of residual persistence is determined based on the co-directional persistence of the current effective residual and the effective residual of the previous sampling period, and the residual amplitude factor is determined based on the effective residual amplitude. The mixing sensitivity is determined based on the nominal fertilizer injection flow rate and the clean water flow rate, and a sensitivity gating coefficient is generated based on the mixing sensitivity. The adaptive perturbation observation gain is determined based on the residual persistence confidence, residual magnitude factor, and sensitivity gating coefficient, and the total perturbation estimate is updated. The prediction time-domain disturbance attenuation coefficient is determined based on the residual persistence reliability, a total disturbance prediction sequence in the prediction time domain is generated, and the total disturbance prediction sequence is superimposed on the nominal EC prediction sequence. The objective function is constructed using the tracking error between the target EC and the compensated EC prediction value, as well as the increment of the driving force, and the optimal control sequence is solved under constraints. Adjust the opening degree or equivalent driving quantity of the fertilizer injection pump according to the first control quantity of the optimal control sequence, so that the mother liquor is injected into the clean water-mother liquor mixing unit at the corresponding flow rate, thereby realizing the EC closed-loop regulation of the mixed liquor.
2. The method according to claim 1, characterized in that, The nominal EC value is updated using a first-order hybrid dynamic update, specifically: , The sampling period is The nominal mixing time constant on the control side. For nominal hybrid EC, For instantaneous mixing EC, For clean water flow, For Qingshui EC, To control the nominal fertilizer injection flow rate on the control side, The nominal mother liquor EC used in the controller; The instantaneous mixing EC is calculated according to the following formula: , The nominal fertilizer injection flow rate on the control side is updated dynamically using the first-order method of the fertilizer injection pump. , To control the nominal fertilizer injection pump time constant on the control side, To control the nominal fertilizer injection pump flow gain on the control side, This refers to the opening degree of the fertilizer injection pump or its equivalent driving force.
3. The method according to claim 2, characterized in that, The steps for obtaining the effective residual include: Construct the output residual: , The measured EC value of the mixture is shown. Estimate residual noise level: , The weighting coefficient for the residual noise index; Generate adaptive dead zone: , To achieve the minimum adaptive dead zone, This is the noise amplification factor; Then when Less than At that time, the effective residual Set to 0; when Not less than At that time, let the effective residual = .
4. The method according to claim 3, characterized in that, The reliability of residual persistence is updated according to the following formula: ; in, The reliability memory coefficient; when the effective residual and When all values are non-zero and have the same sign, the state s(k) continues in the same direction, s(k) = 1; otherwise, s(k) = 0. The residual magnitude factor is: ; Sat(·) is the normalization scale for residual magnitudes, which means restricting the variable to a given upper and lower limit.
5. The method according to claim 4, characterized in that, The mixed sensitivity is: , Then the sensitivity gating coefficient , For reference sensitivity, This is the lower limit of the sensitivity gate.
6. The method according to claim 5, characterized in that, The steps for obtaining the total disturbance estimate include: The confidence-amplitude composite weight is calculated based on the residual persistence confidence and the residual amplitude factor: , Basic weights; Calculate the adaptive perturbation observation gain: , This is the lower limit of the gain limit. This is the upper limit of the perturbation observation gain; Update the total perturbation estimate using adaptive perturbation observation gain: , This represents the upper limit of the estimated amplitude of the total disturbance.
7. The method according to claim 6, characterized in that, The steps for obtaining the nominal EC prediction sequence include: The prediction time-domain perturbation attenuation coefficient is adaptively determined based on the residual persistence reliability: , , These are the upper and lower limits for predicting the attenuation coefficient of the time-domain disturbance; In the i-th prediction step in the prediction time domain, the total perturbation prediction is propagated according to the following formula: , The total disturbance prediction is added to the nominal EC prediction: 。 8. The method according to claim 7, characterized in that, The objective function is: , in, For EC tracking error weights, To control the incremental weight, Representing the prediction time domain, the terminal tracking error term is based on... Sure, This refers to the terminal weighting coefficient; The constraints are , ; At each sampling time k, the optimal control increment sequence is obtained by minimizing the objective function J to solve for the control increment sequence in the predictive control time domain. ; Find the optimal control increment sequence China's first control increment and in accordance with Calculate the current driving quantity of the fertilizer injection pump.
9. An EC adaptive disturbance compensation predictive control system for an integrated water and fertilizer machine, characterized in that, include: The information acquisition unit is used to acquire the target EC setpoint, the measured EC value of the mixed liquor, the clear water flow rate, and the fertilizer injection pump control volume of the previous sampling cycle. The central control unit includes a nominal model state update module, a residual noise estimation and adaptive dead zone module, a residual persistence reliability calculation module, a hybrid sensitivity gating module, a total disturbance observation module, a prediction compensation module, and a model prediction optimization module. The nominal model state update module is used to update the nominal fertilizer injection flow rate and nominal EC value based on the target EC setpoint, clean water flow rate, and the execution driving quantity of the previous sampling period; The residual noise estimation and adaptive dead zone module is used to construct the output residual based on the measured EC value and the nominal EC value of the mixture, estimate the residual noise level online based on the output residual increment, generate an adaptive dead zone based on the residual noise level, and obtain the effective residual. The residual persistence reliability calculation module is used to determine the residual persistence reliability based on the same-direction persistence of the current effective residual and the effective residual of the previous sampling period, and to determine the residual amplitude factor based on the effective residual amplitude. A mixing sensitivity gating module is used to determine the mixing sensitivity based on the nominal fertilizer injection flow rate and the clean water flow rate, and to generate a sensitivity gating coefficient based on the mixing sensitivity. The total disturbance observation module is used to determine the adaptive disturbance observation gain based on the residual persistence confidence, residual magnitude factor and sensitivity gating coefficient, and update the total disturbance estimate; The prediction compensation module is used to determine the prediction time-domain disturbance attenuation coefficient based on the residual persistence reliability, generate the total disturbance prediction sequence in the prediction time domain, and superimpose the total disturbance prediction sequence onto the nominal EC prediction sequence. The model prediction optimization module is used to construct the optimal control sequence under constraints by using the tracking error between the target EC and the compensated EC prediction value and the increment of the execution driving quantity as the objective function. The execution drive unit is used to adjust the opening degree or equivalent drive quantity of the fertilizer injection pump according to the first control quantity of the optimal control sequence, so that the mother liquor is injected into the clean water-mother liquor mixing unit at the corresponding flow rate, thereby realizing the EC closed-loop regulation of the mixed liquor.
