Space photovoltaic cable preparation process control method and space photovoltaic cable

Abstract

This invention discloses a process control method for the fabrication of space photovoltaic cables and a space photovoltaic cable itself, belonging to the field of space photovoltaic cable fabrication technology. The method includes: acquiring the original dielectric loss measurement value of the insulating material at a characteristic frequency during extrusion, obtaining a corrected value through temperature compensation and digital filtering; determining the degree of crosslinking at the current moment based on the corrected value and a dielectric loss-crosslinking kinetic model in the form of a set of differential equations; inputting the degree of crosslinking into a model predictive controller, using a quadratic programming algorithm to solve for a sequence of control quantities including extrusion temperature setpoints and traction speed setpoints, and adjusting the extrusion temperature and traction speed according to the first control quantity. This invention achieves online sensing and closed-loop control of the degree of crosslinking of the insulating material, improving the uniformity of the crosslinking degree of the insulation layer in space photovoltaic cables.

Description

Technical Field

[0001] This invention belongs to the field of space photovoltaic cable manufacturing technology, and relates to a method for controlling the manufacturing process of space photovoltaic cables. Background Technology

[0002] In the extrusion production of insulation layers for space photovoltaic cables, an open-loop control method is typically used, where process parameters such as extrusion temperature and traction speed are set. Alternatively, a closed-loop proportional-integral-derivative control method based on melt temperature and melt pressure feedback is employed. Operators set the temperature values ​​for each heating zone of the extruder and the linear speed values ​​for the traction machine based on experience. This allows the insulating material to melt and plasticize within the extruder barrel, then be formed through the die head, and the cross-linking reaction is completed relying on the material's own thermal history. To ensure that the degree of cross-linking meets the radiation resistance and mechanical performance requirements of the space environment, samples need to be periodically taken from the production line and sent to a laboratory for offline testing of the degree of cross-linking using solvent extraction or thermal analysis methods. The process parameters are then manually adjusted based on the test results.

[0003] The existing process control methods described above have the following drawbacks: First, the offline detection cycle for crosslinking degree is long (usually several hours to several days), and the detection results lag significantly behind the production process. This makes it impossible to detect fluctuations in crosslinking degree caused by batch differences in raw materials, screw wear, or changes in ambient temperature in a timely manner, resulting in a large number of defective products being produced continuously before detection. Second, dielectric loss, as a physical quantity that sensitively reflects the micro-crosslinking structure of the material, is significantly affected by melt temperature fluctuations. Existing control systems do not effectively compensate for and filter this measurement value, resulting in a low signal-to-noise ratio and poor stability of the measurement signal, making it difficult to use for closed-loop control. Third, the effects of extrusion temperature and traction speed on crosslinking degree and dielectric loss exhibit strong coupling, nonlinearity, and large time delay characteristics. Conventional proportional-integral-derivative controllers cannot effectively handle multivariable coupling constraints, easily leading to overshoot or oscillation, affecting the uniformity of the crosslinking degree of the insulation layer.

[0004] Therefore, a process control method for the fabrication of space photovoltaic cables is needed, which can sense the cross-linking state of the insulating material online and perform coordinated optimization control of extrusion temperature and traction speed. Summary of the Invention

[0005] To address the problems existing in the background technology, this invention proposes a method for controlling the manufacturing process of space photovoltaic cables.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: a method for controlling the manufacturing process of space photovoltaic cables, comprising the following steps: Obtain the original dielectric loss measurement of the insulating material at at least one characteristic frequency during the extrusion process; Temperature compensation and digital filtering are performed on the original dielectric loss measurement value to obtain the corrected dielectric loss measurement value. Based on the corrected dielectric loss measurement and the dielectric loss-crosslinking kinetic model in the form of a set of differential equations, the degree of crosslinking at the current moment is determined; The dielectric loss-crosslinking kinetic model includes a crosslinking degree dynamic equation and a dielectric loss output equation. The degree of crosslinking is input into a preset model predictive controller, and a quadratic programming algorithm is used to solve the control quantity sequence output by the model predictive controller. The control quantity sequence includes the extrusion temperature setpoint and the traction speed setpoint. The extrusion temperature and traction speed are adjusted according to the first control quantity in the control quantity sequence.

[0007] Specifically, the crosslinking degree dynamic equation determines the rate of change of the crosslinking degree based on the melt temperature; The dielectric loss output equation determines the predicted dielectric loss value at the corresponding characteristic frequency based on the degree of crosslinking and the rate of change of the degree of crosslinking.

[0008] Specifically, obtaining the original dielectric loss measurement value of the insulating material at at least one characteristic frequency during the extrusion process includes: The original dielectric loss measurements were obtained at three characteristic frequencies, namely a first characteristic frequency, a second characteristic frequency, and a third characteristic frequency. The first characteristic frequency corresponds to the polymer main chain motion, the second characteristic frequency corresponds to the side chain dipole orientation, and the third characteristic frequency corresponds to the micro-interface polarization.

[0009] Specifically, the model prediction controller adopts a cascaded architecture, including an upper-level planner and a lower-level controller; The upper-level planner generates the target dielectric loss trajectory by solving an optimization problem based on the target crosslinking degree and the dielectric loss-crosslinking dynamics model. The lower-level controller takes tracking the dielectric loss target trajectory as its main control objective and solves the control quantity sequence.

[0010] Specifically, the lower-level controller solves for the control quantity sequence, including: Obtain the measured value of the controlled variable at the current moment, wherein the measured value of the controlled variable includes the corrected dielectric loss measurement value and the melt pressure measurement value; Input the measured value of the controlled variable into the prediction model to obtain the predicted value of the controlled variable in the future prediction time domain; Construct an objective function, which includes an error term between the predicted dielectric loss value and the target dielectric loss trajectory, a penalty term for the change in the control quantity, and a penalty term for the absolute value of the control quantity; Under the premise of satisfying the preset constraints, a quadratic programming algorithm is used to solve for the sequence of control variables that minimizes the objective function.

[0011] Specifically, the preset constraints include hard constraints and soft constraints; The hard constraints include the upper and lower limits of the extrusion temperature, the upper and lower limits of the traction speed, and the upper and lower limits of the melt pressure; The soft constraint includes the allowable fluctuation range of dielectric loss, and the objective function also includes a penalty term for violating the soft constraint.

[0012] Specifically, a method for controlling the manufacturing process of space photovoltaic cables also includes the step of updating the prediction model parameters of the model prediction controller online: Obtain the measured value of the controlled variable at the current sampling time; Update the process gain matrix based on the measured values ​​of the controlled variable and the recursive least squares method; The updated process gain matrix is ​​validated, including single-step rate of change validation and physical boundary validation. When the verification passes, the prediction model is corrected based on the updated process gain matrix.

[0013] Specifically, a method for controlling the manufacturing process of space photovoltaic cables also includes steps of model confidence assessment and control mode switching: Calculate the model confidence index based on the error between historical measurements and historical predictions; When the model confidence index falls below the first confidence threshold, the update of the prediction model parameters is stopped. When the model confidence index is lower than the second confidence threshold and the duration exceeds the preset duration, the system switches to conservative control mode, in which the weight of the penalty term for the change in the control quantity in the objective function is increased.

[0014] Specifically, the step of performing temperature compensation and digital filtering on the original dielectric loss measurement value to obtain the corrected dielectric loss measurement value includes: Obtain the real-time temperature of the ring electrode; The original dielectric loss measurement value is corrected for temperature compensation based on the real-time temperature and the preset temperature compensation coefficient. The temperature-compensated dielectric loss measurement is subjected to median filtering and Kalman filtering to obtain the corrected dielectric loss measurement.

[0015] Specifically, the step of obtaining the original dielectric loss measurement value of the insulating material at at least one characteristic frequency during the extrusion process also includes a signal validity verification step: Determine whether the original dielectric loss measurement value is within the preset range; Determine whether the rate of change of the original dielectric loss measurement value exceeds a preset rate of change threshold; If any judgment result is negative, the current original dielectric loss measurement value is marked as invalid, and the valid dielectric loss measurement value of the previous moment is retained.

[0016] This invention provides a space photovoltaic cable, in which the extrusion temperature and traction speed are controlled and adjusted by the above-mentioned space photovoltaic cable manufacturing process control method during traction preparation.

[0017] Compared with the prior art, the present invention has the following beneficial effects: First, by obtaining the original dielectric loss measurement value of the insulating material at the characteristic frequency during the extrusion process and performing temperature compensation and digital filtering, the online and real-time sensing of the micro-crosslinking state of the material is realized. Compared with the existing offline detection method, the problem of the detection result being seriously lagging behind the production process is avoided.

[0018] Second, by using a dielectric loss-crosslinking kinetic model in the form of a set of differential equations to determine the degree of crosslinking at the current moment, and using this as the input to the model predictive controller, a direct closed loop from process parameter control to material performance control is realized, overcoming the defect that existing control methods cannot directly regulate the degree of crosslinking.

[0019] Third, by using a quadratic programming algorithm to solve the control quantity sequence containing the extrusion temperature setpoint and the traction speed setpoint, and adjusting the extrusion temperature and traction speed according to the first control quantity in the sequence, multivariate collaborative optimization control of a strongly coupled, nonlinear, and time-delay process is achieved. Compared with conventional proportional-integral-derivative controllers, this effectively suppresses overshoot and oscillation, and improves the uniformity of crosslinking degree of the insulation layer. Attached Figure Description

[0020] Figure 1 This is a flowchart of a process control method for the fabrication of space photovoltaic cables according to the present invention.

[0021] Figure 2 This is a process control block diagram of a method for controlling the manufacturing process of space photovoltaic cables according to the present invention.

[0022] Figure 3 This is a process control architecture diagram of a method for controlling the manufacturing process of space photovoltaic cables according to the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] like Figures 1-3 As shown, the technical solution adopted by this invention is as follows: A method for controlling the manufacturing process of space photovoltaic cables, comprising the following steps: S1: Obtain the original dielectric loss measurement of the insulating material at at least one characteristic frequency during the extrusion process.

[0025] On the extrusion production line for the insulation layer of space photovoltaic cables, the dielectric loss factor of the X-ETFE insulation material in its molten state at a specific frequency point is collected in real time using a ring electrode integrated into the extruder head and an online dielectric spectrum analyzer (such as Keysight E4990A). The original signal. This measurement is based on the principle of high-frequency dielectric spectroscopy analysis in materials science, which states that the dielectric loss factor of a material is deterministically correlated with its internal microstructure (such as crosslinking density, dipole orientation, and micro-defect distribution). The measurement system hardware includes a custom Inconel 625 ring electrode, RG-402 double-shielded coaxial cable, and a differential amplifier to ensure high signal-to-noise ratio original measurement signals in industrial environments with strong electromagnetic interference. This step is fundamental to achieving online sensing of the material's microscopic properties, providing direct observational data for subsequent model-based predictive control of the closed-loop process.

[0026] The method of obtaining the original dielectric loss measurement value of the insulating material at at least one characteristic frequency during the extrusion process specifically includes: obtaining the original dielectric loss measurement value at three characteristic frequencies, wherein the three characteristic frequencies are a first characteristic frequency, a second characteristic frequency, and a third characteristic frequency, the first characteristic frequency corresponds to the polymer main chain motion, the second characteristic frequency corresponds to the side chain dipole orientation, and the third characteristic frequency corresponds to the micro-interface polarization.

[0027] The acquisition of the original dielectric loss measurements needs to cover three characteristic frequency points with clear physical significance to comprehensively characterize the multi-scale microdynamic behavior of X-ETFE materials during the extrusion process. Specifically, the three acquired original dielectric loss measurements correspond to: The original dielectric loss measurement at the first characteristic frequency: This frequency point was selected as the characteristic frequency that can sensitively reflect the cooperative motion of the X-ETFE polymer backbone segments (i.e., glass transition-related relaxation).

[0028] The original dielectric loss measurement at the second characteristic frequency: This frequency point was selected as a characteristic frequency that can sensitively reflect the orientation polarization of polar groups on the side chains of the material under the action of an applied electric field.

[0029] The original dielectric loss measurement at the third characteristic frequency: This frequency point is selected as the characteristic frequency that can sensitively reflect the relaxation process of microscopic space charge (i.e., the Maxwell-Wagner-Sillars effect) formed inside the material due to fillers, additives or phase interfaces.

[0030] By simultaneously monitoring the dielectric loss dynamics at these three characteristic frequencies, a state vector can be constructed that can fully describe the evolution of the material's multi-level structure from molecular chain segments to micro-interfaces, providing multi-dimensional input for the dielectric loss-crosslinking kinetics model, thereby more accurately inverting the current degree of crosslinking.

[0031] Specifically, the step of obtaining the original dielectric loss measurement value of the insulating material at at least one characteristic frequency during the extrusion process also includes a signal validity verification step: determining whether the original dielectric loss measurement value is within a preset range.

[0032] To ensure the reliability of online measurement data, signal validity verification is performed immediately after acquiring the original dielectric loss measurement value. Specifically, determining whether the original dielectric loss measurement value is within the preset range refers to verifying the dielectric loss factor measured in real time. The numerical value is compared with the system's preset effective measurement range. This preset range is... The lower limit of 0.0001 corresponds to the typical dielectric loss value of fully cross-linked insulating materials, while the upper limit of 0.01 corresponds to an uncross-linked or slightly degraded state. If the measured value exceeds this range, the measurement result is considered invalid, possibly due to sensor malfunction, severe signal interference, or material abnormality. This step is a crucial prerequisite for ensuring the accuracy and reliability of the input data for subsequent control algorithms.

[0033] Determine whether the rate of change of the original dielectric loss measurement value exceeds a preset rate of change threshold.

[0034] As another key criterion for verifying signal validity, determining whether the rate of change of the original dielectric loss measurement exceeds a preset rate of change threshold involves calculating the absolute value of the difference between the dielectric loss measurement at the current sampling time and the previous sampling time, and comparing it with the preset threshold. This preset rate of change threshold is... Since the evolution of a material's microstructure is a relatively continuous physicochemical process, its dielectric loss value will not change abruptly under normal operating conditions. If the rate of change exceeds this threshold, the measurement result is deemed invalid due to the influence of a transient strong disturbance (such as power fluctuations or mechanical vibrations). This criterion effectively filters out non-physical noise spikes.

[0035] If any judgment result is negative, the current original dielectric loss measurement value is marked as invalid, and the valid dielectric loss measurement value of the previous moment is retained.

[0036] If either judgment result is negative (meaning either of the aforementioned two judgment conditions (range check and rate of change check) is not met), the control system marks the current raw dielectric loss measurement value as invalid data. Subsequently, the system retains the valid dielectric loss measurement value from the previous time step, using the verified valid measurement value from the previous sampling period as the replacement value for the current time step, for subsequent temperature compensation, digital filtering, and model predictive control calculations. This strategy avoids drastic disturbances to the entire control loop caused by a single invalid measurement, significantly improving the robustness and stability of the control system.

[0037] S2: Perform temperature compensation and digital filtering on the original dielectric loss measurement value to obtain the corrected dielectric loss measurement value.

[0038] To eliminate the physical interference of melt temperature fluctuations during extrusion on dielectric loss measurements and to suppress random noise in the industrial environment, a series of signal conditioning operations are performed on the raw dielectric loss measurements obtained from the toroidal electrode. This process is a crucial preliminary step to ensure the accuracy and reliability of the data used for subsequent crosslinking degree inversion and model predictive control.

[0039] First, temperature compensation correction based on a physical model is performed, then two-stage cascaded digital filtering and smoothing are performed, and finally the corrected dielectric loss measurement value is output for the control system.

[0040] The process of performing temperature compensation and digital filtering on the original dielectric loss measurement value to obtain the corrected dielectric loss measurement value specifically includes: acquiring the real-time temperature of the ring electrode.

[0041] As the first step in temperature compensation, obtaining the real-time temperature of the toroidal electrode involves using a PT100 platinum resistance temperature sensor integrated into a custom Inconel 625 toroidal electrode to collect the surface temperature of the electrode directly in contact with the molten insulating material. This temperature signal is transmitted to an industrial control computer via a separate signal channel, serving as a necessary input parameter for the temperature compensation algorithm. This real-time temperature accurately reflects the local thermodynamic state at the dielectric loss measurement point and is fundamental to performing effective temperature compensation.

[0042] The original dielectric loss measurement value is corrected for temperature compensation based on the real-time temperature and the preset temperature compensation coefficient.

[0043] After acquiring the real-time temperature, a temperature compensation correction is performed on the original dielectric loss measurement value based on the real-time temperature and a preset temperature compensation coefficient. This temperature compensation correction is implemented using the following formula: ; in, This is the measured dielectric loss value after temperature compensation correction. This is the original dielectric loss measurement value. This represents the real-time temperature of the ring electrode (in degrees Celsius). This is a reference temperature (taken as 25℃). This is a preset temperature compensation coefficient (valued at 0.002 / ℃). This value is based on the linear regression results of the dielectric loss factor of X-ETFE material with temperature variation within the temperature range of 25℃ to 100℃, and verified using a PT100 temperature sensor with a ring electrode. The physical basis of this formula is the characteristic that the dielectric loss factor typically increases approximately linearly with increasing temperature. Through this correction, the dielectric loss values ​​measured at different temperatures can be normalized to the same reference temperature, thereby separating the true value of dielectric loss caused by changes in the material's microstructure.

[0044] The temperature-compensated dielectric loss measurement is subjected to median filtering and Kalman filtering to obtain the corrected dielectric loss measurement.

[0045] After temperature compensation is completed, the obtained signal is subjected to median filtering and Kalman filtering on the temperature-compensated dielectric loss measurement value to obtain the corrected dielectric loss measurement value.

[0046] The digital filtering process employs a two-stage cascaded structure: First, an adaptive median filter with a window length of 5 is applied to effectively remove non-Gaussian spike noise caused by electromagnetic pulses or mechanical vibrations; then, the output of the median filter is used as the observation value and input to a Kalman filter for smoothing, where the observation noise covariance... The value is set to 0.0001. The Kalman filter utilizes the system's dynamic model and noise statistics to optimally estimate the signal, further suppressing Gaussian white noise while preserving the signal's dynamic characteristics. After these two stages of filtering, the output signal is the corrected dielectric loss measurement value used in the model predictive controller, and its signal-to-noise ratio and stability meet the requirements of high-precision closed-loop control.

[0047] S3: Determine the degree of crosslinking at the current moment based on the corrected dielectric loss measurement and the dielectric loss-crosslinking kinetic model in the form of a set of differential equations.

[0048] The dielectric loss-crosslinking kinetic model includes a crosslinking degree dynamic equation and a dielectric loss output equation.

[0049] An online state estimation algorithm based on physical mechanisms is employed. The corrected dielectric loss measurement, after temperature compensation and digital filtering, is used as the observation input and substituted into a pre-established dielectric loss-crosslinking kinetic model. By solving the inverse problem of this model or executing a state observer (such as an extended Kalman filter), the degree of crosslinking of the X-ETFE insulating material at the current moment during the extrusion process is retrieved in real time. This degree of crosslinking is a key process state variable characterizing the extent of the formation of the material's three-dimensional network structure; its accurate acquisition is a prerequisite for achieving high-precision closed-loop control. The dielectric loss-crosslinking kinetic model consists of two coupled differential equations: a dynamic equation for the degree of crosslinking and a dielectric loss output equation.

[0050] The dielectric loss-crosslinking kinetic model is written in continuous-time state-space form. A state vector is selected. ,in .

[0051] Equations of state: ; Observation equations (for three characteristic frequencies): ; in For process noise (covariance matrix) ), To measure noise (covariance matrix) ).

[0052] Discretization and EKF recursive steps (sampling period) Second): Prediction: The continuous state equations are discretized using the fourth-order Runge-Kutta method to obtain the state prediction. Sum of error covariance prediction ,in Let be the state transition Jacobian matrix.

[0053] Calculate the Kalman gain: ,in To observe the Jacobian matrix ( (Value at the predicted point).

[0054] Status correction: ,in It is a vector consisting of the corrected dielectric loss measurements at three characteristic frequencies.

[0055] Output: Current degree of cross-linking (The first component of the state vector). Initial state Set as Initial covariance Set as identity matrix Multiples. Process noise covariance and measurement noise covariance Obtained through offline experimental data calibration.

[0056] The crosslinking degree dynamic equation describes the intrinsic dynamic law of crosslinking degree evolution over time, driven by melt temperature; the dielectric loss output equation establishes a quantitative mapping relationship between the current microstate of the material (characterized by crosslinking degree and its rate of change) and the observable dielectric loss. Together, these two equations constitute a complete and identifiable state-space model, enabling the dielectric loss signal measured externally to uniquely infer the internal, indirectly measurable crosslinking degree state.

[0057] Specifically, the crosslinking degree dynamic equation determines the rate of change of the crosslinking degree based on the melt temperature.

[0058] The chemical kinetics of the crosslinking reaction are governed by Arrhenius's law. The specific form of this dynamic equation is: ; in, The degree of crosslinking (dimensionless). Time (in seconds) This refers to the melt temperature (in degrees Celsius). Pre-exponential factor (with values ​​of...) ), The activation energy is 95 kJ / mol. The universal gas constant ( ), The reaction order is 1.5. This equation shows the rate of change of the degree of crosslinking. Melt temperature and current degree of crosslinking The reaction rate is a function of the temperature. Higher temperatures result in a faster reaction rate; as the degree of crosslinking approaches 1, the reactant concentration decreases, and the rate gradually slows down. This equation provides the time derivatives of the state variables for subsequent dielectric loss prediction.

[0059] The dielectric loss output equation determines the predicted dielectric loss value at the corresponding characteristic frequency based on the degree of crosslinking and the rate of change of the degree of crosslinking.

[0060] The model's output is determined by the dielectric loss output equation, which uses the degree of crosslinking and its rate of change to calculate the predicted dielectric loss at the corresponding characteristic frequency. This output equation models independently for each characteristic frequency, and its general form is: ; in, For the first Predicted dielectric loss values ​​at each characteristic frequency This is a specific nonlinear mapping function at this frequency. For the first characteristic frequency (corresponding to the polymer main chain motion), this function is specified as: ; in, and The model parameters are 0.008 and 0.05s, respectively. The physical meaning of this equation is that uncrosslinked molecular chain segments contribute the majority of the dielectric loss (and...). (Proportional to), while the rearrangement of dipoles during the crosslinking reaction also generates additional losses (compared to) (Proportional). Through this equation, the model can predict the theoretical dielectric loss value based on the currently estimated crosslinking state. This predicted value is compared with the corrected dielectric loss measurement, and the residual is used to drive the state estimation algorithm (such as an extended Kalman filter) to continuously correct the crosslinking. The estimation continues until it converges to the true value.

[0061] S4: Input the degree of crosslinking into a preset model prediction controller, and use a quadratic programming algorithm to solve the control quantity sequence output by the model prediction controller. The control quantity sequence includes the extrusion temperature setpoint and the traction speed setpoint.

[0062] A quadratic programming algorithm is used to solve the control sequence output by the model predictive controller. This control sequence includes the extrusion temperature setpoint and the traction speed setpoint. The degree of crosslinking at the current moment, obtained from the dielectric loss-crosslinking kinetics model inversion, is used as a key state feedback variable and input into a pre-configured model predictive controller (MPC). Based on an internal process model, this MPC constructs and solves an optimal control problem in a finite-time domain within each control cycle.

[0063] The optimization problem is constructed as a standard quadratic programming (QP) problem, where the decision variables are the sequence of control variables in the future control time domain. The specific composition of this control variable sequence is clearly defined as the control input direction... Incremental form It includes temperature adjustment for five extruder heating zones. and traction speed adjustment amount The final extrusion temperature and traction speed settings are obtained by adding the current settings to these adjustments.

[0064] The model predictive controller adopts a cascaded architecture, including an upper-level planner and a lower-level controller. The upper-level planner is a dielectric loss setpoint planner, responsible for handling long-term trajectory planning tasks directly related to the final product quality target; the lower-level controller is a multivariable MPC controller, responsible for executing real-time dynamic tracking control tasks. This cascaded architecture improves the system's control capability and robustness against complex multivariable coupled processes through task decoupling.

[0065] Application of real-time crosslinking degree in cascaded MPC: The current crosslinking degree is obtained from the dielectric loss-crosslinking kinetics model in real time. It is simultaneously fed to the internal state observers of both the upper-level planner and the lower-level controller.

[0066] Upper-level planner: As the initial state for trajectory planning, the degree of interconnection with the user-defined final goal. Together, they constitute a two-point boundary value problem, thereby generating a dielectric loss target trajectory that smoothly transitions from the current degree of crosslinking to the target degree of crosslinking. Specifically, the planner solves the following optimization problem with initial conditions:

[0067] ; in Estimate the degree of cross-linking at the current moment as the initial condition for the trajectory; The smoothing factor (ranging from 0.01 to 0.1) is used to constrain the rate of change of dielectric loss over time, preventing the generation of overly aggressive or non-smooth target trajectories, thereby ensuring the trackability of the lower-level controller.

[0068] The state observer of the lower-level controller: In each control cycle, the extended Kalman filter inside the lower-level MPC will... Crosslinking degree based on thermal history prediction The results are compared, residuals are calculated, and the crosslinking degree state variables in the state vector are corrected. The corrected states are used for forward calculation of the prediction model, thereby improving the accuracy of the dielectric loss prediction. This design allows the MPC control action to indirectly consider the deviation between the current crosslinking degree and the target crosslinking degree, realizing a closed loop from process parameter control to material property control.

[0069] The upper-level planner generates the target dielectric loss trajectory by solving the above optimization problem based on the target crosslinking degree and the dielectric loss-crosslinking dynamics model.

[0070] The lower-level controller takes tracking the dielectric loss target trajectory as its main control objective and solves the control quantity sequence.

[0071] The lower-level controller uses a discrete-time state-space model as its prediction model: ; ; in, For discrete sampling times, the sampling period is... Seconds; State Vector It is 8×1 dimensional, containing three dielectric loss states, one melt pressure state, and four thermal dynamic auxiliary states; the control input vector It is 6×1 dimension and is defined as ; Controlled variable output vector It is 4×1 dimension and is defined as ;matrix , , All were obtained through system identification.

[0072] At each sampling time The lower-level controller solves a problem to control the increment. Quadratic programming optimization problem with decision variables Its goal is to make the predicted output Closely follow the dielectric loss target trajectory provided by the upper-level planner, while satisfying process constraints, and finally output the optimal control quantity sequence.

[0073] The lower-level controller solves the control quantity sequence by: obtaining the measured value of the controlled variable at the current moment, wherein the measured value of the controlled variable includes the corrected dielectric loss measurement value and the melt pressure measurement value.

[0074] Obtaining the measured value of the controlled variable at the current moment is the starting point for optimization calculations. The measured value of the controlled variable here includes two types of key process data: first, the corrected dielectric loss measurement obtained through the aforementioned signal conditioning process, which directly reflects the microscopic cross-linking state of the material; and second, the melt pressure measurement obtained through a pressure sensor installed at the extruder head, which characterizes the flow stability of the extrusion process. These two types of measurements together constitute a complete observation of the current system state.

[0075] The measured value of the controlled variable is input into the prediction model to obtain the predicted value of the controlled variable in the future prediction time domain.

[0076] The prediction model here is a dynamic mathematical model that describes how control variables such as extrusion temperature and traction speed affect the corrected dielectric loss and melt pressure measurements. This model is typically obtained based on system identification or mechanistic modeling. The controller uses this model, combined with the current state and candidate control variable sequences, to continuously predict the future evolution trajectory of all controlled variables in the future prediction time domain (P = 60 sampling periods, corresponding to 30 seconds).

[0077] Construct an objective function, which includes an error term between the predicted dielectric loss value and the target dielectric loss trajectory, a penalty term for the change in the control quantity, and a penalty term for the absolute value of the control quantity.

[0078] To quantify control performance, constructing an objective function is crucial to the optimization problem. This objective function... Designed as: ; in, For the first Predicted dielectric loss value of step This corresponds to the target trajectory value of dielectric loss. This is the error weight matrix; To control the change in quantity, Its penalty weight matrix; It is the absolute value of the control quantity (i.e., the vector composed of the extrusion temperature setpoint and the traction speed setpoint). Its penalty weight matrix; To predict the length of the time domain, To control the time domain length, the objective function comprehensively considers factors such as tracking accuracy, smoothness of control actions, and energy consumption / actuator saturation.

[0079] Under the premise of satisfying the preset constraints, a quadratic programming algorithm is used to solve for the sequence of control variables that minimizes the objective function.

[0080] Since the objective function is a quadratic form with respect to the control variable and the constraints are linear, this optimization problem is classified as a standard quadratic programming problem. The controller calls a QP solver (such as OSQP or qpOASES) to compute the expression that makes the objective function... Minimum optimal control sequence According to the standard implementation of model predictive control, only the first element in the sequence is used. The actual control commands are applied to the extruder and traction machine.

[0081] The preset constraints include hard constraints and soft constraints.

[0082] The controller needs to handle two different types of constraints during the solution process. Hard constraints are physical or safety boundaries that must be strictly satisfied, and violations will lead to process failures or equipment damage; soft constraints are performance indicators that should be satisfied as much as possible but can be violated to a certain extent when necessary, and the degree of violation is reflected by adding a penalty term to the objective function.

[0083] The hard constraints include upper and lower limits for extrusion temperature, upper and lower limits for traction speed, and upper and lower limits for melt pressure.

[0084] For hard constraints, the specific numerical range is: extrusion temperature. satisfy ; traction speed satisfy Melt pressure satisfy The extrusion temperature range is set based on the typical process window (350℃±20℃) for X-ETFE material in the manufacture of space photovoltaic cables. 330℃ is the lower limit for the effective initiation of the crosslinking reaction, and 370℃ is the upper limit for preventing thermal degradation of the material. These constraints are directly embedded in the inequality constraint set of the QP problem, and the solver guarantees that no feasible solution will violate these boundaries.

[0085] The soft constraint includes the allowable fluctuation range of dielectric loss, and the objective function also includes a penalty term for violating the soft constraint.

[0086] For soft constraints, the allowable fluctuation range is defined as the dielectric loss target trajectory. The upper and lower limits are both 5%. To handle this soft constraint, the objective function... An additional penalty item has been added: ; in, This is the soft constraint penalty weight (with a value of 1000). When the predicted dielectric loss exceeds the allowable fluctuation range, this penalty term is activated, prompting the optimizer to prioritize pulling the trajectory back to the allowable range where feasible, but allowing temporary violations in extreme cases such as hard constraint conflicts, thereby improving the feasibility of the controller.

[0087] S5: Adjust the extrusion temperature and traction speed according to the first control quantity in the control quantity sequence.

[0088] After the model predictive controller completes the optimization calculation and outputs a sequence of control variables containing multiple future time-stamped values, only the first element of this sequence is extracted and executed as the actual control command. The first control variable is a two-dimensional vector, whose components correspond to the extrusion temperature setpoint and the traction speed setpoint, respectively. In practice, the extrusion temperature setpoint is sent to the distributed temperature control system of the extruder, and the actual melt temperature is brought close to the setpoint by adjusting the power output or cooling water flow of each heating zone; the traction speed setpoint is sent to the variable frequency drive of the traction machine, and the traction line speed of the cable is precisely tracked by adjusting the motor speed. This rolling time-domain, first-term-only strategy is a standard implementation method of model predictive control. It transforms the open-loop optimization result into closed-loop feedback control, effectively overcoming the influence of model mismatch and external disturbances, ensuring high-precision tracking of the dielectric loss target trajectory, and ultimately guaranteeing the uniformity and consistency of the crosslinking degree of the insulation layer.

[0089] A method for controlling the manufacturing process of space photovoltaic cables further includes the step of updating the prediction model parameters of the model prediction controller online. This sentence defines an additional step in a process control method for manufacturing space photovoltaic cables: online updating of the predictive model parameters of a model predictive controller (MPC). A model predictive controller (MPC) is defined as a model-based predictive control algorithm originating from the field of industrial process control, whose predictive model adopts a discrete-time state-space model form. , The state matrix Input matrix and output matrix Obtained through system identification.

[0090] The purpose of online updating of predictive model parameters is to enable the MPC controller to adapt to disturbances such as batch-to-batch material differences, screw wear, and changes in ambient temperature and humidity, while maintaining control accuracy. Specifically, this online update step uses the recursive least squares (RLS) method to identify the process gain matrix of the predictive model in real time. It also implements parameter verification and protection mechanisms to ultimately correct the prediction model.

[0091] Obtain the measured value of the controlled variable at the current sampling time.

[0092] The measured value of the controlled variable refers to the value at the current sampling time. The physical quantity values ​​read directly from sensors at the production site. Controlled variable output vector. for The dimensional vector specifically includes dielectric loss measurements at three characteristic frequencies. , , and melt pressure measurement values .

[0093] Sampling period The interval is set to 0.5 seconds, meaning data acquisition is performed every 0.5 seconds. This measurement will be used as input to the Recursive Least Squares (RLS) algorithm to update the process gain matrix. After acquiring the measurement, the actual temperature of each heating zone is read. However, this step only involves obtaining the measured values ​​of the controlled variable.

[0094] The process gain matrix is ​​updated based on the measured values ​​of the controlled variables and the recursive least squares method.

[0095] The update formula for Recursive Least Squares (RLS) is as follows: ; in, The current sampling time Estimated process gain matrix This is the estimated value from the previous time step. To update the gain vector, This is the vector of measured values ​​of the controlled variable at the current moment (containing three dielectric loss measurements and melt pressure measurements). This is the regression vector. Update the gain. The calculation formula is: ; covariance matrix ( The update formula for ) is: ; Among them, the forgetting factor The value is set to 0.98. This step enables the process gain of the predictive model to track the dynamic changes of the actual system, achieving adaptive adjustment of the model parameters. This update operation is performed every 10 sampling periods (i.e., every 5 seconds).

[0096] The updated process gain matrix is ​​validated, including single-step rate of change validation and physical boundary validation.

[0097] The verification includes the following two items: Single-step rate of change verification: Calculate each element in the process gain matrix ( The single-step relative rate of change of ) must satisfy: .

[0098] Here, 0.2 represents the maximum permissible rate of change. If the rate of change of an element exceeds this threshold, the update of that element is rejected, and the value from the previous time step is retained.

[0099] Physical boundary verification: Check each element in the process gain matrix ( Whether it is within the preset physical boundary range, that is, whether it satisfies: .

[0100] If an element exceeds the boundary, it is shrunk back to the nearest boundary value. The specific values ​​of the physical boundary are pre-calibrated based on the physical properties of the X-ETFE material. This verification step serves to prevent model parameter drift caused by noise or anomalous data, ensuring the stability and reliability of the prediction model.

[0101] When the verification passes, the prediction model is corrected based on the updated process gain matrix.

[0102] This step describes how, after successful verification, the updated process gain matrix is ​​used. This involves refining the prediction model of the Model Predictive Controller (MPC). The prediction model is in discrete-time state-space form. , The state matrix Input matrix and output matrix Obtained through system identification. Process gain matrix. Reflects control input Changes in the output of the controlled variable The steady-state gain relationship.

[0103] Process gain matrix Defined as the steady-state gain matrix, i.e., the system satisfies the following in steady state: .

[0104] In state-space model , In this context, the theoretical expression for the steady-state gain matrix is: .

[0105] When RLS updates get new Then, the following method was used for correction. (Keep and constant):

[0106] Calculate the theoretical steady-state gain of the current model. .

[0107] Calculate the gain bias matrix .

[0108] Update the input matrix using the pseudo-inverse method: ;in for Moore-Penrose pseudo-inverse (if If the term of office is completed, then ).

[0109] Stability verification: calculation The eigenvalues, if updated This leads to system instability (i.e.) If the eigenvalue modulus is greater than 1, then damped update is used: And re-verify stability.

[0110] If the system remains unstable, reject this update and maintain the existing system. .

[0111] When the verification passes, The values ​​are assigned to the input matrix of the prediction model to complete the correction. If the verification fails (single-step rate of change exceeds the limit or exceeds the physical boundary), the correction is not performed, and the original model parameters are maintained.

[0112] A method for controlling the manufacturing process of space photovoltaic cables also includes steps of model confidence assessment and control mode switching. Model confidence assessment refers to the process of quantitatively evaluating the current accuracy of the predictive model of the Model Predictive Controller (MPC). Control mode switching refers to the operation of switching between normal adaptive control mode and conservative control mode based on the model confidence assessment results.

[0113] Model confidence assessment is based on the error between historical measurements and historical predictions, which is used to calculate the model confidence index. The control mode switching is performed based on the relationship between this indicator and a preset threshold: when the model confidence index... When the value falls below the second confidence threshold (0.7) and the duration exceeds the preset duration (30 sampling periods, i.e., 15 seconds), the system automatically switches to conservative control mode, stops adaptive control, and increases the control increment weight.

[0114] When the model confidence index After restoring the value to above 0.8 and maintaining it for 10 sampling periods, switch back to adaptive mode. This step ensures the stability and reliability of the prediction model under abnormal disturbances or model mismatch, preventing control performance degradation due to model errors.

[0115] The model confidence index is calculated based on the error between historical measurements and historical predictions.

[0116] Model confidence index ( The formula for calculating ) is: ; in, The current discrete sampling time, sampling period It takes 0.5 seconds; This is the length of the historical data window, with a value of 50 (corresponding to a 25-second historical time window). For the first The measured values ​​of the controlled variable at the sampling time (including dielectric loss measurements at three characteristic frequencies) , , and melt pressure measurement value ; For the first Information about the time and preceding moments for the first The predicted value of the controlled variable at any given time; To measure the noise variance; the observation noise covariance of the Kalman filter ( (This can be used as a reference, but the specific value needs to be determined based on the system calibration.)

[0117] This formula maps the normalized sum of squared prediction errors to the (0,1] interval using an exponential function, and the model confidence index... The closer the value is to 1, the more accurate the prediction model; the closer it is to 0, the more severe the model mismatch. This calculation step serves to provide a quantitative basis for subsequent control mode switching.

[0118] When the model confidence index falls below the first confidence threshold, the prediction model parameters are no longer updated, and the system is based on the model confidence index. The first level of judgment and corresponding operation.

[0119] First confidence threshold When the model confidence index At this time, the operation to stop updating the prediction model parameters is executed, without the need for an additional duration condition. This operation only pauses the online identification of the process gain matrix using the recursive least squares (RLS) method, maintaining the current prediction model parameters. Unchanged. The control mode remains normal, and the control increment weights of the Model Predictive Controller (MPC) remain unchanged. This threshold is set based on the following: when the model prediction error variance exceeds 1.2 times the measurement noise variance (corresponding to...). Adaptive updates may introduce noise; a value of 0.85 provides an engineering safety margin.

[0120] When the model confidence index is lower than the second confidence threshold and the duration exceeds the preset duration, the system switches to conservative control mode, in which the weight of the penalty term for the change in the control quantity in the objective function is increased.

[0121] Second confidence threshold When the model confidence index Furthermore, if this state continues for more than 30 sampling periods (15 seconds), the system, having already stopped updating, further switches to a conservative control mode: the control increment weight matrix in the MPC objective function is adjusted. Each element is increased to twice its original size, that is The threshold and duration are set based on the fact that the response time constant of the X-ETFE extrusion process is about 10 seconds, and a low confidence level lasting 15 seconds can eliminate random noise interference.

[0122] Recovery condition: When the model confidence index After 10 sampling cycles, the system switches back to adaptive mode, resumes normal control increment weights, and re-enables RLS parameter updates.

[0123] In conservative control mode, two operations are performed: first, adaptive updates are stopped, i.e., the prediction model parameters are stopped, which is the same as the operation when the value is below the first confidence threshold; second, the weight of the penalty term for the change in control quantity in the objective function is increased.

[0124] Objective function of Model Predictive Controller (MPC) Includes control increment weight matrix Its definition is This corresponds to the temperature control increments and traction speed control increments for the five heating zones. In conservative control mode, the weight of the control increment is increased to twice, i.e. Increasing the weight of the penalty term for changes in the control quantity helps to suppress drastic changes in the control output, making the system response smoother and preventing excessive control actions that could lead to process runaway when the model is severely mismatched.

[0125] Recovery condition: When the model confidence index After 10 sampling cycles, the system switches back to adaptive mode, resumes normal control increment weights, and re-enables RLS parameter updates.

[0126] This embodiment provides a specific implementation of a process control method for the fabrication of space photovoltaic cables. This embodiment is applied to an extrusion production line for producing X-ETFE-insulated space photovoltaic cables for specific types of satellite solar arrays.

[0127] Step S1: Obtain the original dielectric loss measurement value.

[0128] The raw dielectric loss of the insulating material during extrusion is measured at three characteristic frequencies using a ring electrode integrated into the extruder head and an online dielectric spectrum analyzer. The ring electrode is made of custom Inconel 625 material, and the online dielectric spectrum analyzer is a Keysight E4990A model. The three characteristic frequencies are a first characteristic frequency, a second characteristic frequency, and a third characteristic frequency. The first characteristic frequency corresponds to the polymer backbone motion and has a value of 1 kHz. The second characteristic frequency corresponds to the side chain dipole orientation and has a value of 10 kHz. The third characteristic frequency corresponds to the micro-interface polarization and has a value of 100 kHz. The online dielectric spectrum analyzer measures and outputs the raw dielectric loss at the first characteristic frequency with a sampling period of 0.5 seconds. The original dielectric loss measurement value at the second characteristic frequency and the original dielectric loss measurement value at the third characteristic frequency .

[0129] Obtain the original dielectric loss measurement value , , Then, immediately perform a signal validity verification.

[0130] First, determine whether the original dielectric loss measurement value is within a preset range. The preset range is... The lower limit of this range, 0.0001, corresponds to the typical dielectric loss value of a fully cross-linked insulating material, while the upper limit, 0.01, corresponds to the uncross-linked state.

[0131] Second, determine whether the rate of change of the original dielectric loss measurement exceeds a preset rate of change threshold. Calculate the absolute value of the difference between the dielectric loss measurement at the current sampling time and the previous sampling time, and compare it with the preset rate of change threshold of 0.002 / second.

[0132] If the measured value at any characteristic frequency of the original dielectric loss measurement is not within the preset range, or its rate of change exceeds the preset rate of change threshold, the current original dielectric loss measurement is marked as invalid, and the valid dielectric loss measurement value from the previous moment is retained. When all judgment results are yes, the original dielectric loss measurement is marked as valid and sent to step S2.

[0133] Step S2: Temperature compensation and digital filtering.

[0134] The original dielectric loss measurement value , , Temperature compensation and digital filtering are performed to obtain the corrected dielectric loss measurement. , , .

[0135] First, the real-time temperature T of the ring electrode is obtained. The real-time temperature T is acquired using a PT100 platinum resistance temperature sensor integrated inside the custom Inconel625 ring electrode.

[0136] Secondly, the original dielectric loss measurement value is corrected by temperature compensation based on the real-time temperature T and the preset temperature compensation coefficient.

[0137] Temperature compensation correction uses the following formula: ;

[0138] The preset temperature compensation coefficient Values The reference temperature Values .

[0139] Finally, the dielectric loss measurement after temperature compensation correction was performed. Median filtering and Kalman filtering are performed to obtain the corrected dielectric loss measurement. The median filter is an adaptive median filter with a window length of 5. The observation noise covariance of the Kalman filter... Set to 0.0001.

[0140] Step S3: Determine the degree of crosslinking at the current moment.

[0141] Based on the corrected dielectric loss measurement value , , Using a dielectric loss-crosslinking kinetic model in the form of a system of differential equations, the degree of crosslinking at the current moment is determined. The dielectric loss-crosslinking kinetic model includes a crosslinking degree dynamic equation and a dielectric loss output equation.

[0142] The crosslinking degree dynamic equation determines the crosslinking degree based on the melt temperature T. The rate of change of is specifically expressed as: ;

[0143] Among them, the pre-finite factor Values ,activation energy( The value can be... Universal gas constant ( The value can be... Reaction order The value is 1.5.

[0144] The dielectric loss output equation is based on the degree of crosslinking. and the rate of change of the degree of crosslinking Determine the predicted dielectric loss value at the corresponding characteristic frequency. For the first characteristic frequency, the dielectric loss output equation is: ; Where model parameters The value is 0.008, model parameter. The value is 0.05 seconds.

[0145] The corrected dielectric loss measurement value is obtained by using an extended Kalman filter. , , Using the observed input, the inverse problem of the dielectric loss-crosslinking kinetics model is solved to retrieve the degree of crosslinking at the current moment in real time. The sampling period of the extended Kalman filter. It takes 0.5 seconds.

[0146] Step S4: Solve the control quantity sequence using model predictive control.

[0147] The degree of crosslinking A preset model predictive controller is input, and a quadratic programming algorithm is used to solve for the control quantity sequence output by the model predictive controller. The control quantity sequence includes the extrusion temperature setpoint and the traction speed setpoint.

[0148] The model predictive controller adopts a cascaded architecture, including an upper-level planner and a lower-level controller. The upper-level planner generates the target dielectric loss trajectory by solving an optimization problem based on the target crosslinking degree and the dielectric loss-crosslinking kinetics model. The target degree of crosslinking The value is set to 0.95. The optimization problem solved by the upper-level planner is:

[0149] .

[0150] Among them, smoothing factor ( The value is 0.05.

[0151] The lower-level controller tracks the dielectric loss target trajectory. The control quantity sequence is solved to achieve the primary control objective. The specific process of the lower-level controller solving the control quantity sequence is as follows.

[0152] First, obtain the measured value of the controlled variable at the current moment. The measured value of the controlled variable includes the corrected dielectric loss measured value. , , and melt pressure measurement value The melt pressure measurement value is obtained through a pressure sensor installed at the extruder head.

[0153] Second, the measured values ​​of the controlled variable are input into the prediction model to obtain the predicted values ​​of the controlled variable in the future prediction time domain. The prediction model is a discrete-time state-space model. ; Sampling period For 0.5 seconds. State vector. for Dimension, controlling the input vector for dimension, controlled variable output vector for Dimension. Matrix , , Obtained through system identification.

[0154] Third, construct the objective function. The objective function includes an error term between the predicted dielectric loss value and the target dielectric loss trajectory, a penalty term for changes in the control quantity, and a penalty term for the absolute value of the control quantity. ; The prediction time domain length The value is 60, which controls the length of the time domain. It is 30.

[0155] Fourth, under the premise of satisfying the preset constraints, a quadratic programming algorithm is used to solve for the objective function. The minimized sequence of control variables. The preset constraints include hard constraints and soft constraints.

[0156] The hard constraint includes: extrusion temperature satisfy traction speed satisfy Melt pressure satisfy .

[0157] The soft constraint includes the allowable fluctuation range of dielectric loss, defined as the target dielectric loss trajectory. The upper and lower limits are each 5%. To handle the soft constraints, a penalty term is added to the objective function J: ; Among them, the soft constraint penalty weight The value is 1000.

[0158] Step S5: Adjust the extrusion temperature and traction speed.

[0159] The extrusion temperature and traction speed are adjusted according to the first control variable in the control variable sequence. The first control variable is a two-dimensional vector, whose components correspond to the extrusion temperature setpoint and the traction speed setpoint, respectively. The extrusion temperature setpoint is sent to the distributed temperature control system of the extruder, and the traction speed setpoint is sent to the variable frequency drive of the traction machine.

[0160] This embodiment also includes the step of updating the prediction model parameters of the model prediction controller online.

[0161] First, obtain the measured value of the controlled variable at the current sampling time, i.e., the corrected dielectric loss measurement value. , , and the measured value of the melt pressure .

[0162] Second, update the process gain matrix based on the measured values ​​of the controlled variable and the recursive least squares method. The update formula for the recursive least squares method is: ; Update Gains: ; Covariance matrix update: ; Among them, forgetting factor The value is 0.98. This update operation is performed once every 10 sampling periods (i.e., every 5 seconds).

[0163] Third, the updated process gain matrix Verification is performed. The verification includes single-step rate of change verification and physical boundary verification. The single-step rate of change verification requires that the single-step relative rate of change of each element in the process gain matrix satisfy: .

[0164] The physical boundary verification requires that each element in the process gain matrix be within a preset physical boundary range.

[0165] Fourth, when the verification passes, the updated process gain matrix is ​​used. The prediction model is then corrected. The correction method involves calculating the theoretical steady-state gain of the current model. Calculate the gain bias matrix The input matrix is ​​updated using the pseudo-inverse method: .

[0166] in for The Moore-Penrose pseudo-inverse is used. After the update, a stability check is performed. If the system is unstable after the update, a damped update is used; if it is still unstable, the update is rejected.

[0167] This embodiment also includes steps for model confidence assessment and control mode switching.

[0168] First, based on the error between historical measurements and historical predictions, calculate the model confidence index (…). The calculation formula is: ; The historical data window length N is 50.

[0169] Second, when the model confidence index When the confidence level falls below the first confidence threshold of 0.85, the prediction model parameters are no longer updated.

[0170] Third, when the model confidence index If the value falls below the second confidence threshold of 0.70 and the duration exceeds the preset duration of 30 sampling periods (15 seconds), the system switches to conservative control mode. In conservative control mode, the weight of the penalty term for the change in control quantity in the objective function J is increased, and the control increment weight matrix is ​​adjusted. Each element is increased to twice its original size.

[0171] Fourth, when the model confidence index After recovering to above 0.80 and maintaining this level for 10 sampling periods, switch back to adaptive mode.

[0172] This invention provides a space photovoltaic cable, in which the extrusion temperature and traction speed are controlled and adjusted by the above-mentioned space photovoltaic cable manufacturing process control method during traction preparation.

[0173] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for controlling the manufacturing process of space photovoltaic cables, characterized in that, Includes the following steps: Obtain the original dielectric loss measurement of the insulating material at at least one characteristic frequency during the extrusion process; Temperature compensation and digital filtering are performed on the original dielectric loss measurement value to obtain the corrected dielectric loss measurement value. Based on the corrected dielectric loss measurement and the dielectric loss-crosslinking kinetic model in the form of a set of differential equations, the degree of crosslinking at the current moment is determined; The dielectric loss-crosslinking kinetic model includes a crosslinking degree dynamic equation and a dielectric loss output equation. The degree of crosslinking is input into a preset model predictive controller, and a quadratic programming algorithm is used to solve the control quantity sequence output by the model predictive controller. The control quantity sequence includes the extrusion temperature setpoint and the traction speed setpoint. The extrusion temperature and traction speed are adjusted according to the first control quantity in the control quantity sequence.

2. The method for controlling the manufacturing process of a space photovoltaic cable according to claim 1, characterized in that, The dynamic equation for the degree of crosslinking determines the rate of change of the degree of crosslinking based on the melt temperature. The dielectric loss output equation determines the predicted dielectric loss value at the corresponding characteristic frequency based on the degree of crosslinking and the rate of change of the degree of crosslinking.

3. The method for controlling the manufacturing process of a space photovoltaic cable according to claim 1, characterized in that, The acquisition of the original dielectric loss measurement value of the insulating material at at least one characteristic frequency during the extrusion process specifically includes: The original dielectric loss measurements were obtained at three characteristic frequencies, namely a first characteristic frequency, a second characteristic frequency, and a third characteristic frequency. The first characteristic frequency corresponds to the polymer main chain motion, the second characteristic frequency corresponds to the side chain dipole orientation, and the third characteristic frequency corresponds to the micro-interface polarization.

4. The method for controlling the manufacturing process of a space photovoltaic cable according to claim 1, characterized in that, The model prediction controller adopts a cascaded architecture, including an upper-level planner and a lower-level controller; The upper-level planner generates the target dielectric loss trajectory by solving an optimization problem based on the target crosslinking degree and the dielectric loss-crosslinking dynamics model. The lower-level controller takes tracking the dielectric loss target trajectory as its main control objective and solves the control quantity sequence.

5. The method for controlling the manufacturing process of a space photovoltaic cable according to claim 4, characterized in that, The lower-level controller solves for the control quantity sequence, specifically including: Obtain the measured value of the controlled variable at the current moment, wherein the measured value of the controlled variable includes the corrected dielectric loss measurement value and the melt pressure measurement value; Input the measured value of the controlled variable into the prediction model to obtain the predicted value of the controlled variable in the future prediction time domain; Construct an objective function, which includes an error term between the predicted dielectric loss value and the target dielectric loss trajectory, a penalty term for the change in the control quantity, and a penalty term for the absolute value of the control quantity; Under the premise of satisfying the preset constraints, a quadratic programming algorithm is used to solve for the sequence of control variables that minimizes the objective function.

6. The method for controlling the manufacturing process of a space photovoltaic cable according to claim 5, characterized in that, The preset constraints include hard constraints and soft constraints; The hard constraints include the upper and lower limits of the extrusion temperature, the upper and lower limits of the traction speed, and the upper and lower limits of the melt pressure; The soft constraint includes the allowable fluctuation range of dielectric loss, and the objective function also includes a penalty term for violating the soft constraint.

7. The method for controlling the manufacturing process of a space photovoltaic cable according to claim 1, characterized in that, It also includes the step of updating the prediction model parameters of the model prediction controller online: Obtain the measured value of the controlled variable at the current sampling time; Update the process gain matrix based on the measured values ​​of the controlled variable and the recursive least squares method; The updated process gain matrix is ​​validated, including single-step rate of change validation and physical boundary validation. When the verification passes, the prediction model is corrected based on the updated process gain matrix.

8. The method for controlling the manufacturing process of a space photovoltaic cable according to claim 6, characterized in that, It also includes steps for model confidence assessment and control mode switching: Calculate the model confidence index based on the error between historical measurements and historical predictions; When the model confidence index falls below the first confidence threshold, the update of the prediction model parameters is stopped. When the model confidence index is lower than the second confidence threshold and the duration exceeds the preset duration, the system switches to conservative control mode, in which the weight of the penalty term for the change in the control quantity in the objective function is increased.

9. The method for controlling the manufacturing process of a space photovoltaic cable according to claim 3, characterized in that, The original dielectric loss measurement value is subjected to temperature compensation and digital filtering to obtain the corrected dielectric loss measurement value, specifically including: Obtain the real-time temperature of the ring electrode; The original dielectric loss measurement value is corrected for temperature compensation based on the real-time temperature and the preset temperature compensation coefficient. The temperature-compensated dielectric loss measurement is subjected to median filtering and Kalman filtering to obtain the corrected dielectric loss measurement.

10. The method for controlling the manufacturing process of a space photovoltaic cable according to claim 3, characterized in that, The method of obtaining the original dielectric loss measurement value of the insulating material at at least one characteristic frequency during the extrusion process also includes a step of signal validity verification: Determine whether the original dielectric loss measurement value is within the preset range; Determine whether the rate of change of the original dielectric loss measurement value exceeds a preset rate of change threshold; If any judgment result is negative, the current original dielectric loss measurement value is marked as invalid, and the valid dielectric loss measurement value of the previous moment is retained.

11. A space photovoltaic cable, characterized in that, In the traction preparation process, the extrusion temperature and traction speed are adjusted by the process control method for the preparation of space photovoltaic cables as described in any one of claims 1-10.

Images