Liquid metal radiator heat dissipation control method based on model predictive control

Abstract

The invention provides a liquid metal radiator heat dissipation control method based on model predictive control. The control method comprises the following steps that S1, a thermal distribution diagram and liquid metal density distribution are obtained; s2, the path of the liquid metal liquid drops is planned through the thermal distribution diagram, and the positions of the liquid metal liquid drops are determined through liquid metal density distribution; s3, according to the MPC-based electric field control system, the current position information of the liquid metal liquid drops is compared with a set path, and the deviation of the movement path of the liquid metal liquid drops is obtained; and S4, solving a minimum deviation value required by correction by using quadratic programming, taking the minimum deviation value as a correction value, and feeding back the correction value to the MPC-based electric field control system to realize path tracking.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic heat dissipation, and in particular to a heat dissipation control method of a liquid metal radiator based on model predictive control. Background Art

[0002] As electronic devices advance toward miniaturization, integration, and higher frequencies, the heat flux density of these devices is increasing. Thermal failure caused by high temperatures seriously impacts the reliability of these products. As electronic devices heat up, they often form localized high-temperature areas (hot spots), which can lead to performance degradation, aging, and shortened lifespans.

[0003] Currently, the main method to eliminate hot spots is to use thermoelectric cooling sheets and control systems to dynamically control the hot spot temperature. However, this mode is mainly suitable for small devices, and when the temperature difference between the hot and cold ends is small during the thermoelectric cooling process, it will lead to insufficient cooling capacity and thus fail to meet the design requirements.

[0004] Liquid metal droplets have high thermal conductivity and can quickly and efficiently transfer the heat generated by electronic devices. In addition, liquid metal droplets have good shape adaptability. Due to the fluidity of their liquid state, they can flexibly adapt to irregular surfaces and tiny gaps to ensure better thermal contact. However, in order to solve the hot spot problem, it is necessary to accurately control the movement of liquid metal droplets. As an advanced control strategy, Model Predictive Control (MPC) is based on the prediction of the model of system behavior and adjusts the control input through online optimization to enable the system to achieve optimal performance in the future. MPC control has significant advantages in dealing with multivariable systems and considering multiple constraints. It can handle the nonlinearity, time-varying and uncertainty of the system, and adapt to the dynamic changes of the system through continuous online optimization. Therefore, the present invention proposes a liquid metal droplet targeted electronic heat dissipation control method based on MPC, which accurately controls the movement of liquid metal droplets to achieve the purpose of targeted elimination of hot spots, thereby improving the efficiency of the radiator. Summary of the Invention

[0005] The object of the present invention is to provide a liquid metal radiator heat dissipation control method based on model predictive control to solve the problems raised in the above background technology.

[0006] The present invention is achieved through the following technical solutions:

[0007] A liquid metal radiator heat dissipation control method based on model predictive control, the control method comprising the following steps:

[0008] Step S1: Obtaining a thermal distribution map and a liquid metal density distribution;

[0009] Step S2: planning the path of the liquid metal droplet through the thermal distribution map, and confirming the position of the liquid metal droplet through the liquid metal density distribution;

[0010] Step S3: comparing the current position information of the liquid metal droplet with the set path according to the MPC-based electric field control system to obtain the deviation of the liquid metal droplet's motion path;

[0011] Step S4: Use quadratic programming to solve the minimum deviation value required for correction, and use the minimum deviation value as the correction value, and feed this correction value back to the MPC-based electric field control system to achieve path tracking.

[0012] Specifically, the step S1 includes:

[0013] Obtain thermal distribution maps through thermal imaging cameras;

[0014] The density distribution map of the liquid metal is obtained by optical properties or coloring processing of the liquid metal.

[0015] Specifically, the step S2 includes:

[0016] The highest temperature point of each block of the liquid metal radiator is used as a reference for planning the path, and an incremental path planning algorithm is used to plan the path between each highest temperature point. The specific process is as follows:

[0017] Calculate the actual cost from the starting point to node s:

[0018]

[0019] Among them, x s is the horizontal coordinate of node s, y s is the vertical coordinate of node s, x star is the horizontal coordinate of the starting point, y star is the vertical coordinate of the starting point, Temp(s) is the temperature value of node s, Temp(star) is the temperature value of the starting point, K TEM Represents the bias value of the cost function;

[0020] Calculate the expected cost from the starting point to node s:

[0021] rhs(s)=min(c(s′,s)+gs′))rhs(s)

[0022] where rhs(s) is the expected cost of node s, i.e., the estimated minimum cost from the starting point to node s, s′ is the predecessor node of node s, c(s′,s) is the actual cost of moving from predecessor node s′ to node s, and gs′) is the actual cost from the starting point to the predecessor node s′.

[0023] Specifically, step S3 includes the following steps:

[0024] Step S31: establishing a nonlinear model based on the physical formula of motion and the physical constraints to which the liquid metal is subjected;

[0025] Step S32: linearizing the nonlinear model using Taylor's formula to obtain an output matrix change rate error function; discretizing the linear data in the output matrix change rate error function to obtain a discretization equation;

[0026] Step S33: Integrate the discrete data in the discretization equation to obtain a prediction matrix;

[0027] Step S34: inputting the historical data into the MPC-based electric field control system for processing, and obtaining the state variables for a period of time in the future through the prediction matrix;

[0028] Step S35: The current position information of the liquid metal droplet is collected through the liquid metal density distribution map. At the same time, the sensor collects the real-time physical information of the liquid metal droplet, and feeds the current position information of the liquid metal and the real-time physical information back to the MPC-based electric field control system for comparison.

[0029] Specifically, the specific process of step S31 is as follows:

[0030] Establish a nonlinear model of liquid metal motion in the XOY coordinate system;

[0031] Assume that the longitudinal velocity of the liquid metal is v y , the lateral velocity is v x , where the longitudinal velocity v y Controlled by the electric field, the electric field is variable. The longitudinal velocity is a function of the electric field strength E, v. y (E), lateral velocity v x is a constant:

[0032]

[0033] Among them, the longitudinal velocity v y Controlled by the electric field, the electric field is variable. The longitudinal velocity is a function of the electric field strength E, v. y (E), lateral velocity v x is a constant.

[0034] Specifically, the Taylor formula is used in step S32 to linearize the nonlinear model to obtain the output matrix change rate error function. The specific process is as follows:

[0035] Determine the input and output matrices:

[0036] Input Matrix Output Matrix Co-located is the ideal reference state of the system for planning the path, where X is the actual coordinate of the liquid metal droplet in the horizontal direction, Y is the actual coordinate of the liquid metal droplet in the vertical direction, and v yr is the reference velocity of the liquid metal droplet, X r is the reference coordinate of the liquid metal droplet in the horizontal direction, Y r is the reference coordinate of the liquid metal droplet in the longitudinal direction;

[0037] The nonlinear model is linearized using Taylor's formula, which is defined as follows:

[0038]

[0039] Among them, u r The ideal control quantity for planning the path of the ideal system, f(ξ r , u) is the ideal reference state The derivative of the output matrix change rate error function but:

[0040]

[0041] Where A is the coefficient matrix of the output matrix error, and B is the coefficient matrix of the input matrix error; they are defined as follows:

[0042]

[0043] This completes the linearization of the output matrix change rate error function.

[0044] Specifically, in step S32, the linear data in the output matrix change rate error function is discretized to obtain the discretization equation in the following manner:

[0045] Set the sampling period to T, the reference state of the liquid metal in the kth sampling period to ξ(k), the current control quantity in the kth sampling period to u(k), and use the forward Euler formula to discretize:

[0046]

[0047] Among them, Δξ(k) is the output matrix change rate error function of the liquid state in the kth sampling period, Δξ(k+1) is the output matrix change rate error function of the liquid state in the k+1th sampling period, Δu(k) is the current control quantity error in the kth sampling period, is the coefficient matrix of the discretized output matrix error, is the coefficient matrix of the discretized input matrix error.

[0048] Specifically, the specific process of step S33 is as follows:

[0049] The deviation between the error between the current control quantity and the reference value and the error between the control quantity and the reference value at the previous moment is defined as Δ 2 u; if the current control quantity is the reference value of the current control quantity and the historical data of the error between the control quantity and the reference value at the previous moment, then the optimal solution of the current feedback input deviation Δ 2 u(k) is:

[0050] Δ 2 u(k)=Δu(k)-Δu(k-1)

[0051] Where K is the K moment, Δu(k) is the error between the current control quantity and the reference value at the K moment, and Δu(k-1) is the error between the control quantity and the reference value at the k-1 moment;

[0052] And order:

[0053]

[0054] Where x(k+1) is the integration matrix of the output Δξ(k+1) at time k+1 and the error Δu(k) at the current time, and the discretized equation is organized as:

[0055]

[0056] Thus completing the integration of discrete data;

[0057] Through the forward Euler equation of a finite number of x(k), the prediction matrix is ​​obtained, that is:

[0058] D=ψε(k)+θU(k)

[0059] in,

[0060]

[0061] Among them, N p is the number of matrix rows and columns, D is the integration matrix of the prediction matrix, and ψ is the coefficient matrix of the discretized output matrix error , ε(k) is the integration matrix of the state variable x(k) at time k, U(k) is the integration matrix of the current control error of k sampling cycles, and θ is the coefficient matrix of the discretized output matrix error The coefficient matrix of the discretized output matrix error The integration matrix, x(k+N p -1) means k+N p -1 time state variable, x(k) represents the state variable at time k, Δu(k+Np -1) means k+N p -1 moment required state increment, express The square of the matrix, express Matrix N p Power.

[0062] Specifically, the specific process of step S4 is as follows:

[0063] The problem of finding the minimum value of the prediction matrix D that satisfies the constraints is converted into finding the minimum value of the weighted sum of squares of the prediction matrix D. The equation can be listed as follows:

[0064] f[U(k)]=D T QD+U(k) T RU(k)

[0065] Let V = ψε(k), then substitute into D = ψε(k) + θU(k) to obtain:

[0066] D=V+θU(k)

[0067] but:

[0068] f[U(k)]=V T QV+[θU(k)] T Q[θU(k)]+2V T QθU(k)+U(k) T RU(k)

[0069] Among them, Q is the weight matrix of D matrix, R is the weight matrix of U matrix, V T QV is a constant during the feedback process, so the minimum deviation required to correct the constraint is calculated and fed back:

[0070] g[U(k)]=[θU(k)] T Q[θU(k)]+2V T Q[θU(k)]+U(k) T RU(k)

[0071] =U(k) T (θ T Qθ+R)U(k)+2V T QθU(k)

[0072] Among them, g[U(k)] is constrained by the physical constraints of liquid metal as the upper and lower bounds of U(k).

[0073] Compared with the prior art, the present invention has the following beneficial effects:

[0074] The present invention provides a liquid metal radiator heat dissipation control method based on model predictive control. The method plans the path of liquid metal droplets through a thermal distribution map, and confirms the position of the liquid metal droplets through the liquid metal density distribution. The current position information of the liquid metal droplets is compared with the set path according to the MPC-based electric field control system to obtain the deviation of the liquid metal droplet movement path. The quadratic programming is used to solve the minimum deviation value required for correction, and the minimum deviation value is used as the correction value. This correction value is fed back to the MPC-based electric field control system. The MPC-based electric field control system corrects the longitudinal flow path of the liquid metal according to the correction value, increases the liquid metal concentration near the hot spot, thereby enhancing heat transfer, and further improving the heat dissipation efficiency of the electronic radiator, effectively solving the problem of excessively high hot spot temperatures. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only preferred embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0076] Figure 1 A flow chart of the steps of a liquid metal radiator heat dissipation control method based on model predictive control provided by the present invention.

[0077] Figure 2 Schematic diagram for planning the path of a liquid metal droplet using a thermodynamic map. DETAILED DESCRIPTION

[0078] In order to make the purpose, technical solutions and advantages of the present invention more apparent, exemplary embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments of the present invention, and it should be understood that the present invention is not limited to the exemplary embodiments described herein. Based on the embodiments of the present invention described in the present invention, all other embodiments obtained by those skilled in the art without creative work should fall within the scope of protection of the present invention.

[0079] In the following description, numerous specific details are provided to provide a more thorough understanding of the present invention. However, it will be apparent to those skilled in the art that the present invention may be practiced without one or more of these details. In other instances, certain technical features well known in the art are not described to avoid confusion with the present invention.

[0080] It should be understood that the present invention can be implemented in different forms and should not be interpreted as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to make disclosure thorough and complete and to fully convey the scope of the present invention to those skilled in the art.

[0081] The purpose of the terms used herein is only to describe specific embodiments and is not intended to limit the present invention. When used herein, the singular forms "a", "an", and "the" are intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the terms "comprising" and / or "comprising", when used in this specification, determine the presence of the features, integers, steps, operations, elements and / or parts, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, parts and / or groups. When used herein, the term "and / or" includes any and all combinations of the relevant listed items.

[0082] In order to fully understand the present invention, a detailed structure will be provided in the following description to illustrate the technical solution proposed by the present invention. Optional embodiments of the present invention are described in detail below. However, in addition to these detailed descriptions, the present invention may also have other implementations.

[0083] See also Figure 1 A liquid metal radiator heat dissipation control method based on model predictive control, the control method comprising the following steps:

[0084] Step S1: Obtaining a thermal distribution map and a liquid metal density distribution;

[0085] Step S2: planning the path of the liquid metal droplet through the thermal distribution map, and confirming the position of the liquid metal droplet through the liquid metal density distribution;

[0086] Step S3: comparing the current position information of the liquid metal droplet with the set path according to the MPC-based electric field control system to obtain the deviation of the liquid metal droplet's motion path;

[0087] Step S4: Use quadratic programming to solve the minimum deviation value required for correction, and use the minimum deviation value as the correction value, and feed this correction value back to the MPC-based electric field control system to achieve path tracking.

[0088] In this embodiment, the present invention provides a liquid metal heat sink heat dissipation control method based on model predictive control. The method obtains historical data of thermal distribution diagrams and liquid metal density distribution, establishes an MPC-based electric field control system, inputs the historical data into the MPC-based electric field control system for processing, and obtains predictive control parameters through a prediction matrix. The device temperature is collected, and the current device temperature information and real-time temperature are fed back to the MPC-based electric field control system. The MPC-based electric field control system compares the current device temperature with the set temperature, analyzes the deviation of the device temperature, and calculates a correction value. The MPC-based electric field control system corrects the longitudinal flow path of the liquid metal based on the correction value, increases the liquid metal concentration near the hot spot, and thus enhances heat transfer. The present invention improves the heat dissipation efficiency of the electronic heat sink and effectively solves the problem of excessively high hot spot temperatures.

[0089] Specifically, the step S1 includes:

[0090] The thermal distribution map is obtained by a thermal imager; the density distribution map of the liquid metal is obtained by the optical properties or coloring treatment of the liquid metal.

[0091] Specifically, the step S2 includes:

[0092] See also Figure 2 , taking the highest temperature point of each block of the liquid metal radiator as the reference planning path, an incremental path planning algorithm is used to plan the path between each highest temperature point. The specific process is as follows:

[0093] Calculate the actual cost from the starting point to node s:

[0094]

[0095] Among them, x s is the horizontal coordinate of node s, y s is the vertical coordinate of node s, x star is the horizontal coordinate of the starting point, y star is the vertical coordinate of the starting point, Temp(s) is the temperature value of node s, Temp(star) is the temperature value of the starting point, K TEM Represents the bias value of the cost function, when K TEM When it is close to 0, it means that the cost function is more inclined to the distance cost. TEM When it is close to 1, it means that the cost function is more biased towards temperature cost;

[0096] Calculate the expected cost from the starting point to node s:

[0097] rhs(s)=min(c(s′,s)+gs′))rhs(s)

[0098] where rhs(s) is the expected cost of node s, i.e., the estimated minimum cost from the starting point to node s, s′ is the predecessor node of node s, c(s′,s) is the actual cost of moving from predecessor node s′ to node s, and gs′) is the actual cost from the starting point to the predecessor node s′.

[0099] For example, the basic process of planning the path between each highest temperature point using the incremental path planning algorithm is as follows:

[0100] (1) Initialization: Initialize g(s) and rhs(s) of all nodes to infinity (indicating an unknown state), set rhs(s) of the starting point to 0, indicating that the minimum cost from the starting point to itself is 0, add the starting point to the priority queue, and sort it according to the calculated priority (key). Use the k(s) value to calculate the priority. The k(s) value consists of two parts, k1(s) and k2(s). K1(s) is compared first. If the k1(s) of two points are the same, k2(s) is compared. Among them, h(s,goal) is the estimated distance from point s to the target point goal. The present invention uses the absolute distance to the target node to express it, that is,

[0101] (2) Update function: When the environment changes or new path information is available, the rhs(s) value of the node is updated. For node s, the way to update its rhs(s) value is: rhs(s) = min(rhs(s′), c(s′, s) + gs′));

[0102] (3) Main loop: Select the node with the highest priority in the priority queue for processing, update the strategy based on the g(s) and rhs(s) values ​​of the current node and its neighboring nodes, and adjust the path information. If the g(s) value of the current node changes, it may affect the rhs(s) value of its neighboring nodes, so the path information of its neighboring nodes needs to be updated. If the rhs(s) value of the target node is not equal to the g(s) value, or there are still nodes to be processed in the priority queue, continue the loop;

[0103] (4) Path Extraction Once the rhs(s) value of the target node is equal to the g(s) value and the priority queue is empty (i.e., there is no need to change the priority), the optimal path from the start point to the target node can be extracted.

[0104] Specifically, step S3 includes the following steps:

[0105] Step S31: establishing a nonlinear model based on the physical formula of motion and the physical constraints to which the liquid metal is subjected;

[0106] Step S32: linearizing the nonlinear model using Taylor's formula to obtain an output matrix change rate error function; discretizing the linear data in the output matrix change rate error function to obtain a discretization equation;

[0107] Step S33: Integrate the discrete data in the discretization equation to obtain a prediction matrix;

[0108] Step S34: inputting the historical data into the MPC-based electric field control system for processing, and obtaining the prediction control parameters through the prediction matrix;

[0109] Step S35: The current position information of the liquid metal droplet is collected through the liquid metal density distribution map. At the same time, the sensor collects the real-time physical information of the liquid metal droplet, and feeds the current position information of the liquid metal and the real-time physical information back to the MPC-based electric field control system for comparison.

[0110] Specifically, the specific process of step S31 is as follows:

[0111] Establish a nonlinear model of liquid metal motion in the XOY coordinate system;

[0112] Assume that the longitudinal velocity of the liquid metal is v y , the lateral velocity is v x , where the longitudinal velocity v y Controlled by the electric field, the electric field is variable. The longitudinal velocity is a function of the electric field strength E, v. y (E), lateral velocity v x is a constant:

[0113]

[0114] Among them, the longitudinal velocity v y Controlled by the electric field, the electric field is variable. The longitudinal velocity is a function of the electric field strength E, v. y (E), lateral velocity v x is a constant.

[0115] Specifically, the Taylor formula is used in step S32 to linearize the nonlinear model to obtain the output matrix change rate error function. The specific process is as follows:

[0116] Determine the input and output matrices:

[0117] Input Matrix Output Matrix Co-located is the ideal reference state for the system to plan the path, X is the actual coordinate of the liquid metal droplet in the horizontal direction, Y is the actual coordinate of the liquid metal droplet in the vertical direction, and v yr is the reference velocity of the liquid metal droplet, Xr is the reference coordinate of the liquid metal droplet in the horizontal direction, Y r is the reference coordinate of the liquid metal droplet in the longitudinal direction;

[0118] The nonlinear model is linearized using Taylor's formula, which is defined as follows:

[0119]

[0120] u、ξ、ξ r Substituting into Taylor's formula, R n (x) is the remainder of Taylor's formula, which gives:

[0121]

[0122] Among them, u r The ideal control quantity for planning the path of the ideal system, f(ξ r , u) is the ideal reference state Derivative, let the output matrix change rate error function be Therefore:

[0123]

[0124] Where A is the coefficient matrix of the output matrix error, and B is the coefficient matrix of the input matrix error; they are defined as follows:

[0125]

[0126] This completes the linearization of the output matrix change rate error function.

[0127] Specifically, in step S32, the linear data in the output matrix change rate error function is discretized to obtain the discretization equation in the following manner:

[0128] Set the sampling period to T, the reference state of the liquid metal in the kth sampling period to ξ(k), the current control quantity in the kth sampling period to u(k), and use the forward Euler formula:

[0129]

[0130] Discretize and substitute the liquid metal state parameters into:

[0131]

[0132] And organize it into:

[0133]

[0134] Among them, Δξ(k) is the output matrix change rate error function of the liquid state in the kth sampling period, Δξ(k+1) is the output matrix change rate error function of the liquid state in the k+1th sampling period, Δu(k) is the current control quantity error in the kth sampling period, is the coefficient matrix of the discretized output matrix error, is the coefficient matrix of the discretized input matrix error, that is:

[0135]

[0136] This completes the discretization of the output matrix change rate error function and obtains the discretization equation.

[0137] Specifically, the specific process of step S33 is as follows:

[0138] The deviation between the error between the current control quantity and the reference value and the error between the control quantity and the reference value at the previous moment is defined as Δ 2 u; It is known that the current control quantity is the reference value of the current control quantity, the historical data of the error between the control quantity and the reference value at the previous moment, and the optimal solution Δ of the current feedback input deviation 2 u(k), that is:

[0139] u(k)=u r (k)+×u(k-1)+Δ 2 u(k)

[0140] Δ 2 u(k)=Δu(k)-Δu(k-1)

[0141] Among them, K is the Kth sampling period, u r (k) and Δu(k-1) are known data. If we need to find the current input u(k), we need to find Δ 2 u(k);

[0142] Therefore, it is ordered:

[0143]

[0144] Among them, x(k+1) is the integration matrix of the output Δξ(k+1) at time k+1 and the deviation Δu(k) at the current time, and the discretization equation is organized as:

[0145]

[0146] Thus completing the integration of discrete data;

[0147] List a finite number of x(k) forward Euler equations and get the prediction matrix, that is:

[0148]

[0149] The prediction matrix is:

[0150] D=ψε(k)+θU(k)

[0151] in,

[0152]

[0153] Among them, N p is the number of matrix rows and columns, D is the integration matrix of the prediction matrix, and ψ is the coefficient matrix of the discretized output matrix error The integration matrix of ε(k) is the integration matrix of the output Δξ(k) at time k and the deviation Δu(k) at the current moment, U(k) is the integration matrix of the current control error of k sampling cycles, and θ is the coefficient matrix of the discretized output matrix error The coefficient matrix of the discretized output matrix error The integration matrix, x(k+N p -1) means k+N p -1 time state variable, x(k) represents the state variable at time k, Δu(k+N p -1) means k+N p -1 moment required state increment, express The square of the matrix, express Matrix N p Power.

[0154] Specifically, the specific process of step S4 is as follows:

[0155] For example, establish a solution equation; due to the liquid metal state error XX r , YY r , v y -v yr The importance is different in the actual scenario of liquid metal flow. The problem of finding the minimum value of the prediction matrix D that satisfies the constraints is converted into finding the minimum value of the weighted square sum of the prediction matrix D. The equation can be listed as follows:

[0156] f[U(k)]=D T QD+U(k) T RU(k)

[0157] Let V = ψε(k), then substitute into D = ψε(k) + θU(k) to obtain:

[0158] D=V+θU(k)

[0159] but:

[0160] f[U(k)]=[V+θU(k)] T Q[V+θU(k)]+U T RU

[0161] =V T QV+[θU(k)] T Q[θU(k)]+2V T QθU(k)+

[0162] U(k) T TU(k)

[0163] Among them, Q is the weight matrix of D matrix, R is the weight matrix of U matrix, V T QV is a constant in the feedback process, then:

[0164] g[U(k)]=[θU(k)] T Q[θU(k)]+2V T Q[θU(k)]+U(k) T RU(k)

[0165] =U(k) T (θ T Qθ+R)U(k)+2V T QθU(k)

[0166] g[U(k)] is constrained by the physical constraints of liquid metal as the upper and lower bounds of U(k). Finally, U(k) is substituted into the quadratic programming minimum algorithm to obtain and feed back the minimum deviation value required for the correction to satisfy the constraints.

[0167] Exemplarily, the MPC-based electric field control system corrects the longitudinal flow path of the liquid metal according to the minimum deviation value obtained as a correction value, accurately increasing the concentration of the liquid metal near the hot spot, thereby enhancing heat transfer.

[0168] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A liquid metal radiator heat dissipation control method based on model predictive control, It is characterized in that The control method comprises the following steps: Step S1: obtaining a thermal distribution diagram and a liquid metal density distribution; Step S2: planning the path of the liquid metal droplet through the thermal distribution map, and confirming the position of the liquid metal droplet through the liquid metal density distribution; Step S3: comparing the current position information of the liquid metal droplet with the set path according to the MPC-based electric field control system to obtain the deviation of the liquid metal droplet movement path; Step S4: Use quadratic programming to solve the minimum deviation value required for correction, and use the minimum deviation value as the correction value, and feed this correction value back to the MPC-based electric field control system to achieve path tracking.

2. According to claim 1, a liquid metal radiator heat dissipation control method based on model predictive control, It is characterized in that The step S1 specifically includes: Get thermal distribution map through thermal imager; The density distribution map of the liquid metal is obtained through the optical properties or coloring treatment of the liquid metal.

3. According to claim 2, a liquid metal radiator heat dissipation control method based on model predictive control, It is characterized in that The step S2 specifically includes: The highest temperature point of each block of the liquid metal radiator is used as a reference for planning the path, and an incremental path planning algorithm is used to plan the path between each highest temperature point. The specific process is as follows: Calculate the actual cost from the starting point to node s: Among them, x s is the horizontal coordinate of node s, y s is the ordinate of node s, x star is the horizontal coordinate of the starting point, y star is the ordinate of the starting point, Temp(s) is the temperature of node s, Temp(star) is the temperature of the starting point, K TEM Represents the bias value of the cost function; Calculate the expected cost from the starting point to node s: rhs(s)=min(c(s′,s)+g(s′))rhs(s) Among them, rhs(s) is the expected cost of node s, that is, the minimum cost estimate from the starting point to node s, s′ is the predecessor node of node s, c(s′,s) is the actual moving cost from the predecessor node s′ to node s, and g(s′) is the actual cost from the starting point to the predecessor node s′.

4. According to claim 2, a liquid metal radiator heat dissipation control method based on model predictive control, It is characterized in that The step S3 specifically comprises the following steps: Step S31: establishing a nonlinear model according to the physical formula of motion and the physical constraints to which the liquid metal is subjected; Step S32: linearize the nonlinear model using Taylor's formula to obtain an output matrix change rate error function; discretize the linear data in the output matrix change rate error function to obtain a discretization equation; Step S33: Integrate the discrete data in the discretization equation to obtain a prediction matrix; Step S34: inputting the historical data into the MPC-based electric field control system for processing, and obtaining the state variables for a period of time in the future through the prediction matrix; Step S35: The current position information of the liquid metal droplet is collected through the liquid metal density distribution map. At the same time, the sensor collects the real-time physical information of the liquid metal droplet, and the current position information of the liquid metal and the real-time physical information are fed back to the MPC-based electric field control system for comparison.

5. According to claim 4, a liquid metal radiator heat dissipation control method based on model predictive control, It is characterized in that The specific process of step S31 is as follows: Establish a nonlinear model of liquid metal motion in the XOY coordinate system; Assume the longitudinal velocity of the liquid metal is v y , the lateral velocity is v x , where the longitudinal velocity v y Controlled by the electric field, the electric field is variable. The longitudinal velocity is a function of the electric field strength E, v. y (E), lateral velocity v x is a constant: The longitudinal velocity v y Controlled by the electric field, the electric field is variable. The longitudinal velocity is a function of the electric field strength E, v. y (E), lateral velocity v x is a constant.

6. A liquid metal radiator heat dissipation control method based on model predictive control according to claim 5, It is characterized in that In step S32, the nonlinear model is linearized by using Taylor's formula to obtain the output matrix change rate error function. The specific process is as follows: Determine the input matrix and output matrix: Input Matrix Output Matrix Concurrently is the ideal reference state of the system for planning the path, where X is the actual coordinate of the liquid metal droplet in the horizontal direction, Y is the actual coordinate of the liquid metal droplet in the vertical direction, and v yr is the reference velocity of the liquid metal droplet, X r is the reference coordinate of the liquid metal droplet in the horizontal direction, Y r is the reference coordinate of the liquid metal droplet in the longitudinal direction; The nonlinear model is linearized using Taylor's formula, which is defined as follows: Among them, u r The ideal control quantity for planning the path of the ideal system, f(ξ r , u) is the ideal reference state The derivative of the output matrix change rate error function but: Among them, A is the coefficient matrix of the output matrix error, and B is the coefficient matrix of the input matrix error; they are defined as follows: This completes the linearization of the output matrix change rate error function.

7. A liquid metal radiator heat dissipation control method based on model predictive control according to claim 6, It is characterized in that In step S32, the linear data in the output matrix change rate error function is discretized to obtain the discretization equation as follows: The sampling period is set to T, the reference state of the liquid metal in the kth sampling period is ξ(k), the current control quantity in the kth sampling period is u(k), and the forward Euler formula is used for discretization: Among them, Δξ(k) is the output matrix change rate error function of the liquid in the kth sampling period, Δξ(k+1) is the output matrix change rate error function of the liquid in the k+1th sampling period, Δu(k) is the current control amount error in the kth sampling period, is the coefficient matrix of the discretized output matrix error, is the coefficient matrix of the discretized input matrix error.

8. The method for controlling heat dissipation of a liquid metal radiator based on model predictive control according to claim 4, It is characterized in that The specific process of step S33 is as follows: Define the deviation between the error between the current control amount and the reference value and the error between the control amount and the reference value at the previous moment as Δ 2 u; if the current control quantity is the reference value of the current control quantity and the historical data of the error between the control quantity and the reference value at the previous moment, then the optimal solution of the current feedback input deviation is Δ 2 u(k) is: D 2 u(k)=Δu(k)-Δu(k-1) Where K is the K moment, Δu(k) is the error between the current control value and the reference value at the K moment, and Δu(k-1) is the error between the control value and the reference value at the k-1 moment; And order: Among them, x(k+1) is the integration matrix of the output Δξ(k+1) at time k+1 and the error Δu(k) at the current time, and the discretization equation is organized as: Thereby completing the integration of discrete data; Through the forward Euler equation of a finite number of x(k), the prediction matrix is ​​obtained, that is: D=ψε(k)+θU(k) in, Among them, N p is the number of matrix rows and columns, D is the integration matrix of the prediction matrix, and ψ is the coefficient matrix of the discretized output matrix error , ε(k) is the integration matrix of the state variable x(k) at time k, U(k) is the integration matrix of the current control error of k sampling cycles, and θ is the coefficient matrix of the discretized output matrix error The coefficient matrix of the error with the discretized output matrix The integration matrix, x(k+N p -1) means k+N p -1 time, x(k) represents the state variable at time k, Δu(k+N p -1) means k+N p -1 The required state increment at time express The square of the matrix, express Matrix N p Power.

9. The method for controlling heat dissipation of a liquid metal radiator based on model predictive control according to claim 1, It is characterized in that The specific process of step S4 is as follows: The problem of finding the minimum value of the prediction matrix D that satisfies the constraints is converted into finding the minimum value of the weighted sum of squares of the prediction matrix D. The equation can be listed as follows: f[U(k)]=D T QD+U(k) T RU(k) Let V = ψε(k), then substitute into D = ψε(k) + θU(k) to obtain: D=V+θU(k) but: f[U(k)]=V T QV+[θU(k)] T Q[θU(k)]+2V T QθU(k)+U(k) T RU(k) Among them, Q is the weight matrix of D matrix, R is the weight matrix of U matrix, V T QV is a constant during the feedback process, so the minimum deviation required for the correction to satisfy the constraint is calculated and fed back: g[U(k)]=[θU(k)] T Q[θU(k)]+2V T Q[θU(k)]+U(k) T RU(k) =U(k) T (θ T Qθ+R)U(k)+2V T QθU(k) Among them, g[U(k)] is constrained by the physical constraints of liquid metal as the upper and lower bounds of U(k).

Images