A high-voltage variable-frequency power supply thermal management control method for improving energy efficiency

Abstract

The application relates to the technical field of power heat management control, and discloses a high-voltage variable-frequency power heat management control method for improving energy efficiency, which comprises the following steps: combining a solar high-temperature heat power generation system to obtain a low-dimension model through order reduction; using a Gaussian process regression and Kalman filtering algorithm to perform data fusion and reconstruct temperature field information of the system; establishing a multi-objective optimization function to dynamically adjust the weight optimization target; using an improved particle swarm optimization algorithm to generate a control parameter sequence, and adjusting the control decision in real time through iterative optimization; inputting the optimized control parameter into a multi-rate hierarchical control system to perform work in shifts; and detecting a fault type and switching to a corresponding fault-tolerant control mode. The application combines a multi-physical field coupling model and intrinsic orthogonal decomposition technology to establish an accurate coupling model of an electromagnetic field, a temperature field and a flow field in a high-voltage variable-frequency power generation system, the dimension of the model is reduced through order reduction, the calculation complexity is significantly reduced, and the calculation efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power heat management control, in particular to a high-voltage variable-frequency power supply heat management control method for improving energy efficiency. BACKGROUND

[0002] As a key equipment in modern industry and energy systems, high-voltage variable-frequency power supply is widely used in power, chemical industry, mining and other fields. Its main function is to realize precise control of power by adjusting voltage and frequency to meet the power demand of different loads and working conditions. With the improvement of industrial automation level and the increasingly stringent requirements of energy saving and emission reduction, the energy efficiency and stability of high-voltage variable-frequency power supply have become the core factors to ensure the efficient operation of the system.

[0003] The existing heat management technology mainly relies on temperature field model or flow field model, which adjusts the cooling system through simplified numerical analysis of temperature change or fluid cooling effect. Common cooling methods include air cooling, liquid cooling, etc. These methods usually adjust the fan speed, liquid flow and other means to keep the temperature of the power supply equipment within a safe range.

[0004] However, the existing heat management control technology only analyzes the temperature field or flow field independently, lacks comprehensive consideration of the mutual coupling of electromagnetic field, temperature field and flow field, resulting in low model accuracy and difficulty in fully reflecting the thermal state of the system. Secondly, the traditional multi-objective optimization method usually focuses on the optimization of a single target, ignoring the comprehensive optimization of multiple targets. Therefore, the present application provides a high-voltage variable-frequency power supply heat management control method for improving energy efficiency to solve the problems existing in the prior art. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a high-voltage variable-frequency power supply heat management control method for improving energy efficiency, which solves the problems of single physical field analysis, lack of multi-field coupling, insufficient precision and insufficient multi-objective optimization of the prior art.

[0006] To achieve the above purpose, the present application realizes the following technical scheme: a high-voltage variable-frequency power supply heat management control method for improving energy efficiency, comprising the following steps:

[0007] According to the principles of electromagnetism, heat conduction and fluid mechanics, combined with the structure and operating environment of the solar high-temperature thermal power generation system, a coupled model of electromagnetic field, temperature field and flow field is established, and a low-dimensional model is obtained through eigenvalue orthogonal decomposition for reduction;

[0008] Based on the obtained low-dimensional model, temperature sensors, heat flow sensors and flow rate sensors are arranged at different positions in the solar thermal power generation system, the data of the power generation system is collected, and the collected data is fused using Gaussian process regression and Kalman filter algorithm to reconstruct the temperature field of the power generation system.

[0009] By reconstructing the temperature field information, a multi-objective optimization function is established, which comprehensively considers the objectives of temperature difference, cooling power, temperature field uniformity and thermal energy storage efficiency, dynamically adjusts the weights, and optimizes the objective priority in real time.

[0010] Based on the established multi-objective optimization function, an improved particle swarm optimization algorithm is used to generate a sequence of control parameters. The control parameters are then iteratively optimized through particle swarm optimization, and control decisions are adjusted based on real-time data and prediction results.

[0011] The optimized control parameters are input into the multi-rate hierarchical control system, which is executed by high-speed loop, medium-speed loop and low-speed loop respectively, and the data is synchronized by the state buffer.

[0012] By executing the control process in real time, the system status is monitored and the types of faults occurring in the power generation system are detected. Based on the type of fault, the system switches to the corresponding fault-tolerant control mode.

[0013] Preferably, the establishment of the coupled electromagnetic field, temperature field, and flow field model includes the following steps:

[0014] By combining the photothermal conversion process of solar high-temperature thermal power generation system with the principles of heat conduction and fluid mechanics, the mutual coupling between light radiation, temperature field and cooling flow field is first considered in the physical model.

[0015] Partial differential equations are used to describe the coupling relationship between light radiation, temperature and fluid field, and the transient response of each physical field is calculated. The focus is on the heat conduction and heat recovery processes of solar collectors, thermal energy storage systems and heat exchangers.

[0016] The high-dimensional coupled equations are reduced in order by using the intrinsic orthogonal decomposition method, preserving the main physical features and reducing the computational load, resulting in a low-dimensional coupled model of photothermal conversion, temperature field and flow field.

[0017] Preferably, the process of fusing the collected data using Gaussian process regression and Kalman filtering algorithms includes the following steps:

[0018] After obtaining the low-dimensional model, temperature sensors, heat flow sensors, and flow velocity sensors are placed at different locations in the solar high-temperature thermal power generation system to collect system status information in real time, with a focus on the status of the collector, heat exchanger, and thermal energy storage system.

[0019] By using the Gaussian process regression algorithm to spatially interpolate the sensor data, the accurate values ​​of the temperature field and flow field are inferred, ensuring the accurate reconstruction of the collector surface temperature and heat exchange efficiency.

[0020] By combining the Kalman filter algorithm, the sensor data is dynamically updated over time to reconstruct the temperature field of the solar power generation system and adapt to changes in different weather and light intensity.

[0021] Preferably, establishing the multi-objective optimization function includes the following steps:

[0022] Based on the reconstructed temperature field information, a multi-objective optimization function is established to simultaneously optimize temperature difference, cooling power, temperature field uniformity, and thermal energy storage efficiency.

[0023] The multi-objective optimization function is expressed as follows:

[0024] J(t) = α·max(ΔT) + β·∫P cool dt+γ·σ T ;

[0025] Where ΔT is the maximum temperature difference within the system, P cool For the system cooling power, σ T For the uniformity of the temperature field, α, β, γ are dynamically adjusted weighting coefficients, dt is the time step, and J(t) is the value of the multi-objective optimization function.

[0026] The weighting coefficients are dynamically adjusted based on real-time operating conditions and environmental changes, especially optimized according to light intensity, ambient temperature, and system load conditions.

[0027] Preferably, generating the control parameter sequence using the improved particle swarm optimization algorithm includes the following steps:

[0028] In the particle swarm optimization algorithm, the initial position and velocity of each particle in the swarm are first set, and the fitness of the current particle is calculated. The fitness function is defined according to the temperature field, cooling power and thermal energy storage target.

[0029] In each iteration, the particle's velocity and position are updated, where the particle velocity update equation is:

[0030]

[0031] in, This represents the velocity update value of particle i in the (k+1)th generation, ω is the inertial weight, c1, c2, and c3 are constant acceleration factors, and pbest i Let be the optimal position reached by particle i in the historical search. Let r1 be the current position of particle i in the kth generation, gbest be the globally optimal position found by all particles in the current generation, and r1, r2, and r3 be the coefficients in the particle velocity update equation. The gradient term of the cost function is represented by the inertia weight ω, which is adjusted according to the convergence requirements of the algorithm.

[0032] Through multiple particle swarm iterations, a set of optimized control parameters is finally obtained to adapt to the adjustment of solar collectors and heat exchangers under different lighting conditions.

[0033] Preferably, inputting the optimized control parameters into the multi-rate hierarchical control system includes the following steps:

[0034] The optimized control parameters are input into a multi-rate hierarchical control system, which includes a high-speed loop, a medium-speed loop, and a low-speed loop.

[0035] The high-speed loop performs real-time calculations of control parameters for particle swarm optimization, and adjusts the operation of the collector and cooling system. The medium-speed loop dynamically predicts the thermal field model of the power generation system and adjusts the heat release of the thermal energy storage system. The low-speed loop is responsible for executing physical drive and cooling operations.

[0036] By synchronizing data across loops through the state buffer, the control operations of each loop are coordinated and consistent.

[0037] Preferably, the step of switching to the corresponding fault-tolerant control mode according to the fault type includes the following steps:

[0038] By monitoring the real-time status of the system, a residual detection mechanism is used to determine whether a fault exists. If the system detects a fault type, a fault-tolerant control mode is triggered.

[0039] In the event of sensor failure, model predictive control is used to replace the function of the damaged sensor, and the thermal management process is controlled by the predicted value.

[0040] In the event of an actuator failure, the system switches to a safe mode and controls system operation by setting limits.

[0041] Preferably, the step of reducing the order of the obtained high-dimensional coupled equation system using the intrinsic orthogonal decomposition method includes:

[0042] Collect a snapshot matrix X = [x1, x2, ..., x] of the state variables of the coupled model over time. m ], where X represents a snapshot matrix of state variables collected over time, each Indicates the state of the power generation system at a certain moment, x i This represents the state of the power generation system at time i, and n represents the dimension of the state variables of the power generation system.

[0043] Perform singular value decomposition on the snapshot matrix:

[0044] X=U∑V T ;

[0045] Where U is the left singular vector matrix, ∑ is the singular value diagonal matrix, and V TIt is the transpose of the right singular vector matrix;

[0046] Select the first r principal modes to form the dimensionless basis Φ = [u1, u2, ..., u r Projecting the state of a high-dimensional system onto a low-dimensional subspace:

[0047] x(t)≈Φa(t);

[0048] Where r represents the number of principal components, Φ = [u1, u2, ..., u r ] represents the most representative basis vector extracted from the original state data, a(t) is a low-dimensional state variable, Φ is the POD mode matrix, and x(t)≈Φa(t) indicates that the POD method was used.

[0049] Preferably, the dynamic adjustment of the weight coefficients is driven by a Lyapunov-guided strategy, and the adjustment formula is as follows:

[0050]

[0051] Where α, β, and γ represent three dynamic weighting factors in multi-objective optimization. This represents the rate of change of these three weights over time, with k1, k2, k3 > 0 being adjustment factors. This represents the partial derivative of the objective function with respect to each sub-objective.

[0052] A high-voltage variable frequency power supply thermal management control system for improving energy efficiency is also provided, including:

[0053] The multiphysics modeling and order reduction module is used to establish a coupled model of electromagnetic field, temperature field and flow field based on the principles of electromagnetics, thermal conductivity and fluid mechanics, combined with the system structure and operating conditions, and to reduce the order of the model based on the intrinsic orthogonal decomposition method.

[0054] The sensor acquisition and temperature rise detection module is used to deploy temperature sensors, heat flow sensors and flow velocity sensors in key components and overheat-sensitive areas of the solar thermal power generation system to collect multi-point operating data and analyze it by fusion through Gaussian process regression and Kalman filter algorithm.

[0055] The thermal safety assessment and protection strategy module is used to combine real-time temperature field reconstruction results with historical thermal runaway characteristic curves to determine whether the preset temperature protection threshold has been reached, construct a multi-objective protection function, and form a dynamic safety control strategy.

[0056] The control parameter optimization generation module is used to generate control parameter sequences based on thermal safety control strategies using an improved particle swarm optimization algorithm, and introduces temperature rise weighting factors and gradient feedback to improve control accuracy.

[0057] The multi-rate thermal control execution module is used to receive optimized control parameters and execute them in layers according to the control strategies of high-speed loop, medium-speed loop and low-speed loop, while using the state buffer to realize cross-rate control data synchronization.

[0058] The fault identification and redundancy fault tolerance module is used to identify and locate sensor faults, actuator failures, or temperature anomalies that may occur during system operation, and switch to model prediction fault tolerance control mode or enable redundant equipment according to the fault type.

[0059] This invention provides a thermal management control method for high-voltage variable frequency power supplies to improve energy efficiency. It has the following beneficial effects:

[0060] 1. This invention establishes an accurate coupled model of electromagnetic, temperature, and flow fields in a high-voltage variable frequency power generation system by combining a multiphysics coupling model with intrinsic orthogonal decomposition technology. This model can accurately simulate the interactions between different physical fields, especially the impact of electromagnetic field changes on temperature distribution and fluid flow. Through order reduction processing, the model's dimensionality is reduced, significantly lowering computational complexity and improving computational efficiency. It comprehensively considers the coupling effects of multiple physical fields in the system, avoiding the high computational cost and low efficiency problems caused by complex interactions in existing technologies.

[0061] 2. This invention employs a combined algorithm of Gaussian process regression and Kalman filtering, which can optimize the prediction accuracy of temperature and flow fields in real time even when sensor data is incomplete or noisy. Through data fusion, the system can still provide accurate temperature reconstruction information even when temperature sensors malfunction or are lost. This is superior to traditional prediction methods based on a single sensor, avoiding inaccurate temperature predictions caused by sensor failure or data loss, thereby improving the reliability of the system's thermal management.

[0062] 3. This invention employs a particle swarm optimization algorithm to dynamically generate control parameter sequences, achieving optimization of multiple objectives in the high-voltage variable frequency power supply system, including temperature difference, cooling power, and temperature field uniformity. The iterative optimization of the particle swarm optimization algorithm not only adjusts the control strategy based on real-time data and system prediction results but also adapts to changes in system load and environment, optimizing thermal management, reducing energy consumption, and improving system stability. Compared to traditional fixed control strategies, the control scheme of this invention is more flexible and precise, and can dynamically adjust to changes in load and environmental conditions to ensure the system always remains within a safe temperature range.

[0063] 4. This invention, through the combination of residual analysis and model predictive control, can identify sensor or actuator faults in real time and automatically switch to fault-tolerant control mode. This mechanism enables the system to maintain temperature control functionality and prevent overheating caused by faults even in the event of sensor or actuator failure. Compared with traditional fault-tolerant control methods, this invention ensures stable system operation and avoids overheating or equipment damage even when equipment malfunctions by dynamically adjusting the weights of the optimization objectives. Attached Figure Description

[0064] Figure 1 This is a flowchart of the method steps of the present invention;

[0065] Figure 2 This is a system architecture diagram of the present invention. Detailed Implementation

[0066] The technical solutions in 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.

[0067] Please see the appendix Figure 1 This invention provides a high-voltage variable frequency power supply thermal management control method to improve energy efficiency, comprising the following steps:

[0068] S1. Based on the principles of electromagnetism, thermal conductivity and fluid mechanics, and combined with the structure and operating environment of the solar high-temperature thermal power generation system, a coupled model of electromagnetic field, temperature field and flow field is established. The model is then reduced in order through intrinsic orthogonal decomposition to obtain a low-dimensional model.

[0069] S2. Based on the obtained low-dimensional model, temperature sensors, heat flow sensors, and flow velocity sensors are placed at different locations in the solar thermal power generation system to collect data from the power generation system. Gaussian process regression and Kalman filter algorithms are used to fuse the collected data and reconstruct the temperature field of the power generation system.

[0070] S3. By reconstructing the temperature field information, a multi-objective optimization function is established, which comprehensively considers the objectives of temperature difference, cooling power, temperature field uniformity and thermal energy storage efficiency, dynamically adjusts the weights, and optimizes the objective priority in real time.

[0071] S4. Based on the established multi-objective optimization function, the improved particle swarm optimization algorithm is used to generate a sequence of control parameters. The control parameters are then iteratively optimized through particle swarm optimization, and the control decisions are adjusted based on real-time data and prediction results.

[0072] S5. Input the optimized control parameters into the multi-rate hierarchical control system, which is executed by the high-speed loop, medium-speed loop and low-speed loop, and the data is synchronized through the state buffer.

[0073] S6. By executing the control process in real time, the system status is monitored and the type of fault that occurs in the power generation system is detected. The system then switches to the corresponding fault-tolerant control mode according to the type of fault.

[0074] For step S1, in this embodiment, a coupled model of the electromagnetic field, temperature field, and flow field is first established based on the principles of electromagnetics, thermal conductivity, and fluid mechanics, combined with the specific design parameters of the high-voltage variable frequency power supply. Generally, these physical fields influence each other; changes in the electromagnetic field lead to changes in the temperature field distribution, while fluid flow affects heat conduction and dissipation. In high-voltage variable frequency power supply systems, special attention needs to be paid to overheat protection and safety to ensure that the system can respond promptly and take effective protective measures in the event of abnormal temperatures.

[0075] Specifically, the model building process includes the following aspects:

[0076] Electromagnetic field model establishment: High-voltage variable frequency power generation systems typically contain multiple power electronic devices, which generate strong electromagnetic fields during operation. To accurately simulate the performance of the power generation system, the influence of electromagnetic fields on power equipment must be considered, especially the effect of current changes on motor windings and heat sources. During this process, the system needs to monitor changes in the electromagnetic field in real time to prevent electromagnetic interference and thermal damage caused by overload. This part of the model is described using Maxwell's equations from electromagnetism:

[0077]

[0078] in, B represents the curl of the electric field E; E is the electric field strength; B is the magnetic field strength. Represents the gradient operator; This represents the partial derivative of the magnetic field B with respect to time; This indicates that the divergence of the magnetic field is zero.

[0079] Establishment of the temperature field model: During the operation of the variable frequency power supply, the heat exchange between electrical components and air or coolant determines the temperature field distribution. This temperature field is jointly determined by heat conduction and convection processes, and is typically modeled using a combination of heat conduction equations and fluid dynamics equations. Real-time monitoring of the temperature field is particularly important to prevent overheating-related malfunctions or safety risks. The temperature field distribution directly affects the effectiveness of thermal management strategies. This temperature field can be described by the following equation:

[0080]

[0081] Where ρ is density; c is specific heat capacity; k is thermal conductivity; T is temperature; and Q is the heat source term. This represents the rate of change of temperature over time. This represents the diffusion of heat in space. The distribution of the temperature field plays a crucial role in the thermal management of power generation systems.

[0082] Establishing the flow field model: The fluid dynamics section primarily considers the effect of coolant flow, as changes in the flow field directly impact heat conduction and dissipation. Fluid flow follows the Navier-Stokes equations, describing the fluid's motion. Here, the flow velocity and flow rate of the cooling fluid are emphasized to ensure timely heat removal and prevent overheating due to heat accumulation. The flow field model typically takes the following form:

[0083]

[0084] Where v is the fluid velocity; p is the fluid pressure; μ is the dynamic viscosity of the fluid; and F is the external force term (such as gravity). Indicates the instantaneous rate of change of flow velocity; The change in velocity of a particle due to its motion; This indicates the direction and magnitude of acceleration caused by uneven pressure distribution; This indicates the diffusion of momentum within a fluid caused by viscosity.

[0085] Time-varying characteristics and dynamic response: As voltage and load change, the electromagnetic characteristics, temperature distribution, and cooling effect of the power generation system will dynamically change. Therefore, when building the model, the time-varying characteristics of the system must be considered to ensure that the model can adapt to changes under different operating conditions. Especially under high load or low temperature environments, the system response needs to have high flexibility and accuracy to avoid failures caused by abnormal temperature or overload.

[0086] By establishing the physical model described above, the coupling effects between the electromagnetic field, temperature field, and flow field can be accurately simulated and solved. It is important to note that the electromagnetic characteristics, temperature distribution, and cooling effect of the power generation system will dynamically change with variations in voltage and load. Therefore, when establishing the model, it is also necessary to consider the time-varying characteristics of the system so that the model can adapt to changes under different operating conditions.

[0087] In some embodiments, the Orthogonal Eigenfactor Decomposition (POD) method is used to reduce the order of high-dimensional coupled equations. A snapshot matrix is ​​formed by capturing snapshots of the state variables of the coupled model over time.

[0088] X = [x1, x2, ..., xm];

[0089] Where X is the snapshot matrix; x i(where i = 1, 2, ..., m) represents the state vector of the system at the i-th time step; m is the number of snapshots; each x i This represents the state vector of the power generation system at a certain moment. Singular Value Decomposition (SVD) is then performed on this matrix:

[0090] X=U∑V T ;

[0091] Where U is the left singular vector matrix; ∑ is the singular value diagonal matrix; V T It is the transpose of the right singular vector matrix.

[0092] Extract the main feature U and select the first r modes to form a reduced basis:

[0093] Φ=[u1,u2,…,u r ];

[0094] Where Φ is the reduced-dimensional modal basis matrix; u i (where i = 1, 2, ..., n); r is the number of selected principal modes.

[0095] By projecting the high-dimensional state into a low-dimensional subspace, the state variables of the low-dimensional system are obtained, thus greatly reducing computational complexity. This reduced-order model not only improves the efficiency of numerical solutions but also ensures accurate thermal management of the high-voltage variable frequency power supply system under different operating conditions.

[0096] In this embodiment, the established electromagnetic field, temperature field and flow field coupled model provides basic data support for subsequent multi-objective optimization and thermal safety protection. Through the dynamic optimization and updating of this model, the system can monitor and reflect the thermal state of the power generation system in real time. When the temperature is abnormal or overloaded, it can respond in time and trigger the overheat protection mechanism, thereby ensuring that the system is always within the safe operating range.

[0097] In step S2, this embodiment utilizes Gaussian process regression (GPR) and Kalman filtering algorithms to process sensor data, thereby improving the reconstruction accuracy of the temperature and flow fields of the power generation system and providing more accurate real-time data support for subsequent thermal management control, particularly for overheat protection and fault diagnosis. The data fusion process integrates data collected from multiple sensors (such as temperature sensors and flow rate sensors), and further optimizes the prediction effects of the temperature and flow fields through mathematical modeling, providing data support for overheat warning and thermal protection response.

[0098] Data on temperature, flow rate, and other relevant parameters collected by sensors are often affected by noise, sensor errors, and system anomalies. Therefore, Gaussian process regression and Kalman filtering algorithms must be used to process the data to eliminate noise, compensate for missing data, and enhance the system's ability to respond to fault warnings.

[0099] Specifically, the combination of Gaussian process regression and Kalman filtering provides a powerful data fusion method:

[0100] Gaussian Process Regression (GPR): Gaussian process regression is a nonparametric regression method based on Bayesian theory, widely used to solve regression problems caused by noise, incomplete data, and other factors. It helps systems handle sensor errors and predict real-time trends in temperature and flow fields, ensuring timely system response to thermal runaway. The specific regression model can be represented by the following equation:

[0101]

[0102] Where y(x) is the predicted temperature value; K(x,X) is the kernel function matrix, representing the similarity between the test point and the training point; K(X,X) is the kernel function matrix among the training set itself; and I is the identity matrix. y is the noise variance; y is the temperature value of the training data; X is the training data point.

[0103] Kalman Filtering: Kalman filtering is a recursive optimization estimation method widely used in state estimation problems of dynamic systems. It can effectively adjust the temperature field estimate in real-time monitoring, optimizing fault warning and overheat protection functions. By combining historical estimates and new observation data and minimizing the error covariance, Kalman filtering enhances the system's dynamic prediction capability of temperature changes, ensuring the system can respond to temperature anomalies and faults in real time. The basic formula for Kalman filtering is as follows:

[0104]

[0105] in, It is the estimated value at time k. K is a predicted value based on previous data. k It is the Kalman gain, z k This is the new observation at time K, H k It is the observation matrix.

[0106] By combining the algorithmic advantages of GPR and Kalman filtering, this invention effectively utilizes the characteristics of data from various sensors, reduces noise interference, and improves the reconstruction accuracy of the temperature and flow fields. During real-time dynamic monitoring, the data fusion method provides precise input data for subsequent control strategies, enabling the system to intelligently adjust according to different operating conditions, ensuring a balanced temperature field distribution and preventing system failure due to overheating.

[0107] Especially in fault detection and fault-tolerant control, data fusion technology can effectively monitor anomalies in sensor data and provide early warnings of overheating faults. For example, when some sensors malfunction or lose data, the system can reconstruct an accurate prediction of the temperature field through data fusion and trigger fault-tolerant control through redundancy mechanisms to ensure stable system operation. Furthermore, after data fusion, the system can accurately predict the internal temperature change trend of the power supply and adjust the operating status of the cooling system in a timely manner based on the prediction results to avoid thermal runaway.

[0108] In step S3, this embodiment constructs a multi-objective optimization function to comprehensively optimize various thermal management indicators of the system, particularly considering overheat protection and fault response requirements. Specifically, the optimization function not only focuses on a single objective but also considers three important parameters simultaneously: temperature difference, cooling power, and temperature field uniformity. By dynamically adjusting the weight coefficients of each objective, the optimization process can automatically adjust the priority of the optimization objectives according to the needs under different operating conditions, thereby achieving optimal thermal management performance under different environmental and load conditions and ensuring that the system can respond promptly and take fault protection measures when temperature anomalies occur.

[0109] In some embodiments, the multi-objective optimization function can be represented by the following mathematical model:

[0110] J(t) = α·max(ΔT) + β·∫P cool dt+γ·σ T ;

[0111] Where ΔT is the maximum temperature difference within the system; P cool σ is the system cooling power; T Let α, β, and γ represent the uniformity of the temperature field; α, β, and γ are dynamically adjusted weighting coefficients; dt is the time step; and J(t) is the value of the multi-objective optimization function. The weighting coefficients α, β, and γ for each objective are dynamically adjusted and updated in real time according to actual working conditions (such as load changes and ambient temperature), thereby ensuring that the optimization process can meet the thermal management requirements under different operating conditions.

[0112] In another embodiment, the adjustment of the weighting coefficients can be driven by a Lyapunov-guided strategy to ensure that the system can respond promptly to temperature anomalies or load changes. In this case, the variation of the weighting coefficients α, β, and γ is described by the following equation:

[0113]

[0114] Where α, β, and γ represent three dynamic weighting factors in multi-objective optimization; This represents the rate at which these three weights change over time; k1, k2, k3 > 0 are adjustment factors; These equations represent the partial derivatives of the objective function with respect to each sub-objective. Through these equations, the optimization process can automatically adjust the weights of each optimization objective based on the real-time state of the system, thereby achieving more precise temperature control and improved energy efficiency.

[0115] In multi-objective optimization, temperature difference is the first consideration, as it is a key factor in determining whether local overheating exists in the power generation system. In high-voltage variable frequency power generation systems, temperature difference typically represents uneven temperature distribution in certain critical components (such as between power electronic devices and heat dissipation devices). Excessive temperature difference can lead to equipment overheating or malfunction; therefore, it is necessary to minimize the temperature difference to maintain system stability and safety.

[0116] Secondly, cooling power is a key factor affecting system energy efficiency. While excessive cooling power can keep equipment operating at a safe temperature, excessive cooling consumption will reduce the overall system energy efficiency. Therefore, optimization should not only control the temperature difference but also balance cooling power consumption to avoid unnecessary energy waste and ensure that the cooling system can be quickly adjusted in the event of abnormal temperatures.

[0117] Temperature field uniformity refers to the degree of uniformity of temperature distribution within a system. A uniform temperature field can effectively prevent localized overheating, thereby extending the service life of equipment and avoiding equipment damage due to uneven temperature. Temperature field uniformity is another key optimization objective; by optimizing this parameter, the stability and reliability of the system can be further improved.

[0118] In practical applications, temperature difference, cooling power, and temperature field uniformity have certain interrelationships. For example, reducing the temperature difference may require increasing cooling power; conversely, achieving temperature field uniformity may necessitate adjustments to the cooling system to optimize temperature distribution. To address this issue, this invention dynamically adjusts weighting coefficients to ensure priority changes under different operating environments, thereby achieving the dual goals of precise temperature control and improved energy efficiency.

[0119] Meanwhile, after each optimization calculation, the system automatically adjusts the cooling system's operating status based on real-time temperature data and load changes. When the temperature reaches the set safety threshold, the system activates an overheat protection mechanism, such as increasing cooling power or adjusting the workload, to ensure the system remains in a safe state and prevent equipment damage due to overheating.

[0120] In step S4, in this embodiment, the Particle Swarm Optimization (PSO) algorithm is used to generate a sequence of control parameters, thereby optimizing the thermal management and overheat protection performance of the high-voltage variable frequency power supply system. PSO is a global optimization algorithm that simulates the foraging behavior of bird flocks. It can find the optimal solution within a given search space and is particularly suitable for handling high-dimensional complex problems, such as the multi-objective thermal management problem in high-voltage variable frequency power generation systems.

[0121] The core of particle swarm optimization (PSO) is to find the optimal control parameters for thermal management objectives (such as temperature difference, cooling power, and temperature field uniformity) by simulating the motion and mutual learning of a swarm of particles. PSO not only focuses on energy efficiency but also considers the system's safety under abnormal operating conditions, particularly its role in overheat protection and fault-tolerant control. Through iterative optimization, PSO can provide the system with optimal control parameters, ensuring efficient operation under different loads and environmental conditions, and immediately activating fault protection mechanisms in the event of temperature anomalies.

[0122] In particle swarm optimization (PSO), the initial positions and velocities of the particles are first set, and each particle corresponds to a set of control parameters. These control parameters directly affect the system's temperature management performance, including the cooling system's fan speed and coolant flow rate. Through iteration, the particles adjust their positions based on their current fitness values ​​to find the optimal control parameters. The particle velocity update equation is:

[0123]

[0124] in, This represents the velocity update value of particle i in the (k+1)th generation; ω is the inertial weight; c1, c2, and c3 are constant acceleration factors; pbest i This represents the best position reached by particle i during the historical search. Let r1 be the current position of particle i in the kth generation; gbest is the globally optimal position found by all particles in the current generation; r1, r2, and r3 are the coefficients in the particle velocity update equation. The gradient term of the cost function is ω; the inertia weight ω is adjusted according to the convergence requirements of the algorithm.

[0125] The fitness function of a particle is typically given by a multi-objective optimization function, which comprehensively considers multiple objectives such as temperature difference, cooling power, and temperature field uniformity. The fitness function can be expressed as:

[0126] J(t) = α·max(ΔT) + β·∫P cool dt+γ·σ T ;

[0127] Where ΔT is the maximum temperature difference within the system; P cool σ is the system cooling power;T Let represent the uniformity of the temperature field; α, β, and γ be dynamically adjusted weighting coefficients; dt be the time step; and J(t) be the multi-objective optimization function value. In this way, the particle swarm optimization algorithm can simultaneously optimize multiple thermal management objectives and adjust the priority of these objectives based on the current operating state of the system.

[0128] In each particle swarm optimization iteration, particles adjust control parameters based on their updated velocity and position. These control parameters include specific operating parameters of the cooling system, such as fan speed and pump flow rate. These parameters directly affect the temperature field distribution, thereby achieving the optimization goal of reducing temperature differences, optimizing cooling power, and improving the uniformity of the temperature field. Especially under high load or extreme environmental conditions, the particle swarm optimization algorithm can dynamically adjust cooling parameters to meet temperature control requirements while preventing overheating and avoiding thermal runaway.

[0129] Furthermore, during the optimization process, the system not only considers minimizing the temperature difference but also dynamically adjusts the weighting coefficients using a particle swarm optimization algorithm. This ensures that the priority of the optimization objective can switch in real time under load changes or environmental fluctuations. For example, optimizing the temperature difference may have a higher priority under high load, while optimizing cooling power may become the priority under low load. Through such adjustments, the system can avoid unnecessary energy waste, ensure temperature safety, and adjust the operating status of the cooling system in a timely manner to prevent overheating.

[0130] The power of particle swarm optimization lies in its ability to continuously adjust control parameters based on real-time feedback, ensuring the system remains in optimal operating condition even under complex and dynamically changing circumstances. For example, under high load conditions, the system may need to increase cooling power, while under low load conditions, it can reduce cooling power to save energy. This flexible adjustment method ensures efficient thermal management and long-term system stability.

[0131] By employing particle swarm optimization (PSO) algorithms, the system can dynamically adjust cooling parameters in real time while ensuring that the overheat protection mechanism remains active, preventing equipment damage caused by abnormal temperatures. Furthermore, PSO algorithms can optimize the energy efficiency of the cooling system, avoiding energy waste due to inappropriate cooling power settings.

[0132] During the optimization process, the particle swarm optimization algorithm not only focuses on temperature control but also involves fault detection and fault-tolerant control of the system. When a fault is detected in the system (such as a temperature sensor malfunction or abnormal coolant flow), the optimization algorithm can automatically adjust control parameters, for example, by switching to an alternative cooling path or adjusting the cooling power, to ensure that the equipment will not overheat and be damaged due to a single point of failure, thereby improving the fault tolerance and reliability of the system.

[0133] In step S5, in this embodiment, the optimized control parameters are input to the multi-rate hierarchical control system to ensure accurate execution in each different stage. The multi-rate hierarchical control system mainly includes three stages: a high-speed loop, a medium-speed loop, and a low-speed loop, each responsible for different tasks.

[0134] The high-speed loop undertakes the high-frequency control tasks of the system, primarily focusing on real-time regulation of modules extremely sensitive to temperature changes (such as power converters, power electronic switches, and heat sinks). Based on the latest control parameters output by the Particle Swarm Optimization (PSO) algorithm, the high-speed loop rapidly executes control commands for critical components in the cooling system, such as fan speed and coolant pump speed, ensuring millisecond-level response in the event of overheating. Specifically, the control function of the high-speed loop can be expressed as:

[0135] u high =f high (P opt ,T real ,Δt);

[0136] Among them, u high For high-speed loop control commands; f high P is the high-speed loop control function; opt The control parameters obtained from particle swarm optimization; T real Δt represents the current actual temperature of the system; Δt represents the control execution cycle.

[0137] The intermediate-speed loop is primarily responsible for the dynamic prediction of the thermal field model of the power generation system, forecasting changes in the temperature field based on real-time data. In this stage, combining feedback data from temperature and heat flux sensors, the system uses algorithms such as Model Predictive Control (MPC) to simulate and predict temperature changes in the power generation system. Specifically, the control function of the intermediate-speed loop is:

[0138] u mid =f mid (T pred ,T real ,Δt);

[0139] Among them, u mid For the control command of the medium-speed loop; f high T is the high-speed loop control function; pred For the predicted temperature field; T real The current actual temperature is represented by Δt, which is the control period. The intermediate-speed loop adjusts the control strategy based on real-time temperature field predictions to reduce temperature deviation.

[0140] The low-speed loop primarily addresses long-term temperature control management and equipment-level control execution at the physical level, including overall start-stop scheduling of the cooling system and updates to energy efficiency optimization strategies for power-level equipment. The control function of the low-speed loop can be expressed as:

[0141] u low =f low (P cool ,T real ,Δt);

[0142] Among them, u low For low-speed loop control commands; f high P is the high-speed loop control function; cool This refers to the cooling power or the status of the cooling equipment; T real Δt represents the current actual temperature; Δt represents the control period.

[0143] To ensure consistency in timing, commands, and states across control loops, the system introduces a state cache synchronization mechanism to guarantee data consistency and coordination across loops.

[0144] Each control loop has a local status buffer that periodically uploads its own control results and status indicators.

[0145] The buffer contains information such as current temperature, cooling status, warning status, and control output.

[0146] Before performing prediction and equipment management tasks, the medium-speed and low-speed loops prioritize reading the most recent state of the high-speed loop from the cache.

[0147] In the event of command conflicts or rapid changes in trends (such as a sudden temperature rise), the medium-speed ring can temporarily relinquish some decision-making authority to the high-speed ring to ensure that the priority of real-time overheating emergency response is maximized.

[0148] For step S6, in this embodiment, the monitoring system continuously assesses the health status of the system through real-time status monitoring and residual detection mechanisms to accurately determine whether there are sensor or actuator faults. Once a fault is detected, the system immediately triggers the corresponding fault-tolerant control mode to ensure the temperature safety and stable operation of the power generation system and prevent overheating risks caused by equipment failure.

[0149] When a sensor in the system fails, traditional sensor data becomes unreliable, necessitating an alternative strategy. In some embodiments, this invention replaces the function of the damaged sensor with model predictive control (MPC). Specifically, the system predicts the temperature or heat flux during the sensor failure period based on existing temperature and flow field models and known control parameters. This predicted value is used to adjust the operation of the cooling equipment during control to ensure that the system's thermal management is unaffected. The general form of model predictive control is as follows:

[0150]

[0151] Among them, u MPC For model predictive control output; This indicates that MPC selects a set of control strategies within the rolling prediction interval to minimize the cost function; T pred (k) represents the predicted temperature value; T ref The desired temperature value; T pred (k)-T ref 2 The squared loss term is the temperature error; λ is the weighting coefficient; u(k) is the control input; u previous The previous control input is denoted by ; N is the prediction time step. Using this formula, the system can adjust the cooling control process based on the predicted temperature and known parameters in the event of a sensor failure.

[0152] When a fault is detected in an actuator (such as a cooling fan or electric pump), the system automatically switches to safety mode. In safety mode, the system strictly limits the adjustment range of control inputs to prevent abnormal actuator movements caused by the fault from leading to overheating or excessive cooling of the equipment, ensuring that the equipment operates within a safe temperature range. For example, the maximum limit of the control input can be expressed by the following formula:

[0153] u safe =min(max(u) opt ,u min ),u max );

[0154] Among them, u safe For control input in safe mode; u opt For the optimized control parameters; u min and u max These are the minimum and maximum control input limits in safe mode, respectively. In this way, the system can ensure that it will not exceed the safe operating range even in the event of actuator failure, thus ensuring the safety and stability of the power generation system.

[0155] The system identifies faults by monitoring the operational status of each component in real time and incorporating a residual detection mechanism. When the output of a sensor or actuator deviates significantly from the predicted value, the residual detection mechanism issues an alarm and determines if a fault exists. If the system confirms a fault, it automatically switches to the appropriate fault-tolerant control mode based on the fault type.

[0156] Specifically, the basic formula for residual detection in the system is as follows:

[0157]

[0158] Where r(k) is the residual, and y(k) is the actual output of the sensor or actuator. The system predicts the output. If the residual r(k) exceeds a set threshold, the system will identify a fault and activate the corresponding fault-tolerant mechanism. For sensor faults, the system will activate model predictive control; for actuator faults, the system will enter a safe mode and restrict control input.

[0159] The energy-efficient high-voltage variable frequency power supply thermal management control system described below can be referred to in correspondence with the energy-efficient high-voltage variable frequency power supply thermal management control method described above.

[0160] Please see the appendix Figure 2 The present invention also provides a high-voltage variable frequency power supply thermal management control system for improving energy efficiency, comprising:

[0161] The multiphysics modeling and order reduction module is used to establish a coupled model of electromagnetic field, temperature field and flow field based on the principles of electromagnetics, thermal conductivity and fluid mechanics, combined with the system structure and operating conditions, and to reduce the order of the model based on the intrinsic orthogonal decomposition method.

[0162] The sensor acquisition and temperature rise detection module is used to deploy temperature sensors, heat flow sensors and flow velocity sensors in key components and overheat-sensitive areas of the solar thermal power generation system to collect multi-point operating data and analyze it by fusion through Gaussian process regression and Kalman filter algorithm.

[0163] The thermal safety assessment and protection strategy module is used to combine real-time temperature field reconstruction results with historical thermal runaway characteristic curves to determine whether the preset temperature protection threshold has been reached, construct a multi-objective protection function, and form a dynamic safety control strategy.

[0164] The control parameter optimization generation module is used to generate control parameter sequences based on thermal safety control strategies using an improved particle swarm optimization algorithm, and introduces temperature rise weighting factors and gradient feedback to improve control accuracy.

[0165] The multi-rate thermal control execution module is used to receive optimized control parameters and execute them in layers according to the control strategies of high-speed loop, medium-speed loop and low-speed loop, while using the state buffer to realize cross-rate control data synchronization.

[0166] The fault identification and redundancy fault tolerance module is used to identify and locate sensor faults, actuator failures, or temperature anomalies that may occur during system operation, and switch to model prediction fault tolerance control mode or enable redundant equipment according to the fault type.

[0167] The system in this embodiment can be used to execute the above method embodiments, and its principle and technical effect are similar, so they will not be described again here.

[0168] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A thermal management control method for high-voltage variable frequency power supplies to improve energy efficiency, characterized in that, Includes the following steps: Based on the principles of electromagnetism, thermal conductivity, and fluid mechanics, and combined with the structure and operating environment of the solar high-temperature thermal power generation system, a coupled model of electromagnetic field, temperature field, and flow field is established. The model is then reduced in order through intrinsic orthogonal decomposition to obtain a low-dimensional model. Based on the obtained low-dimensional model, temperature sensors, heat flow sensors, and flow velocity sensors are placed at different locations in the solar thermal power generation system to collect data from the power generation system. Gaussian process regression and Kalman filter algorithms are used to fuse the collected data and reconstruct the temperature field of the power generation system. By reconstructing the temperature field information, a multi-objective optimization function is established, which comprehensively considers the objectives of temperature difference, cooling power, temperature field uniformity and thermal energy storage efficiency, dynamically adjusts the weights, and optimizes the objective priority in real time. Based on the established multi-objective optimization function, an improved particle swarm optimization algorithm is used to generate a sequence of control parameters. The control parameters are then iteratively optimized through particle swarm optimization, and control decisions are adjusted based on real-time data and prediction results. The optimized control parameters are input into the multi-rate hierarchical control system, which is executed by high-speed loop, medium-speed loop and low-speed loop respectively, and the data is synchronized by the state buffer. By executing the control process in real time, the system status is monitored and the type of fault in the power generation system is detected. The system then switches to the corresponding fault-tolerant control mode based on the type of fault. The establishment of the coupled electromagnetic field, temperature field, and flow field model includes the following steps: By combining the photothermal conversion process of solar high-temperature thermal power generation system with the principles of heat conduction and fluid mechanics, the mutual coupling between light radiation, temperature field and cooling flow field is first considered in the physical model. Partial differential equations are used to describe the coupling relationship between light radiation, temperature and fluid field, and the transient response of each physical field is calculated. The focus is on the heat conduction and heat recovery processes of solar collectors, thermal energy storage systems and heat exchangers. The high-dimensional coupled equations are reduced in order by using the intrinsic orthogonal decomposition method to reduce the amount of computation and obtain a low-dimensional photothermal conversion-temperature field-flow field coupled model. The establishment of the multi-objective optimization function includes the following steps: Based on the reconstructed temperature field information, a multi-objective optimization function is established to simultaneously optimize temperature difference, cooling power, temperature field uniformity, and thermal energy storage efficiency. The multi-objective optimization function is expressed as follows: ； in, This represents the maximum temperature difference within the system. For system cooling power, For the uniformity of the temperature field, The weighting coefficients are dynamically adjusted. For time step, Optimize function values ​​for multiple objectives; The weighting coefficients are dynamically adjusted based on light intensity, ambient temperature, and system load conditions.

2. The high-voltage variable frequency power supply thermal management control method for improving energy efficiency according to claim 1, characterized in that, The process of fusing the collected data using Gaussian process regression and Kalman filtering algorithms includes the following steps: After obtaining the low-dimensional model, temperature sensors, heat flow sensors, and flow velocity sensors are placed at different locations in the solar high-temperature thermal power generation system to collect system status information in real time, with a focus on the status of the collector, heat exchanger, and thermal energy storage system. By using the Gaussian process regression algorithm to spatially interpolate the sensor data, the accurate values ​​of the temperature field and flow field are inferred, ensuring the accurate reconstruction of the collector surface temperature and heat exchange efficiency. By combining the Kalman filter algorithm, the sensor data is dynamically updated over time to reconstruct the temperature field of the solar power generation system and adapt to changes in different weather and light intensity.

3. The high-voltage variable frequency power supply thermal management control method for improving energy efficiency according to claim 1, characterized in that, The process of generating the control parameter sequence using the improved particle swarm optimization algorithm includes the following steps: In the particle swarm optimization algorithm, the initial position and velocity of each particle in the swarm are first set, and the fitness of the current particle is calculated. The fitness function is defined according to the temperature field, cooling power and thermal energy storage target. In each iteration, the particle's velocity and position are updated, where the particle velocity update equation is: ； in, Represents particles In the The speed update value of the generation, For inertial weights, , , The constant acceleration factor For particles The best location reached in the history search For particles In the The current position of the generation This represents the globally optimal position found for all particles in the current generation. 、 、 These are the coefficients in the particle velocity update equation. The gradient term of the cost function, and the inertia weights. Adjustments should be made based on the algorithm's convergence requirements; Through multiple particle swarm iterations, a set of optimized control parameters is finally obtained to adapt to the adjustment of solar collectors and heat exchangers under different lighting conditions.

4. The high-voltage variable frequency power supply thermal management control method for improving energy efficiency according to claim 1, characterized in that, The process of inputting the optimized control parameters into the multi-rate hierarchical control system includes the following steps: The optimized control parameters are input into a multi-rate hierarchical control system, which includes a high-speed loop, a medium-speed loop, and a low-speed loop. The high-speed loop performs real-time calculations of control parameters for particle swarm optimization, and adjusts the operation of the solar collector and cooling system. The medium-speed loop dynamically predicts the thermal field model of the power generation system and adjusts the heat release of the thermal energy storage system. The low-speed loop is responsible for executing physical drive and cooling operations. By synchronizing data across loops through the state buffer, the control operations of each loop are coordinated and consistent.

5. The high-voltage variable frequency power supply thermal management control method for improving energy efficiency according to claim 1, characterized in that, The step of switching to the appropriate fault-tolerant control mode based on the fault type includes the following steps: By monitoring the real-time status of the system, a residual detection mechanism is used to determine whether a fault exists. If the system detects a fault type, a fault-tolerant control mode is triggered. In the event of sensor failure, model predictive control is used to replace the function of the damaged sensor, and the thermal management process is controlled by the predicted value. In the event of an actuator failure, the system switches to a safe mode and controls system operation by setting limits.

6. The high-voltage variable frequency power supply thermal management control method for improving energy efficiency according to claim 1, characterized in that, The process of reducing the order of the high-dimensional coupled equation system obtained by the intrinsic orthogonal decomposition method includes: Collect snapshot matrix of state variables of the coupled model over time. ,in This represents a snapshot matrix of state variables collected over time, each... This indicates the state of the power generation system at a certain moment. Indicates the first The state of the power generation system at any given moment. This represents the dimension of the state variables in the power generation system; Perform singular value decomposition on the snapshot matrix: ； in, It is a left singular vector matrix. It is a singular value diagonal matrix. It is the transpose of the right singular vector matrix; Before selection The main modes form a dimensionality reduction base. Projecting the state of a high-dimensional system onto a low-dimensional subspace: ； in, Indicates the number of principal components. This represents the most representative basis vector extracted from the original state data. For low-dimensional state variables, The POD mode matrix, This indicates that the POD method was used.

7. The high-voltage variable frequency power supply thermal management control method for improving energy efficiency according to claim 1, characterized in that, The dynamic adjustment of the weight coefficients is driven by a Lyapunov-guided strategy, and the adjustment formula is as follows: ； in, , , These represent the three dynamic weighting factors in multi-objective optimization. 、 、 This represents the rate at which these three weights change over time. As a regulating factor, , , This represents the partial derivative of the objective function with respect to each sub-objective.

8. A high-voltage variable frequency power supply thermal management control system for improving energy efficiency, applied to the high-voltage variable frequency power supply thermal management control method for improving energy efficiency as described in any one of claims 1-7, characterized in that, include: The multiphysics modeling and order reduction module is used to establish a coupled model of electromagnetic field, temperature field and flow field based on the principles of electromagnetics, thermal conductivity and fluid mechanics, combined with the system structure and operating conditions, and to reduce the order of the model based on the intrinsic orthogonal decomposition method. The sensor acquisition and temperature rise detection module is used to deploy temperature sensors, heat flow sensors and flow velocity sensors in key components and overheat-sensitive areas of the solar thermal power generation system to collect multi-point operating data and analyze it by fusion through Gaussian process regression and Kalman filter algorithm. The thermal safety assessment and protection strategy module is used to combine real-time temperature field reconstruction results with historical thermal runaway characteristic curves to determine whether the preset temperature protection threshold has been reached, construct a multi-objective protection function, and form a dynamic safety control strategy. The control parameter optimization generation module is used to generate control parameter sequences based on thermal safety control strategies using an improved particle swarm optimization algorithm, and introduces temperature rise weighting factors and gradient feedback to improve control accuracy. The multi-rate thermal control execution module is used to receive optimized control parameters and execute them in layers according to the control strategies of high-speed loop, medium-speed loop and low-speed loop, while using the state buffer to realize cross-rate control data synchronization. The fault identification and redundancy fault tolerance module is used to identify and locate sensor faults, actuator failures, or temperature anomalies that may occur during system operation, and switch to model prediction fault tolerance control mode or enable redundant equipment according to the fault type.

Images