A method and system for adaptive model predictive control of building HVAC system

Abstract

The present invention proposes an adaptive model predictive control method and system for a building HVAC system, which belongs to the technical field of intelligent building adaptive control, and includes: obtaining operation data of a HVAC system under different working conditions and preprocessing the obtained data; calculating the marginal fitness metric of each data point for each class for the preprocessed data, calculating the fitness metric of the data point for each class by combining the marginal fitness metrics of all descriptors, and obtaining a global fitness metric; setting a global fitness metric threshold, and classifying the current operating state of the system according to a classification rule; after classification, the model parameters of each class are fixed, so as to adopt an adaptive control strategy for different operating states.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent building adaptive control, and in particular relates to an adaptive model predictive control method and system for a building heating, ventilation and air conditioning system. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] With the continuous increase of global energy consumption and the increasing environmental problems, building energy conservation has become one of the important challenges facing today's society. In modern building management, the energy consumption of HVAC systems accounts for a large part of the total energy consumption of buildings. With the increase of energy costs and the improvement of environmental protection awareness, improving the energy efficiency of HVAC systems has become a key issue.

[0004] Since traditional HVAC control methods are usually based on fixed control strategies, such as model predictive control (MPC), these strategies will be difficult to adapt to the complex working conditions of HVAC systems, and with the continuous changes in system parameters and the surrounding environment, the system will be inefficient and waste energy. At the same time, in nonlinear systems, traditional MPC is easily affected by noise and measurement errors during state estimation, resulting in poor control effects. When faced with rapidly changing operating conditions, the optimization of control strategies may lag behind, making it difficult to make the best decisions in complex and changing environments. When dealing with uncertainty and external interference, the system has low robustness and reliability, and may also lead to high energy consumption due to lack of precision. Summary of the invention

[0005] In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides an adaptive model predictive control method for a building HVAC system to ensure the efficient operation of the HVAC system under different operating conditions, which can not only improve the energy efficiency and operation stability of the system, but also ensure the comfort of the indoor environment.

[0006] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0007] In a first aspect, a method for adaptive model predictive control of a building HVAC system is disclosed, comprising:

[0008] Obtain the operating data of the HVAC system under different working conditions and pre-process the acquired data;

[0009] Calculate the marginal fitness measure of each data point for each class for the preprocessed data, and calculate the fitness measure of the data point for each class by combining the marginal fitness measures of all descriptors to obtain the global fitness measure;

[0010] Set the global fitness measurement threshold and classify the current operating status of the system according to the classification rules;

[0011] After classification, the model parameters of each class are fixed to adopt adaptive control strategies for different operating states;

[0012] In each control cycle, based on the classification results, the state estimate updated by the Kalman filter is used as the current state input of the adaptive model. Based on the current state and the prediction model, the optimal control input sequence in the future is solved to adjust the operating parameters of the HVAC system.

[0013] As a further technical solution, the model parameters of each category include a state transfer matrix, a control input matrix, a control output matrix, a process noise covariance matrix and a measurement noise covariance matrix.

[0014] As a further technical solution, it also includes: updating the center of each category according to the weighted average of the membership of the data points, that is, the mean value in the membership function.

[0015] As a further technical solution, the state estimation process for the HVAC system includes:

[0016] Construct a state-space model of the HVAC system;

[0017] Linearize the state transition equations in the state space model;

[0018] Linearize the output equations in the state-space model;

[0019] At runtime, the preset parameters of the linearized state space model are linearized and adjusted in real time according to the current state and control input;

[0020] The adjusted parameters are used for the prediction and update steps of the Kalman filter to provide accurate state estimation and dynamically update the model parameters so that the prediction model can accurately reflect the dynamic behavior of the current system.

[0021] As a further technical solution, the state transfer equation in the state space model is linearized. The specific process includes:

[0022] Perform first-order Taylor expansion on the nonlinear state transfer equations in the state-space model at the current state estimate and control input;

[0023] The Jacobian matrix is ​​calculated using the current state and control input, the parameters of the preset classification are modified, and the partial derivative of the state transfer function with respect to the control input is obtained to obtain the Jacobian matrix;

[0024] Finally, the linearized state transfer equation is obtained.

[0025] As a further technical solution, the output equation in the state space model is linearized, specifically including:

[0026] Perform a first-order Taylor expansion on the nonlinear output equation at the current state estimate to obtain the linearized output equation.

[0027] As a further technical solution, when the state estimate after Kalman filtering is input into MPC, the state estimate provided by the Kalman filter is used As the current initial state, we get the cost function of MPC;

[0028] The constraints of the optimization problem, including state constraints and control input constraints;

[0029] By minimizing the cost function, the room temperature is made as close to the target temperature as possible to ensure comfort, and the cooling power and fan speed are controlled to meet the constraints and solve the optimal control input sequence.

[0030] In a second aspect, an adaptive model predictive control system for a building HVAC system is disclosed, comprising:

[0031] The operation data acquisition module is configured to: acquire the operation data of the HVAC system under different working conditions and pre-process the acquired data;

[0032] A global fitness metric calculation module is configured to: calculate the marginal fitness metric of each data point for each class for the preprocessed data, calculate the fitness metric of the data point for each class by combining the marginal fitness metrics of all descriptors, and obtain a global fitness metric;

[0033] The classification module is configured to: set a global fitness measurement threshold and classify the current operation state of the system according to the classification rule;

[0034] The control module,is configured as follows: after classification, the model parameters of each,category are fixed to adopt adaptive control strategies for different,operating states;

[0035] In each control cycle, based on the classification results, the state estimate updated by the Kalman filter is used as the current state input of the adaptive model. Based on the current state and the prediction model, the optimal control input sequence in the future is solved to adjust the operating parameters of the HVAC system.

[0036] One or more of the above technical solutions have the following beneficial effects:

[0037] The technical solution of the present invention can adapt to environmental changes: combining the state recognition of the LAMDA algorithm, the state estimation of the Kalman filter and the optimization control of the MPC, the system can provide real-time decision support in a complex and changing environment. Through real-time data collection and analysis, the present invention can dynamically adjust the operating parameters of the HVAC system, so that the system can meet the comfort requirements while minimizing energy consumption. This helps to significantly reduce the overall energy consumption of the building and improve energy efficiency.

[0038] The technical solution of the present invention can solve the interference of nonlinear systems: the sub-technical solution of this embodiment can optimize the control strategy in real time by combining the LAMDA method, Kalman filtering and model predictive control. The LAMDA method provides accurate state recognition and classification information, making the prediction model more accurate, thereby improving the control accuracy; the Kalman filter algorithm provides the best estimate of the system state by combining the system model and real-time measurement data, filtering out the measurement noise in the non-linear system, and improving the accuracy of state estimation.

[0039] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0041] Figure 1 Design a block diagram for an embodiment of the present invention;

[0042] Figure 2 The figure is a flow chart of a method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0043] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0044] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.

[0045] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0046] In order to overcome the limitations of traditional control methods and enhance the adaptability and intelligence level of the system, this embodiment proposes an energy-saving control method for HVAC system using adaptive model predictive control (MPC) combining LAMDA method and Kalman filtering.

[0047] The sub-technical scheme of this embodiment introduces the LAMDA method and the extended Kalman filter method. First, based on the LAMDA algorithm for real-time classification, combined with the classification results of the LAMDA algorithm, the system can effectively reduce unnecessary energy consumption and improve energy efficiency. Secondly, the Kalman filter provides the best estimate of the system state by combining the system model and real-time measurement data, effectively filters out measurement noise and process noise, and improves the accuracy of state estimation. Combined with the state estimation of the Kalman filter and the real-time classification of the LAMDA, the system can provide real-time decision support in a complex and changeable environment, and the MPC can perform dynamic optimization control based on the latest system state and operating environment to ensure that the control strategy is always in the best state, and the Kalman filter can enhance the system's ability to handle noise and uncertainty, while the real-time state recognition of the LAMDA algorithm improves the system's adaptability to environmental changes, and can effectively handle nonlinear complex systems, so that the robustness and reliability of the system are greatly improved.

[0048] The sub-technical solution of this embodiment combines the LAMDA (Learning Algorithm for Multivariable Data Analysis) method and the adaptive model predictive control (MPC) of the extended Kalman filter method for energy-saving control of HVAC (Heating, Ventilation, and Air Conditioning) systems.

[0049] By introducing the LAMDA method, the system can identify and classify operating states in real time, dynamically adjust control strategies, and continuously update model parameters and optimize control rules through the extended Kalman filter method, thereby achieving efficient energy management and stable system operation.

[0050] Specific solutions: data preprocessing and classification, state estimation, model adjustment, and dynamic optimization control strategies to ensure the efficient operation of the HVAC system under different operating conditions, which can not only improve the energy efficiency and operational stability of the system, but also ensure the comfort of the indoor environment.

[0051] Name explanation:

[0052] LAMDA: Learning Algorithm for Multivariable Data Analysis, Chinese: Learning Algorithm for Multivariable Data Analysis;

[0053] MPC: A control method: adaptive model predictive control;

[0054] HVAC: Heating, Ventilation, and Air Conditioning, Chinese: Heating, Ventilation, and Air Conditioning.

[0055] Embodiment 1

[0056] See attached Figure 1 As shown, this embodiment discloses an adaptive model predictive control method for a building HVAC system, which optimizes system performance by real-time monitoring and adjusting control strategies to cope with environmental changes and changes in system dynamic characteristics. Adaptive adjustment can be achieved by combining the LAMDA method and model predictive control (MPC), where y(t) is specifically the control action of the HVAC system, such as the start and stop of the compressor, the adjustment of the damper, etc. The prediction model is specifically to establish a state matrix for a period of time in the future, that is, a prediction model, which predicts the future state based on the current state and control input.

[0057] The specific steps are as follows:

[0058] Step 1: Data update;

[0059] During the operation of the building HVAC system, the model parameters are continuously updated through real-time data accumulation and online learning algorithms. According to the new data, the parameters of LAMDA's marginal fitness metric (MAD) and global fitness metric (GAD) are adjusted to enable the model to adapt to environmental and state changes. See Appendix Figure 2 As shown, the specific steps are as follows:

[0060] Step 1-1): Data preprocessing and classification:

[0061] By classifying the operating conditions of the HVAC system through the LAMDA method, different model parameters can be selected according to the system state, so as to adopt adaptive control strategies for different operating states. The selection of classification nodes should be based on system characteristics and control objectives. The model parameters of each category are usually fixed after classification to achieve the best control effect under specific conditions.

[0062] Step 1-1-1): Data collection:

[0063] Install sensors at key locations in the HVAC system to collect real-time data; for example: temperature sensor: indoors installed in the center of the room, outdoor installed away from the building and any equipment that may generate heat; humidity sensor: indoors installed in the center of the room, outdoor humidity sensor and outdoor temperature sensor can be installed together.

[0064] In this implementation example, a data acquisition system is established to collect operating data to monitor and record the above parameters in real time.

[0065] The above-mentioned data acquisition system includes a data acquisition sensor, a signal conditioning, a processor, a memory and a communication interface for processing the data collected by the sensor. This part can be constructed by technical personnel in this field according to actual needs and belongs to the existing technical means.

[0066] By changing the operating conditions of the HVAC system, such as load, ambient temperature, humidity, etc., different working conditions are simulated, and the operating data (temperature data, humidity data) of the HVAC system under different working conditions is collected using the data acquisition system; the collected data (data point is a single, specific measurement value, which is the raw data directly obtained from the sensor or data source. For example, a data point can be a temperature reading or humidity reading at a certain moment) is securely stored (standardized storage data) in a database or other storage medium for subsequent analysis.

[0067] Each data point can be represented as a vector:

[0068]

[0069] in, Indicates Moment The sensor readings, , T represents the upper limit of the time interval.

[0070] Step 1-1-2): Standardize data: Obtain real-time data simulating different working conditions from various sensors of the HVAC system, and pre-process the collected data, including denoising, outlier processing, data cleaning, etc.

[0071] By cleaning the data, noise and outliers are removed.

[0072] Each data point is standardized to scale the data to a uniform range for easy processing and comparison; the features of the object are normalized to [0,1] to improve performance through the following formula:

[0073]

[0074] in, Features The minimum value of Features The maximum value of Yes Category The The normalized value of the feature vector of the sample points.

[0075] Step 1-1-3): Marginal fitness measure (MAD): By calculating the fitness of each descriptor in each class, the descriptor is a structure used to describe the attributes and characteristics of the data. It can provide information about the type, size, storage location, etc. of the data, and can achieve effective classification and state identification of the data, thereby providing a basis for dynamic adjustment and optimization of the system.

[0076] Marginal fitness measure (MAD): The marginal fitness measure of each data point for each class is calculated using a probability density function such as the normal distribution. The formula is as follows:

[0077]

[0078] in, and They are Corresponding to The mean and standard deviation of the descriptors.

[0079] Step 1-1-4): Global Fitness Metric (GAD): Combine the MAD of all descriptors to calculate the fitness metric of the data point for each class.

[0080] In adaptive control systems, GAD calculation and update can be performed in real time, helping the system to dynamically adjust the control strategy and optimize system performance;

[0081] The fitness metric of the data point for each class is calculated by combining the marginal fitness metric (MAD) of all descriptors: MAD is required to calculate GAD: the MAD value of its descriptor and the center point of each class is calculated to obtain a fitness vector, which indicates the probability or fitness of the data point belonging to each class;

[0082]

[0083] in, is the normalized data point, Yes Category The The normalized value of the feature vector of the sample points, is a parameter for adjusting the fuzzy division, usually between 0 and 1. , is Descriptor Pair Class The marginal fitness measure of is the number of descriptors, where and They are Corresponding to The mean and standard deviation of the descriptors.

[0084] Step 1-1-5): Classification selection: When applying the LAMDA method for classification, the specific classification and the selection of classification nodes depend on the characteristics of the system and the control objectives. After classification, the model parameters of each category are usually fixed so that adaptive control strategies can be adopted for different operating states;

[0085] For example, the samples (preprocessed data) can be divided into three categories: : Category 1 (C1): low fitness; Category 2 (C2): medium fitness; Category 3 (C3): high fitness.

[0086] Step 1-1-6): Classification node: preset the corresponding GAD threshold. The standard for selecting the GAD threshold can be achieved through various methods such as empirical method, statistical method, cluster analysis and cross-validation. This design classifies the system status through empirical method: threshold 1 = 0.3; threshold 2 = 0.6.

[0087] Step 1-1-7): Classification rules:

[0088] Low fitness: If , then the sample belong ;

[0089] Medium fitness: If , then the sample belong ;

[0090] High fitness: If , then the sample belong .

[0091] Divide the system into Class and record it as .

[0092] Step 1-1-8): Model parameters after classification:

[0093] After classification, the model parameters of each category are usually fixed so that adaptive control strategies can be adopted for different operating states. Specific parameters include the state transfer matrix , control input matrix , control output matrix , process noise covariance matrix and the measurement noise covariance matrix .

[0094] Low fitness : ;

[0095] Medium fitness : ;

[0096] High adaptability : .

[0097] Step 1-1-9): Determine the class to which the data point belongs: With the data measured in real time, the system is able to classify the current operating state according to the classification rules. By calculating the GAD for each class, the data point is assigned to the class with the highest GAD value:

[0098]

[0099] Step 1-1-10): Update the category center: Update the center of each category according to the weighted average of the membership of the data points, that is, the mean in the membership function:

[0100]

[0101] in, Yes Category The The normalized value of the feature vector of the sample points, Yes Category The number of sample points in .

[0102] The above method is used to determine the category to which the data point belongs, and the data point is assigned to the class with the highest GAD value. The classification results are used for state estimation and control strategy adjustment.

[0103] Step 2: State estimation (Extended Kalman Filter, EKF), based on the classification results, select appropriate system parameters and noise matrix, and use the adjusted parameters to perform the prediction and update steps of the Kalman filter to provide accurate state estimation.

[0104] Step 2-1): State space model, construct the state space model of the HVAC system: Assume that the nonlinear state space model of the system is as follows:

[0105]

[0106] in, is the state vector, considering the indoor temperature and humidity ; is the control input vector, considering the cooling power and fan speed ; is the process noise, is the measurement noise; is the output vector, considering the indoor temperature and humidity , is the nonlinear state transfer function, is a nonlinear output function.

[0107] Step 2-2): Perform first-order Taylor expansion linearization on the state transfer equation and output equation in the nonlinear state space model, perform Taylor expansion linearization, and use the linearized equations for Kalman filtering.

[0108] Step 2-2-1): Linearization of state transfer equation;

[0109] For the nonlinear state transfer equation In the current state of estimation and control input Perform a first-order Taylor expansion at:

[0110]

[0111] in, is the state transition function About Status The Jacobian matrix of the current state and control input is used to calculate the Jacobian matrix, and the parameters of the preset classification are modified to improve the accuracy, which is recorded as :

[0112] .

[0113] State transfer function About control input Find the partial derivatives and get the Jacobian matrix :

[0114] .

[0115] The linearized state transfer equation is:

[0116] .

[0117] Step 2-2-2): Linearize the output equation;

[0118] For the nonlinear output equation In the current state of estimation Perform a first-order Taylor expansion at:

[0119]

[0120] in, is the output function About Status The Jacobian matrix of :

[0121]

[0122] The output equation after linearization is:

[0123] .

[0124] Step 2-3): Extended Kalman filter step: performs prediction and update steps to provide accurate state estimation.

[0125] At runtime, preset parameters are adjusted according to the current state and control input Perform real-time linearization and adjustments to improve control accuracy.

[0126] According to the classification results, appropriate system parameters and noise matrix are selected and linearization is completed to adjust the parameters of Kalman filter and MPC to adapt them to the current operating state. The adjusted parameters are used to perform the prediction and update steps of Kalman filter, provide accurate state estimation, and dynamically update model parameters so that the prediction model can accurately reflect the dynamic behavior of the current system. The prediction and update steps of extended Kalman filter are as follows:

[0127] Step 2-3-1): Initialization;

[0128] Initial state estimation ; Initial covariance matrix ; Process noise covariance matrix ; Measurement noise covariance matrix .

[0129]

[0130] Step 2-3-2): Prediction step;

[0131] Predict the next state:

[0132] .

[0133] Prediction covariance matrix:

[0134] .

[0135] State transition matrix :For the state transfer function About Status The Jacobian matrix of is used to linearize the nonlinear system.

[0136] Process noise covariance matrix : The process noise covariance matrix describes the process noise The statistical characteristics of the process noise are usually assumed to be zero-mean Gaussian white noise.

[0137] Measurement noise covariance matrix : The measurement noise covariance matrix describes the measurement noise The statistical characteristics of the measurement noise are usually assumed to be zero-mean Gaussian white noise.

[0138] Prior error covariance matrix : The prior error covariance matrix represents the covariance of the predicted state estimation error, which reflects the uncertainty of the state

[0139] Step 2-3-3): Update step;

[0140] Calculate Kalman gain :

[0141] .

[0142] Update state estimate :

[0143] .

[0144] Update the covariance matrix :

[0145] .

[0146] Kalman Gain : The Kalman gain balances the predicted state and the measured value, and determines the weight of the predicted value and the measured value when updating the estimated state.

[0147] Output Matrix :For the output function About Status The Jacobian matrix of , which is used to linearize the output model.

[0148] Posterior error covariance matrix : The posterior error covariance matrix represents the covariance of the updated state estimation error.

[0149] The above steps 2-3) are used to perform state estimation on the data previously collected by the sensor.

[0150] Steps 2-4): Dynamically adjust control inputs: At each control moment, repeat the above process and use the new state estimates and measurement data for the next optimization to ensure continuous optimization of the control strategy in the time-varying system. In nonlinear time-varying systems, the extended Kalman filter can provide accurate current state estimates, and MPC uses these state estimates and system changes in future moments for optimal control. By classifying and selecting appropriate system parameters and dynamically updating the state transfer matrix and control input matrix at each control moment, MPC can effectively respond to dynamic changes in the system and provide optimal control inputs. This method can significantly improve control effects and system performance in HVAC systems.

[0151] Model Adjustment: By dynamically updating the category centers, the parameters of LAMDA’s marginal fitness metric (MAD) and global fitness metric (GAD) can be adjusted according to new data. This dynamic adjustment ensures the accuracy of the classification and enables the system to continuously optimize according to new data.

[0152] Through the above method, accurate state estimation is achieved, model parameters are dynamically updated, and the state estimation results are used for MPC cost function calculation and optimization control strategy.

[0153] Step 3: Dynamically optimize control strategy;

[0154] Based on the new data and the identified states, the control rules and strategies of MPC are optimized. In each control cycle, the extended Kalman filter is used to update the model parameters so that the system model can dynamically reflect the actual situation and further improve the control accuracy. Based on the states and classifications identified by LAMDA, and based on the updated state estimates and model parameters of the Kalman filter, the control strategy is dynamically optimized.

[0155] In the above steps, after classification, each category corresponds to a different operating state, so that an adaptive control strategy can be adopted for different operating states.

[0156] The updated state estimate from the Kalman filter is used as the current state input to the MPC. Based on the current state and the prediction model, the MPC is used to solve the optimal control input sequence for a period of time in the future to adjust the operating parameters of the HVAC system. Inputting this accurate state estimate into the MPC can significantly improve the performance of the MPC. The specific steps are as follows:

[0157] Step 3-1: Define cost function: Using the state estimate as the initial state, define the cost function of the MPC.

[0158] When the Kalman filtered state estimate is input into the MPC, the cost function of the MPC will use the state estimate provided by the Kalman filter. As the current initial state. This means that each predicted state in the MPC optimization process The calculations are based on the current state estimate given by the Kalman filter, which gives the cost function:

[0159]

[0160] in, The state estimation is done by Kalman filtering. To make predictions; is the target state; is the control input vector (such as cooling power, fan speed, etc.); is the temperature error weight matrix; is the control input weight matrix; is the length of the prediction horizon.

[0161] Step 3-2: Optimize the constraints of the problem: including state constraints and control input constraints.

[0162] The optimization problem needs to satisfy the physical constraints and operational constraints of the system, including state constraints and control input constraints; the constraints can be expressed as:

[0163]

[0164] in, and are the upper and lower limits of the state vector; and are the upper and lower limits of the control input vector; and are the lower and upper bounds of the output vector.

[0165] Step 3-3: Use quadratic programming to solve the HVAC system optimization problem: minimize the cost function and solve for the optimal control input sequence.

[0166] By minimizing the cost function, the room temperature can be made as close to the target temperature as possible to ensure comfort, and the cooling power and fan speed can be controlled to reduce energy consumption and improve the energy efficiency of the system. By minimizing the cost function and satisfying the constraints, the optimal control input sequence can be solved. .

[0167] For example: Implementing optimal control strategies to adjust cooling power and fan speed; if you want to minimize cooling power at the same time and fan speed The energy consumption can be expanded into a quadratic programming form:

[0168]

[0169] in, is the control input vector, including and , is the coefficient matrix of the quadratic terms, is the coefficient vector of the linear terms.

[0170] Constraints:

[0171] Equality constraints: can be used to express the dynamic equations of the system

[0172] (Energy Balance)

[0173] in, Define equality constraints.

[0174] Inequality constraints can be used to express physical or operational restrictions:

[0175] Control input limits:

[0176] (temperature and power limits)

[0177] State restrictions: The specific form depends on the upper and lower limits of the state variable:

[0178]

[0179] in, Define inequality constraints.

[0180] According to the objective function The matrix (symmetric positive definite matrix) and vector:

[0181]

[0182] in, is the state transition matrix, is the control input matrix, is the state bias weight matrix, is the control input weight matrix, is the target state.

[0183] The optimal control input sequence is obtained in the above manner to adjust the operating parameters of the HVAC system.

[0184] Step 3-4: Implement the control input and apply the first control input in the optimal control input sequence calculated by MPC to the HVAC system. Monitor the actual output and predicted output of the system in real time, calculate the error, and feed the error information back to the model update and control strategy adjustment module to achieve closed-loop control.

[0185] Embodiment 2

[0186] The purpose of this embodiment is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the steps of the above method are implemented when the processor executes the program.

[0187] Embodiment 3

[0188] The purpose of this embodiment is to provide a computer-readable storage medium.

[0189] A computer-readable storage medium stores a computer program, which executes the steps of the above method when executed by a processor.

[0190] Embodiment 4

[0191] The purpose of this embodiment is to provide an adaptive model predictive control system for a building HVAC system, including:

[0192] The operation data acquisition module is configured to: acquire the operation data of the HVAC system under different working conditions and pre-process the acquired data;

[0193] A global fitness metric calculation module is configured to: calculate the marginal fitness metric of each data point for each class for the preprocessed data, calculate the fitness metric of the data point for each class by combining the marginal fitness metrics of all descriptors, and obtain a global fitness metric;

[0194] The classification module is configured to: set a global fitness measurement threshold and classify the current operation state of the system according to the classification rule;

[0195] The control module,is configured as follows: after classification, the model parameters of each,category are fixed to adopt adaptive control strategies for different,operating states;

[0196] In each control cycle, based on the classification results, the state estimate updated by the Kalman filter is used as the current state input of the adaptive model. Based on the current state and the prediction model, the optimal control input sequence in the future is solved to adjust the operating parameters of the HVAC system.

[0197] Embodiment 5

[0198] The purpose of this embodiment is to provide a computer program product containing instructions, which, when running on a computer, enables the computer to execute the methods and functions involved in any of the above embodiments.

[0199] The steps involved in the apparatus of the above embodiment correspond to the method embodiment 1, and the specific implementation method can refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0200] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0201] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. An adaptive model predictive control method for a building HVAC system, characterized in that: include: Obtain the operating data of the HVAC system under different working conditions and pre-process the acquired data; Calculate the marginal fitness measure of each data point for each class for the preprocessed data, and calculate the fitness measure of the data point for each class by combining the marginal fitness measures of all descriptors to obtain the global fitness measure; Set the global fitness measurement threshold and classify the current operating status of the system according to the classification rules; After classification, the model parameters of each class are fixed to adopt adaptive control strategies for different operating states; Use the classification results for state estimation and control strategy adjustment; In each control cycle, based on the classification results, the state estimate updated by the Kalman filter is used as the current state input of the adaptive model. Based on the current state and the prediction model, the optimal control input sequence in the future period is solved to adjust the operating parameters of the HVAC system. When the Kalman filtered state estimate is input into the MPC, the state estimate provided by the Kalman filter is used. As the current initial state, we get the cost function of MPC; The cost function is: in, The Kalman filter state estimation To make predictions; is the target state; is the control input vector; is the temperature error weight matrix; is the control input weight matrix; is the length of the prediction horizon; The constraints of the optimization problem, including state constraints and control input constraints; The optimal control input sequence is solved by minimizing the cost function and controlling the cooling power and fan speed to meet the constraints.

2. The adaptive model predictive control method for a building HVAC system according to claim 1, characterized in that: The model parameters of each category include a state transfer matrix, a control input matrix, a control output matrix, a process noise covariance matrix and a measurement noise covariance matrix.

3. The adaptive model predictive control method for a building HVAC system according to claim 1, characterized in that: include: The center of each category is updated according to the weighted average of the membership of the data points, that is, the mean in the membership function.

4. The adaptive model predictive control method for a building HVAC system according to claim 1, characterized in that: The state estimation process for HVAC systems includes: Construct a state-space model of the HVAC system; Linearize the state transition equations in the state space model; Linearize the output equations in the state-space model; At runtime, the preset parameters of the linearized state space model are linearized and adjusted in real time according to the current state and control input; The adjusted parameters are used for the prediction and update steps of the Kalman filter to provide accurate state estimation and dynamically update the model parameters so that the prediction model can accurately reflect the dynamic behavior of the current system.

5. The adaptive model predictive control method for a building HVAC system according to claim 4, characterized in that: Linearize the state transfer equation in the state space model. The specific process includes: Perform first-order Taylor expansion on the nonlinear state transfer equations in the state-space model at the current state estimate and control input; The Jacobian matrix is ​​calculated using the current state and control input, the parameters of the preset classification are modified, and the partial derivative of the state transfer function with respect to the control input is obtained to obtain the Jacobian matrix; Finally, the linearized state transfer equation is obtained; Linearize the output equations in the state-space model, including: Perform a first-order Taylor expansion on the nonlinear output equation at the current state estimate to obtain the linearized output equation.

6. An adaptive model predictive control system for a building HVAC system, comprising: The operation data acquisition module is configured to: acquire the operation data of the HVAC system under different working conditions and pre-process the acquired data; A global fitness metric calculation module is configured to: calculate the marginal fitness metric of each data point for each class for the preprocessed data, calculate the fitness metric of the data point for each class by combining the marginal fitness metrics of all descriptors, and obtain a global fitness metric; The classification module is configured to: set a global fitness measurement threshold and classify the current operation state of the system according to the classification rule; The control module is configured as follows: after classification, the model parameters of each category are fixed so as to adopt adaptive control strategies for different operating states; the classification results are used for state estimation and adjustment of control strategies; In each control cycle, based on the classification results, the state estimate updated by the Kalman filter is used as the current state input of the adaptive model. Based on the current state and the prediction model, the optimal control input sequence in the future period is solved to adjust the operating parameters of the HVAC system. When the Kalman filtered state estimate is input into the MPC, the state estimate provided by the Kalman filter is used. As the current initial state, we get the cost function of MPC; The cost function is: in, The Kalman filter state estimation To make predictions; is the target state; is the control input vector; is the temperature error weight matrix; is the control input weight matrix; is the length of the prediction horizon; The constraints of the optimization problem, including state constraints and control input constraints; The optimal control input sequence is solved by minimizing the cost function and controlling the cooling power and fan speed to meet the constraints.

7. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method described in any one of claims 1 to 5 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 5 are performed.

Images