Operating parameter optimization control method and system of air condenser

Abstract

The invention relates to the technical field of optimization control, in particular to an operation parameter optimization control method and system for an air condenser, and the method comprises the steps: firstly obtaining environment data, processing the environment data to obtain a self-adaptive environment monitoring value, and constructing a non-uniform frequency conversion strategy according to the self-adaptive environment monitoring value to control the optimization frequency of the operation parameters of the air condenser; then obtaining structural features and operation features of the air condenser, obtaining a first to-be-optimized parameter vector according to a principal component analysis method, obtaining the first to-be-optimized parameter vector, and carrying out sensitivity analysis to obtain a second to-be-optimized parameter vector; and then the optimization parameter prediction model predicts a parameter value vector of the second to-be-optimized parameter vector, the distributed parameter optimization control model sets a local agent of the to-be-optimized parameter, the parameter value vector of the second to-be-optimized parameter vector is acquired, and parameter optimization control is performed through a reinforcement learning algorithm. According to the method, through optimal control over the operation parameters, the parameter optimization speed and stability are improved, and the energy consumption of the air condenser is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of optimization control, and in particular to an operation parameter optimization control method and system for an air condenser. Background Art

[0002] At present, the optimization control methods for the operating parameters of air condensers mainly include using fixed set values ​​or simple feedback control, simple PID control and timed start and stop control. For example, frequently starting and stopping the condenser fan within a certain time interval can easily lead to poor control effects and increase equipment wear.

[0003] Traditional control algorithms find it difficult to achieve fast and stable optimization control of operating parameters under different operating conditions. Therefore, it is necessary to use adaptive control methods such as deep reinforcement learning and fuzzy control, combined with the use of the equipment itself, to dynamically adjust the optimization frequency of the operating parameters of the air condenser to achieve accurate and stable optimization control of the operating parameters.

[0004] Therefore, an operating parameter optimization control method and system for an air condenser are proposed. Summary of the invention

[0005] The purpose of the present invention is to provide an operating parameter optimization control method and system for an air condenser. First, environmental data is obtained and processed to obtain an adaptive environmental monitoring value, and a non-uniform frequency conversion strategy is constructed according to the adaptive environmental monitoring value to control the optimized frequency of the operating parameters of the air condenser; then, the structural characteristics and operating characteristics of the air condenser are obtained and a first parameter vector to be optimized is obtained according to the principal component analysis method, and the first parameter vector to be optimized is obtained and a sensitivity analysis is performed to obtain a second parameter vector to be optimized; then, an optimization parameter prediction model is used to predict the parameter value vector of the second parameter vector to be optimized, and a distributed parameter optimization control model sets a local intelligent agent of the parameter to be optimized, obtains the parameter value vector of the second parameter vector to be optimized, and performs parameter optimization control through a reinforcement learning algorithm.

[0006] To achieve the above object, the present invention provides the following technical solutions: An operating parameter optimization control method for an air condenser, comprising: Acquire and process environmental data to obtain adaptive environmental monitoring values, and construct a non-uniform frequency conversion strategy to control the optimal frequency of operating parameters of the air condenser according to the adaptive environmental monitoring values; Furthermore, the adaptive environmental monitoring value includes a thermal stress term, a radiation term and a time-varying correction term, and the calculation formula of the adaptive environmental monitoring value is: ; in, Represents the adaptive environmental monitoring value, and represents the adaptive weight coefficient, Indicates temperature, represents the temperature nonlinear coefficient, represents the symbolic function, Indicates the preset temperature, represents the humidity coupling coefficient, Indicates humidity, Indicates saturated humidity, represents the wind speed attenuation coefficient, represents the square root of wind speed, represents the radiant heat intensity, and represents the pressure modulation parameter, represents the natural base, Indicates atmospheric pressure, represents standard atmospheric pressure, and represents the historical trend sensitivity coefficient, represents the hyperbolic tangent function, Indicates the temperature change trend.

[0007] Furthermore, the non-uniform frequency conversion strategy includes a transient response term, a memory decay integral term and a non-linear compensation term; the calculation formula of the non-uniform frequency conversion strategy is: ; in, represents the non-uniform frequency conversion strategy, and represents the transient response term coefficient and the memory decay integral term coefficient, represents the derivative of the adaptive environmental monitoring value, represents Gaussian kernel convolution, represents the bandwidth adaptively adjusted according to the signal-to-noise ratio, represents the adaptive environmental monitoring value function, represents the memory decay coefficient, Indicates time, represents the integral variable, represents the coefficient of nonlinear compensation term, represents the symbolic function, represents the gradient operator, represents the Laplace term.

[0008] Acquire structural characteristics and operating characteristics of the air condenser and obtain a first parameter vector to be optimized according to a principal component analysis method; acquire the first parameter vector to be optimized and perform a sensitivity analysis to obtain a second parameter vector to be optimized; Furthermore, the parameter values ​​in the first parameter vector to be optimized are obtained and the sensitivity index of the parameter values ​​is calculated through sensitivity analysis to obtain a sensitivity index set; the elements in the sensitivity index set are obtained and sorted and the cumulative contribution is calculated, and the second parameter vector to be optimized is constructed according to the parameter to be optimized corresponding to the sensitivity index when the cumulative contribution exceeds a threshold.

[0009] The optimization parameter prediction model obtains all parameters and characteristics of the air condenser and performs prediction to obtain a parameter value vector of the second parameter vector to be optimized; Furthermore, the step of constructing the optimization parameter prediction model includes: Step S1: Establish a comprehensive target evaluation mechanism to obtain a multi-target scoring label, wherein the comprehensive target evaluation mechanism includes a heat exchange target, an energy-saving target, and an equipment status target; set the initial weight through the AHP hierarchical analysis method, and use an adaptive fuzzy reasoning system to adjust the weight value in real time; Step S2: Obtain historical environmental data, operating parameters, and equipment status data and capture the equipment response characteristics in the time series through a multi-layer LSTM network and an attention enhancement mechanism to obtain equipment features; Step S3: training the XGBoost regression model; Step S4: Obtain the trained XGBoost regression model and perform prediction, where the model input includes environmental characteristics, equipment characteristics, historical parameters, and multi-objective scoring labels; the model output includes the parameter value vector of the second parameter vector to be optimized.

[0010] The distributed parameter optimization control model includes a local agent that sets the parameters to be optimized. The local agent obtains the parameter value vector of the second parameter vector to be optimized and performs parameter optimization control through a reinforcement learning algorithm.

[0011] Further, the steps of constructing the distributed parameter optimization control model include: Step S10: configuring a local agent for each parameter, and using a deep reinforcement learning algorithm for local control; the state space of the local agent includes the state information of the current environment data, the equipment operation state and the parameter value vector; the action space includes the adjustment range of the corresponding parameter; Step S20: constructing a reward function for the local agent according to the heat exchange efficiency, energy consumption and parameter change amplitude; Step S30: Perform training according to the deep reinforcement learning algorithm to obtain a trained local agent; Step S40: obtaining the parameter value vector of the trained local agent and the second parameter vector to be optimized, and adopting a progressive adjustment strategy if the deviation between the optimized parameter value and the current actual parameter value exceeds a set threshold; Step S50: Obtain the cumulative reward of the local agent. After reaching the set objective function convergence condition or reaching the maximum number of iterations, stop the iteration and the local agent control execution ends.

[0012] The present invention also provides an operating parameter optimization control system for an air condenser, comprising: A control frequency optimization module, including an adaptive environmental monitoring value calculation unit and a non-uniform frequency conversion strategy control unit; The module for selecting parameters to be optimized is used to obtain the structural characteristics and operating characteristics of the air condenser and obtain a first parameter vector to be optimized according to the principal component analysis method; obtain the first parameter vector to be optimized and perform sensitivity analysis to obtain a second parameter vector to be optimized; An optimization parameter prediction module is used to obtain all parameters and features and predict through multi-objective modeling, feature extraction and XGBoost regression model to obtain a parameter value vector of the second parameter vector to be optimized; The distributed control module includes a distributed parameter optimization control model, which is used to set a local agent of the parameters to be optimized, obtain the parameter value vector of the second parameter vector to be optimized, and perform parameter optimization control through a reinforcement learning algorithm.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. By acquiring and processing real-time environmental data, an adaptive environmental monitoring value including thermal stress terms, radiation terms and time-varying correction terms is constructed. The system can accurately reflect the dynamic changes of multi-dimensional environmental factors such as temperature, humidity, wind speed, radiation intensity and atmospheric pressure. The non-uniform frequency conversion strategy constructed using adaptive environmental monitoring values ​​combines the transient response term, memory decay integral term and nonlinear compensation term, which can take into account both the short-term rapid response of the system and the influence of long-term historical trends, thereby ensuring that the optimized frequency of the operating parameters can respond quickly to sudden changes.

[0014] 2. Through principal component analysis and sensitivity index calculation, the parameters that have the greatest impact on heat exchange, energy saving and equipment status are selected to form the second parameter vector to be optimized, which can ensure that the optimization control focus is on the most sensitive and critical parameters, thereby improving the overall control effect and robustness; adopting the target comprehensive evaluation mechanism, multiple key targets such as heat exchange, energy saving and equipment status are unified into a model, and a multi-target scoring label is constructed. Then, the XGBoost regression model is used for prediction, which can integrate multi-dimensional features and achieve high-precision prediction of key parameter value vectors.

[0015] 3. By constructing a reward function based on heat exchange efficiency, energy consumption and parameter change range, it is possible to pursue the best energy efficiency and heat exchange performance while ensuring equipment safety and stable operation. This reward mechanism ensures that the control strategy achieves global optimization; the distributed parameter optimization control model can achieve balanced optimization of multiple objectives such as heat exchange and energy consumption while ensuring system safety and equipment protection, thereby improving overall control efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A flow chart of an operating parameter optimization control method for an air condenser provided in an embodiment of the present invention; Figure 2 A flowchart of constructing an optimization parameter prediction model provided by an embodiment of the present invention; Figure 3 A schematic structural diagram of an operating parameter optimization control system for an air condenser provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] Embodiment 1 In order to improve the working efficiency of the cooling system and improve energy efficiency, a data center introduced an air condenser operating parameter optimization control method provided by the present invention. The method flow is as follows: Figure 1 As shown, the specific implementation is as follows: Acquire and process environmental data to obtain adaptive environmental monitoring values, and construct a non-uniform frequency conversion strategy to control the optimal frequency of operating parameters of the air condenser according to the adaptive environmental monitoring values; Furthermore, the adaptive environmental monitoring value includes a thermal stress term, a radiation term and a time-varying correction term, corresponding to the first, second and third terms in the formula, and the calculation formula of the adaptive environmental monitoring value is: ; in, Represents the adaptive environmental monitoring value, and represents the adaptive weight coefficient, Indicates temperature, represents the temperature nonlinear coefficient, represents the symbolic function, Indicates the preset temperature, represents the humidity coupling coefficient, Indicates humidity, Indicates saturated humidity, represents the wind speed attenuation coefficient, represents the square root of wind speed, represents the radiant heat intensity, and represents the pressure modulation parameter, represents the natural base, Indicates atmospheric pressure, represents standard atmospheric pressure, and represents the historical trend sensitivity coefficient, represents the hyperbolic tangent function, Indicates the temperature change trend; Furthermore, the calculation formula for the temperature change trend is: ; in, represents the time window, Indicates time, represents the temperature function; Furthermore, the adaptive weight and Dynamically updated through Kalman filtering. Table 1 shows the adaptive environmental monitoring values ​​at different time nodes.

[0019] Table 1. Adaptive environmental monitoring values

[0020] The adaptive environmental monitoring values ​​integrate multiple factors such as temperature, humidity, wind speed, radiant heat and atmospheric pressure, and introduce parameters such as temperature nonlinearity, humidity coupling and historical trend sensitivity coefficient, so that the monitoring values ​​can more accurately reflect the dynamic changes of the actual environment, help avoid the deviations that may be introduced by a single indicator, and improve the accuracy of environmental monitoring.

[0021] Furthermore, the non-uniform frequency conversion strategy includes a transient response term, a memory decay integral term and a non-linear compensation term, corresponding to the first, second and third terms in the formula, and the calculation formula of the non-uniform frequency conversion strategy is: ; in, represents the non-uniform frequency conversion strategy, and represents the transient response term coefficient and the memory decay integral term coefficient, represents the derivative of the adaptive environmental monitoring value, represents Gaussian kernel convolution, represents the bandwidth adaptively adjusted according to the signal-to-noise ratio, represents the adaptive environmental monitoring value function, represents the memory decay coefficient, Indicates time, represents the integral variable, represents the coefficient of nonlinear compensation term, represents the symbolic function, represents the gradient operator, represents the Laplace term; Furthermore, it is determined whether to optimize the operating parameters of the air condenser according to the non-uniform frequency conversion strategy.

[0022] This non-uniform frequency conversion strategy achieves a comprehensive response to environmental signals through three complementary parts. It can intelligently determine whether it is necessary to optimize the operating parameters of the air condenser. When the environment changes greatly, the optimization frequency is automatically increased to respond quickly; when the environment is stable, the optimization frequency is reduced to avoid frequent adjustments, thereby saving computing resources and reducing equipment wear.

[0023] Acquire structural characteristics and operating characteristics of the air condenser and obtain a first parameter vector to be optimized according to a principal component analysis method; acquire the first parameter vector to be optimized and perform a sensitivity analysis to obtain a second parameter vector to be optimized; Furthermore, the parameter values ​​in the first parameter vector to be optimized are obtained and the sensitivity index of the parameters is calculated through sensitivity analysis to obtain a sensitivity index set; the elements in the sensitivity index set are obtained and sorted and the cumulative contribution is calculated, and the second parameter vector to be optimized is constructed according to the parameters to be optimized corresponding to the sensitivity index when the cumulative contribution exceeds the threshold.

[0024] Furthermore, the calculation formula of the sensitivity index is: ; in, represents the first parameter vector to be optimized. Parameters to be optimized The sensitivity index, represents the system response variable, represents the variance, express expectations; The sensitivity index can measure the contribution of each parameter to the system response variable. By sorting and calculating the cumulative contribution, the parameters with less impact on the response can be eliminated, thereby reducing the dimension of the parameters to be optimized, which helps to reduce the computational complexity of the subsequent optimization control algorithm and improve the control speed.

[0025] The parameter prediction model is optimized to obtain all parameters and features and to perform predictions through multi-objective modeling, feature extraction, and XGBoost regression model to obtain a parameter value vector of the second parameter vector to be optimized; Furthermore, the steps of constructing the optimization parameter prediction model are as follows: Figure 2 As shown, including: Step S1: Establish a comprehensive target evaluation mechanism to obtain a multi-target scoring label, wherein the comprehensive target evaluation mechanism includes a heat exchange target, an energy-saving target, and an equipment status target; set the initial weight through the AHP hierarchical analysis method, and use adaptive fuzzy reasoning to adjust the weight value in real time; Furthermore, the heat transfer target ,in represents the maximum heat transfer efficiency, Indicates lower condensing temperature Weight coefficient of energy saving target ,in Indicates the minimum operating power consumption, Energy saving benefits Weight coefficient of device status target ,in Indicates the degree of equipment wear. Indicates equipment maintenance cost The weight coefficient of Furthermore, the weight coefficient , and It is determined by AHP and adaptive fuzzy reasoning.

[0026] Step S2: Obtain historical environmental data, operating parameters, and equipment status data and capture the equipment response characteristics in the time series through a multi-layer LSTM network and an attention enhancement mechanism to obtain equipment features; Step S3: training the XGBoost regression model, the training data set includes multi-objective scoring labels and data sets, and the data sets include environmental features, equipment features, and historical parameters; Step S4: Obtain the trained XGBoost regression model and perform prediction, where the model input includes environmental characteristics, equipment characteristics, historical parameters, and multi-objective scoring labels; the model output includes the parameter value vector of the second parameter vector to be optimized.

[0027] A comprehensive target evaluation mechanism is used to obtain multi-target scoring labels, and multiple key targets such as heat exchange, energy saving and equipment status are unified into a model, so that the system does not favor a single target during the optimization process, thereby achieving a balanced improvement in overall performance; the initial weights are set through the AHP hierarchical analysis method, and the weight values ​​are adjusted in real time using the adaptive fuzzy reasoning system, which can flexibly modify the target importance according to actual operating conditions and environmental changes; the XGBoost regression model is used for prediction, which can integrate multi-dimensional features and achieve high-precision prediction of key parameter value vectors.

[0028] The distributed parameter optimization control model obtains the parameter value vector of the second parameter vector to be optimized and performs parameter optimization control; the distributed parameter optimization control model includes a local agent that sets the parameters to be optimized and performs parameter optimization control through a reinforcement learning algorithm.

[0029] Furthermore, the step of constructing the distributed parameter optimization control model includes: Step S10: a local agent is configured for each parameter, and a deep reinforcement learning algorithm is used for local control; the state space of the local agent includes the state information of the current environment data, the equipment operation status and the parameter value vector; the action space includes the adjustment range of the corresponding parameter; Step S20: constructing a reward function for the local agent according to the heat exchange efficiency, energy consumption and parameter change amplitude; Furthermore, the formula of the reward function is: ; in, represents the reward function, represents the hyperbolic tangent function, represents the actual heat transfer efficiency, represents the benchmark heat transfer efficiency, represents the maximum heat transfer efficiency, represents the logarithmic function, Indicates the actual power, Indicates rated power, represents the parameter change between adjacent time steps, represents the two-norm.

[0030] Step S30: training the local agent according to the deep reinforcement learning algorithm to obtain a trained local agent; Step S40: obtaining the parameter value vector of the trained local agent and the second parameter vector to be optimized, and adopting a progressive adjustment strategy if the deviation between the optimized parameter value and the current actual parameter value exceeds a set threshold; Furthermore, the progressive adjustment strategy means that if the deviation exceeds the set threshold, the deviation is segmented into multiple levels, and then the parameter optimization control is performed step by step.

[0031] Step S50: Obtain the cumulative reward of the local agent, and stop the iteration after reaching the set objective function convergence condition or reaching the maximum number of iterations, and the local agent control execution ends. Table 2 shows the data when the local agent performs parameter optimization control.

[0032] Table 2. Operation parameter control data

[0033] Each parameter is managed by a dedicated local agent, whose state space covers the current environmental data, equipment operating status and parameter value information, so that each local agent can make fine adjustments to the specific dynamics of the parameter, thereby achieving more precise local control; by constructing a reward function based on heat transfer efficiency, energy consumption and parameter change amplitude, it is possible to pursue optimal energy efficiency and heat exchange performance and achieve global optimization while ensuring the safety and stable operation of the equipment; when the cumulative reward reaches the set convergence condition or the upper limit of the number of iterations, the system automatically stops adjusting to ensure that the entire control process converges to the optimal state quickly and efficiently.

[0034] First, by collecting environmental data in real time and processing it to obtain adaptive environmental monitoring values, it is possible to capture dynamic changes in the environment, thereby constructing a non-uniform frequency conversion strategy and flexibly adjusting the parameter optimization frequency to cope with external disturbances; secondly, principal component analysis and sensitivity analysis are used to screen out the key parameters that have the greatest impact on system response from a number of structural and operational characteristics, significantly reducing the dimensions of the parameters to be optimized and improving the computational efficiency and accuracy of subsequent optimization; then, the target comprehensive evaluation mechanism combines feature extraction and the XGBoost regression model to accurately predict the optimal values ​​of key parameters and achieve a comprehensive balance of thermal efficiency, energy saving and equipment status; finally, through the distributed parameter optimization control model, each local intelligent agent performs parameter optimization based on a deep reinforcement learning algorithm to achieve real-time, robust and safe optimization control of global parameters. In summary, the method provided by the present invention greatly improves the response speed, energy efficiency and stability of parameter optimization, and reduces energy consumption and equipment wear.

[0035] Embodiment 2 The present invention also provides an operating parameter optimization control system for an air condenser, the system structure is as follows: Figure 3 As shown, the specific implementation is as follows: A control frequency optimization module, including an adaptive environmental monitoring value calculation unit and a non-uniform frequency conversion strategy control unit; Further, environmental data is acquired and processed to obtain adaptive environmental monitoring values, and a non-uniform frequency conversion strategy is constructed according to the adaptive environmental monitoring values ​​to control the optimized frequency of the operating parameters of the air condenser; Furthermore, the adaptive environmental monitoring value includes a thermal stress term, a radiation term and a time-varying correction term, and the calculation formula of the adaptive environmental monitoring value is: ; in, Represents the adaptive environmental monitoring value, and represents the adaptive weight coefficient, Indicates temperature, represents the temperature nonlinear coefficient, represents the symbolic function, Indicates the preset temperature, represents the humidity coupling coefficient, Indicates humidity, Indicates saturated humidity, represents the wind speed attenuation coefficient, represents the square root of wind speed, represents the radiant heat intensity, and represents the pressure modulation parameter, represents the natural base, Indicates atmospheric pressure, represents standard atmospheric pressure, and represents the historical trend sensitivity coefficient, represents the hyperbolic tangent function, Indicates the temperature change trend.

[0036] Furthermore, the non-uniform frequency conversion strategy includes a transient response term, a memory decay integral term and a non-linear compensation term; the calculation formula of the non-uniform frequency conversion strategy is: ; in, represents the non-uniform frequency conversion strategy, and represents the transient response term coefficient and the memory decay integral term coefficient, represents the derivative of the adaptive environmental monitoring value, represents Gaussian kernel convolution, represents the bandwidth adaptively adjusted according to the signal-to-noise ratio, represents the adaptive environmental monitoring value function, represents the memory decay coefficient, Indicates time, represents the integral variable, represents the coefficient of nonlinear compensation term, represents the symbolic function, represents the gradient operator, Denotes the Laplace term. As shown in Table 3, whether to perform parameter optimization control is determined according to the non-uniform frequency conversion strategy at different time nodes, and the control accuracy is improved to 98.3%, the false trigger rate is reduced to 3.1%, and the response delay is controlled within 50ms.

[0037] Table 3. Parameter optimization control decisions for some time nodes

[0038] The module for selecting parameters to be optimized is used to obtain the structural characteristics and operating characteristics of the air condenser and obtain a first parameter vector to be optimized according to the principal component analysis method; obtain the first parameter vector to be optimized and perform sensitivity analysis to obtain a second parameter vector to be optimized; Furthermore, the parameter values ​​in the first parameter vector to be optimized are obtained and the sensitivity index of the parameters is calculated through sensitivity analysis to obtain a sensitivity index set; the elements in the sensitivity index set are obtained and sorted and the cumulative contribution is calculated, and the second parameter vector to be optimized is constructed according to the parameters to be optimized corresponding to the sensitivity index when the cumulative contribution exceeds the threshold.

[0039] An optimization parameter prediction module is used to obtain all parameters and features and predict through multi-objective modeling, feature extraction and XGBoost regression model to obtain a parameter value vector of the second parameter vector to be optimized; Furthermore, the step of constructing the optimization parameter prediction model includes: Step S1: Establish a comprehensive target evaluation mechanism to obtain a multi-target scoring label, wherein the comprehensive target evaluation mechanism includes a heat exchange target, an energy-saving target, and an equipment status target; set the initial weight through the AHP hierarchical analysis method, and use an adaptive fuzzy reasoning system to adjust the weight value in real time; Step S2: Obtain historical environmental data, operating parameters, and equipment status data and capture the equipment response characteristics in the time series through a multi-layer LSTM network and an attention enhancement mechanism to obtain equipment features; Step S3: training the XGBoost regression model; Step S4: Obtain the trained XGBoost regression model and perform prediction, where the model input includes environmental characteristics, equipment characteristics, historical parameters, and multi-objective scoring labels; the model output includes the parameter value vector of the second parameter vector to be optimized.

[0040] A distributed control module is used to obtain the parameter value vector of the second parameter vector to be optimized and perform parameter optimization control; the distributed parameter optimization control model includes setting a local intelligent agent for the parameters to be optimized and performing parameter optimization control through a reinforcement learning algorithm.

[0041] Furthermore, the step of constructing the distributed parameter optimization control model includes: Step S10: a local agent is configured for each parameter, and a deep reinforcement learning algorithm is used for local control; the state space of the local agent includes the state information of the current environment data, the equipment operation status and the parameter value vector; the action space includes the adjustment range of the corresponding parameter; Step S20: constructing a reward function for the local agent according to the heat exchange efficiency, energy consumption and parameter change amplitude; Step S30: Perform training according to the deep reinforcement learning algorithm to obtain a trained local agent; Furthermore, there are reinforcement learning algorithms such as PPO algorithm, DDPG algorithm and TD3 algorithm.

[0042] Step S40: obtaining the parameter value vector of the trained local agent and the second parameter vector to be optimized, and adopting a progressive adjustment strategy if the deviation between the optimized parameter value and the current actual parameter value exceeds a set threshold; Step S50: Obtain the cumulative reward of the local agent. After reaching the set objective function convergence condition or reaching the maximum number of iterations, stop the iteration and the local agent control execution ends.

[0043] The system proposed in the present invention realizes real-time intelligent optimization control of the operating parameters of the air condenser by controlling the frequency optimization module, the parameter selection module to be optimized, the optimization parameter prediction module and the distributed control module, which significantly enhances the adaptability, robustness and operation efficiency of the system.

[0044] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for optimizing and controlling operating parameters of an air condenser, characterized in that: include: Acquire and process environmental data to obtain adaptive environmental monitoring values, and construct a non-uniform frequency conversion strategy to control the optimal frequency of operating parameters of the air condenser according to the adaptive environmental monitoring values; Acquire structural characteristics and operating characteristics of the air condenser and obtain a first parameter vector to be optimized according to a principal component analysis method; acquire the first parameter vector to be optimized and perform a sensitivity analysis to obtain a second parameter vector to be optimized; The optimization parameter prediction model obtains all parameters and characteristics of the air condenser and performs prediction to obtain a parameter value vector of the second parameter vector to be optimized; The distributed parameter optimization control model includes a local agent that sets the parameters to be optimized. The local agent obtains the parameter value vector of the second parameter vector to be optimized and performs parameter optimization control through a reinforcement learning algorithm.

2. The operating parameter optimization control method of an air condenser according to claim 1, characterized in that: The adaptive environmental monitoring value includes a thermal stress term, a radiation term and a time-varying correction term. The calculation formula of the adaptive environmental monitoring value is: ； in, Represents the adaptive environmental monitoring value, and represents the adaptive weight coefficient, Indicates temperature, represents the temperature nonlinear coefficient, represents the symbolic function, Indicates the preset temperature, represents the humidity coupling coefficient, Indicates humidity, Indicates saturated humidity, represents the wind speed attenuation coefficient, represents the square root of wind speed, represents the radiant heat intensity, and represents the pressure modulation parameter, represents the natural base, Indicates atmospheric pressure, represents standard atmospheric pressure, and represents the historical trend sensitivity coefficient, represents the hyperbolic tangent function, Indicates the temperature change trend.

3. The operating parameter optimization control method of an air condenser according to claim 1, characterized in that: The non-uniform frequency conversion strategy includes a transient response term, a memory decay integral term and a non-linear compensation term; the calculation formula of the non-uniform frequency conversion strategy is: ； in, represents the non-uniform frequency conversion strategy, and represents the transient response term coefficient and the memory decay integral term coefficient, represents the derivative of the adaptive environmental monitoring value, represents Gaussian kernel convolution, represents the bandwidth adaptively adjusted according to the signal-to-noise ratio, represents the adaptive environmental monitoring value function, represents the memory decay coefficient, Indicates time, represents the integral variable, represents the coefficient of nonlinear compensation term, represents the symbolic function, represents the gradient operator, represents the Laplace term.

4. The operating parameter optimization control method of an air condenser according to claim 1, characterized in that: It also includes obtaining the parameter values ​​in the first parameter vector to be optimized and calculating the sensitivity index of the parameter value through sensitivity analysis to obtain a sensitivity index set; obtaining the elements in the sensitivity index set and sorting and calculating the cumulative contribution, and constructing a second parameter vector to be optimized based on the parameter to be optimized corresponding to the sensitivity index when the cumulative contribution exceeds a threshold.

5. The operating parameter optimization control method of an air condenser according to claim 1, characterized in that: The steps of constructing the optimization parameter prediction model include: step S1: establishing a comprehensive target evaluation mechanism to obtain a multi-target scoring label, wherein the comprehensive target evaluation mechanism includes a heat exchange target, an energy-saving target, and an equipment status target; setting initial weights by using an AHP hierarchical analysis method, and using an adaptive fuzzy inference system to adjust the weight values ​​in real time; Step S2: Obtain historical environmental data, operating parameters, and equipment status data and capture the equipment response characteristics in the time series through a multi-layer LSTM network and an attention enhancement mechanism to obtain equipment features; Step S3: training the XGBoost regression model; Step S4: Obtain the trained XGBoost regression model and perform prediction, where the model input includes environmental characteristics, equipment characteristics, historical parameters, and multi-objective scoring labels; the model output includes the parameter value vector of the second parameter vector to be optimized.

6. The operating parameter optimization control method of an air condenser according to claim 1, characterized in that: The steps of constructing the distributed parameter optimization control model include: step S10: configuring a local agent for each parameter, and using a deep reinforcement learning algorithm for local control; the state space of the local agent includes the state information of the current environment data, the equipment operation state and the parameter value vector; the action space includes the adjustment range of the corresponding parameter; Step S20: constructing a reward function for the local agent according to the heat exchange efficiency, energy consumption and parameter change amplitude; Step S30: Perform training according to the deep reinforcement learning algorithm to obtain a trained local agent; Step S40: obtaining the parameter value vector of the trained local agent and the second parameter vector to be optimized, and adopting a progressive adjustment strategy if the deviation between the optimized parameter value and the current actual parameter value exceeds a set threshold; Step S50: Obtain the cumulative reward of the local agent. After reaching the set objective function convergence condition or reaching the maximum number of iterations, stop the iteration and the local agent control execution ends.

7. An operating parameter optimization control system for an air condenser, characterized in that: include: A control frequency optimization module, including an adaptive environmental monitoring value calculation unit and a non-uniform frequency conversion strategy control unit; The module for selecting parameters to be optimized is used to obtain the structural characteristics and operating characteristics of the air condenser and obtain a first parameter vector to be optimized according to the principal component analysis method; obtain the first parameter vector to be optimized and perform sensitivity analysis to obtain a second parameter vector to be optimized; An optimization parameter prediction module is used to obtain all parameters and features and predict through multi-objective modeling, feature extraction and XGBoost regression model to obtain a parameter value vector of the second parameter vector to be optimized; The distributed control module includes a distributed parameter optimization control model, which is used to set a local agent of the parameters to be optimized, obtain the parameter value vector of the second parameter vector to be optimized, and perform parameter optimization control through a reinforcement learning algorithm.

Images