Variable frequency fan control method and system

Abstract

The invention discloses a frequency conversion fan control method and system, and the method comprises the steps: obtaining the related parameters of a frequency conversion fan, and the related parameters comprise an operation parameter and an environment parameter; inputting relevant parameters of the frequency conversion fan into the frequency conversion fan neural network model, and outputting a target frequency; the variable frequency fan neural network model comprises a hidden layer 1, a hidden layer 2 and an output layer, a weight matrix and a deviation term vector of each layer are optimized by minimizing a loss function, nonlinearity is introduced into the hidden layer 1, so that the neural network can learn a complex mode, and the variable frequency fan neural network model can learn a non-linear mode; the hidden layer 2 controls the response range of neurons and provides smoother signals for the subsequent output layer, and the output layer outputs the target frequency most suitable for the current condition. According to the method, the complex nonlinear mapping relation from the current input state to the single optimal target frequency is achieved, the target frequency can adapt to variable condition parameters, and the adaptability of the draught fan is improved.

Description

Technical Field

[0001] This invention belongs to the field of wind turbine control technology, specifically, it relates to a variable frequency wind turbine control method and system. Background Technology

[0002] In large-scale settings such as industrial plants, civil buildings, or agricultural farms, environmental control is of paramount importance. Among these, the ventilation system is a key component, requiring the installation of fans to meet the needs of ventilation, heat dissipation, and pressure balance.

[0003] Traditional fan control schemes typically rely on fixed rules or simple feedback control (such as PID control), which makes it difficult to optimize airflow output in real time to adapt to changing environmental conditions (such as temperature, humidity, pressure, etc.) and cannot dynamically adjust airflow and pressure according to actual needs, resulting in energy waste or failure to meet environmental requirements.

[0004] In addition, these traditional methods perform poorly when dealing with nonlinear and time-varying problems, resulting in high system energy consumption, low operating efficiency, and may cause equipment overload or performance degradation.

[0005] The information disclosed in this background section is only intended to enhance the understanding of the background technology of this application, and therefore may include prior art that is not known to those skilled in the art. Summary of the Invention

[0006] This invention proposes a variable frequency fan control method and system to solve the technical problem that existing fan control cannot dynamically adjust air volume and pressure according to actual needs, resulting in energy waste or substandard environmental conditions.

[0007] To achieve the above-mentioned invention / design objectives, the present invention adopts the following technical solution:

[0008] A variable frequency fan control method, the control method comprising:

[0009] Obtain relevant parameters of the variable frequency fan, including operating parameters and environmental parameters;

[0010] Input the relevant parameters of the variable frequency fan into the neural network model of the variable frequency fan, and output the target frequency. ;

[0011] According to the target frequency Control the operation of the variable frequency fan;

[0012] The neural network model for variable frequency fans includes:

[0013] Hidden layer 1: , This is a vector of relevant parameters for the variable frequency fan; This is the weight matrix. The vector of deviation terms;

[0014] Hidden layer 2: , This is the weight matrix. The vector of deviation terms;

[0015] Output layer: , This is the weight matrix. The vector of deviation terms;

[0016] The weight matrix and bias term vector are obtained by minimizing the loss function. Obtained through training. It is the predicted target frequency of the model for the i-th sample. It is the actual target frequency of the i-th sample.

[0017] The variable frequency fan control method described above obtains the actual output air volume of the variable frequency fan. / wind pressure Determine the load power of the variable frequency fan. Determine the current state vector , the current state vector Input a Q-network, which is designed for several possible frequency variations. Calculate the corresponding Q value Select the frequency change corresponding to the highest Q value Calculate the corrected target frequency ;

[0018] Based on the corrected target frequency Control the operation of the variable frequency fan.

[0019] The variable frequency fan control method described above updates the Q-network periodically based on historical data. The updated Q-network is as follows:

[0020] ;

[0021] Among them, the reward function , and These are weighting coefficients. For learning rate, As a discount factor, ∈[0, 1], For target air volume / air pressure, This represents the maximum power of the fan.

[0022] The variable frequency fan control method described above includes the actual fan speed n(t) / frequency f(t), the actual fan output air volume Q(t) / air pressure P(t), the fan current, and the fan voltage; the environmental parameters include at least one of ambient temperature, ambient humidity, air pressure, and gas concentration.

[0023] In the variable frequency fan control method described above, the operating parameters include at least one of motor temperature and motor vibration signal.

[0024] A variable frequency fan control system, the control system comprising:

[0025] The parameter acquisition module is used to acquire relevant parameters of the variable frequency fan. The parameter acquisition module includes an operating parameter acquisition module and an environmental parameter acquisition module.

[0026] A neural network model for variable frequency fans is used to receive the relevant parameters and output the target frequency. ;

[0027] The neural network model for variable frequency fans includes:

[0028] Hidden layer 1: , This is a vector of relevant parameters for the variable frequency fan; This is the weight matrix. This is the deviation term;

[0029] Hidden layer 2: , This is the weight matrix. This is the deviation term;

[0030] Output layer: , This is the weight matrix. This is the deviation term;

[0031] The weight matrix and bias term vector are obtained by minimizing the loss function. Obtained through training. It is the predicted target frequency of the model for the i-th sample. It is the actual target frequency of the i-th sample;

[0032] The control module, based on the target frequency Control the operation of the variable frequency fan.

[0033] The variable frequency fan control system described above includes:

[0034] The module for acquiring the actual output air volume / pressure of a variable frequency fan is used to obtain the actual output air volume of the variable frequency fan. / wind pressure ;

[0035] The load power determination module is used to determine the load power of the variable frequency fan. ;

[0036] Current state vector The module determines the output target frequency. Actual output air volume Q(t) / air pressure P(t) of variable frequency fan, load power Determine the current state vector with relevant parameters. ;

[0037] The Q network receives the current state vector. For several possible frequency variations Calculate the corresponding Q value Select the frequency change corresponding to the highest Q value ;

[0038] The target frequency correction module calculates the corrected target frequency. ;

[0039] The control module, based on the corrected target frequency Control the variable frequency fan.

[0040] The variable frequency fan control system described above includes a Q-network timed update module, which updates the Q-network as follows:

[0041] ;

[0042] Among them, the reward function , and These are weighting coefficients. For learning rate, As a discount factor, ∈[0, 1], For target air volume / air pressure, This represents the maximum power of the fan.

[0043] As described above, the variable frequency fan control system includes a module for acquiring the actual fan speed n(t) / frequency f(t), a module for acquiring the actual fan output air volume Q(t) / air pressure P(t), and modules for acquiring the fan current and fan voltage.

[0044] The environmental parameter acquisition module includes at least one of an environmental temperature acquisition module, an environmental humidity acquisition module, an air pressure acquisition module, and a gas concentration acquisition module.

[0045] In the variable frequency fan control system described above, the operating parameter acquisition module includes at least one of a motor temperature detection module and a motor vibration detection module.

[0046] Compared with existing technologies, the advantages and positive effects of this invention are as follows: The variable frequency fan control method includes: acquiring relevant parameters of the variable frequency fan, including operating parameters and environmental parameters; inputting the relevant parameters of the variable frequency fan into a variable frequency fan neural network model and outputting a target frequency; controlling the operation of the variable frequency fan according to the target frequency. The variable frequency fan neural network model includes a hidden layer 1, a hidden layer 2, and an output layer. The weight matrix and bias term vector of each layer are optimized by minimizing the loss function. Hidden layer 1 introduces nonlinearity, enabling the neural network to learn complex patterns. Hidden layer 2 controls the response range of neurons and provides a smoother signal for the subsequent output layer. The output layer outputs the target frequency best suited to the current conditions. This invention realizes a complex nonlinear mapping relationship from the current input state to a single optimal target frequency. The target frequency can adapt to changing condition parameters, improving the adaptability of the fan.

[0047] The variable frequency fan control system includes: a parameter acquisition module for acquiring relevant parameters of the variable frequency fan; a variable frequency fan neural network model for receiving relevant parameters and outputting a target frequency; and a control module for controlling the operation of the variable frequency fan according to the target frequency. The variable frequency fan neural network model includes hidden layer 1, hidden layer 2, and an output layer. The weight matrix and bias term vector of each layer are optimized by minimizing the loss function. Hidden layer 1 introduces nonlinearity, enabling the neural network to learn complex patterns. Hidden layer 2 controls the response range of neurons and provides a smoother signal for the subsequent output layer. The output layer outputs the target frequency best suited to the current conditions. This invention realizes a complex nonlinear mapping relationship from the current input state to a single optimal target frequency. The target frequency can adapt to varying condition parameters, improving the adaptability of the fan.

[0048] Other features and advantages of the present invention will become clearer after reading the detailed embodiments of the invention in conjunction with the accompanying drawings. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a flowchart of a specific embodiment of the present invention.

[0051] Figure 2 This is a flowchart of another specific embodiment of the present invention.

[0052] Figure 3 This is a principle block diagram of a specific embodiment of the present invention.

[0053] Figure 4 This is a principle block diagram of another specific embodiment of the present invention. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] In the description of this invention, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0056] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "joining" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. In the description of embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0057] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature.

[0058] In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0059] Artificial intelligence models are introduced into the control of wind turbines to build an intelligent control system that can sense environmental parameters in real time, automatically optimize control strategies, and dynamically adjust airflow output. This not only significantly improves wind turbine operating efficiency but also reduces energy consumption, extends equipment life, and adapts to various complex operating conditions.

[0060] A variable frequency fan control method includes:

[0061] Obtain relevant parameters of the variable frequency fan, including operating parameters and environmental parameters;

[0062] Input the relevant parameters of the variable frequency fan into the neural network model of the variable frequency fan, and output the target frequency. ;

[0063] According to the target frequency Control the operation of the variable frequency fan;

[0064] The neural network model for variable frequency fans includes:

[0065] Hidden layer 1: , This is a vector of relevant parameters for the variable frequency fan; This is the weight matrix. The vector of deviation terms;

[0066] Hidden layer 2: , This is the weight matrix. The vector of deviation terms;

[0067] Output layer: , This is the weight matrix. The vector of deviation terms;

[0068] The weight matrix and bias term vector are obtained by minimizing the loss function. Obtained through training. It is the predicted target frequency of the model for the i-th sample. It is the actual target frequency of the i-th sample.

[0069] The relationship between the output air volume Q(t) and the rotational speed n(t) is as follows: .

[0070] The relationship between output air pressure P(t) and rotational speed n(t) is as follows: .

[0071] Where k1 and k2 are proportionality coefficients.

[0072] The relationship between rotational speed and frequency is as follows: .

[0073] Where f(t): frequency (Hz), p: number of motor poles;

[0074] The relationship between air volume and frequency: Substituting the rotational speed into the air volume formula, we get... ;

[0075] Relationship between wind pressure and frequency: .

[0076] In some embodiments, the operating parameters include the actual fan speed n(t) / frequency f(t), the actual fan output air volume Q(t) / air pressure P(t), the fan current, and the fan voltage.

[0077] The actual fan speed n(t) / frequency f(t) is the basic feedback quantity for achieving closed-loop control.

[0078] The actual output air volume Q(t) / air pressure P(t) of the fan is the direct control target of the system and the fundamental basis for evaluating the control effect.

[0079] One of the following is essential for realizing the basic variable frequency closed-loop control of the fan: the actual fan speed n(t) / frequency f(t) and the actual fan output air volume Q(t) / air pressure P(t). They constitute direct feedback of the system state.

[0080] Air volume control - suitable for scenarios with direct requirements for ventilation (such as farms and workshop fresh air systems).

[0081] Pressure control - suitable for scenarios where a specific pressure needs to be maintained within the duct system (such as cleanrooms and building ventilation).

[0082] Fan current and voltage are used to calculate fan power, enabling energy-saving optimization and equipment status monitoring. This optimizes load matching, prevents overload, and extends equipment life. Abnormal current can indicate fan overload, stall, or mechanical failure, improving system reliability.

[0083] Environmental parameters include at least one of ambient temperature, ambient humidity, air pressure, and gas concentration.

[0084] Ambient temperature can improve environmental temperature adaptability, enabling the fan adjustment to respond to temperature changes.

[0085] Ambient humidity can improve the adaptability to ambient humidity, enabling the fan to adjust in response to humidity changes.

[0086] In high-humidity environments (such as textile workshops and greenhouses), humidity is a key factor affecting perceived temperature and ventilation needs, enabling airflow adjustment to better meet actual comfort requirements.

[0087] Air pressure can improve the adaptability to environmental air pressure, enabling the fan to adjust in response to changes in air pressure.

[0088] Gas concentration can improve the adaptability to environmental gas concentration, enabling the fan to adjust in response to changes in air pressure.

[0089] In densely populated areas (such as offices and shopping malls), CO2 concentration is a direct indicator of air quality. The system can dynamically adjust the airflow based on this concentration to achieve energy savings (on-demand ventilation) while ensuring air quality.

[0090] Other harmful gas / dust concentrations: For specific industrial environments (such as chemical plants and welding workshops), the introduction of appropriate sensors can enable the control system to directly respond to environmental safety requirements.

[0091] In some embodiments, the operating parameters include at least one of motor temperature and motor vibration signal.

[0092] Motor temperature and vibration signals are used to enable predictive maintenance, improving system reliability and safety.

[0093] The motor temperature can be either the temperature of the fan body or the temperature of the motor bearings, which can prevent the motor from overheating and burning out.

[0094] The goal of a neural network model for variable frequency wind turbines is to find the mapping relationship between relevant parameters and the operating frequency of the wind turbine by learning from historical data. By training the model and utilizing the experience gained from historical data—that is, under corresponding historical environmental conditions—the target frequency can be predicted. Approximate to the actual target frequency that can meet air volume requirements and save energy. Actual target frequency It is determined based on the information from the i-th sample. For example, one of the values ​​within the target frequency setting range that can meet the minimum airflow requirement in the i-th sample is used as the actual target frequency. .

[0095] In real-time control, when relevant parameters are detected, the variable frequency fan neural network model can quickly predict the optimal target frequency of the fan and control the fan frequency to reach the optimal target frequency, thereby optimizing environmental conditions.

[0096] The training process for a neural network model involves preprocessing historical data, including denoising and normalization.

[0097] Historical data includes relevant parameters of the wind turbine at several points in time t, including operating parameters and environmental parameters.

[0098] Data preprocessing involves cleaning and standardizing the collected historical data to remove noise and make it suitable for model input.

[0099] Noise reduction: Use digital filters (such as moving average filtering or Kalman filtering) to remove high-frequency noise.

[0100] Normalization: Normalize all signals to the interval [0,1] in order to train the neural network model.

[0101] Deep neural networks (DNNs) include the following layers:

[0102] Input layer: Receives the normalized input signal ;

[0103] Hidden layer 1: Fully connected layer, with ReLU activation function; ;

[0104] Formula Explanation: This formula calculates the output of each neuron in hidden layer 1. First, the input signal x(t) and the weight matrix... Multiply, and add the deviation term This process yields intermediate results. Then, the ReLU activation function (Rectified LinearUnit) is applied, taking the larger of the maximum value 0 and the intermediate result. This step introduces non-linearity into the neural network, enabling it to learn complex patterns.

[0105] Hidden layer 2: Fully connected layer, with Sigmoid activation function; ;

[0106] Formula explanation: First, the output of hidden layer 1 (t) and weight matrix Multiply, and add the deviation term. .

[0107] Next, the sigmoid function is applied, which compresses the input values ​​to between 0 and 1. This helps control the range of neuron responses and provides a smoother signal for subsequent output layers, contributing to network stability.

[0108] Output layer: Linear output layer, directly outputs the target frequency. (t), .

[0109] Formula explanation: It is the weight matrix connecting hidden layer 2 and the output layer. It is the deviation term of the output layer. The output value represents the system's recommended operating frequency for the exhaust fan at time t, in order to achieve effective environmental control. The output value represents the recommended operating speed of the exhaust fan at time t, in order to achieve effective environmental control.

[0110] It is a "high-level feature representation" that has been refined through layers of the network, and it condenses the essence of all input information.

[0111] The core function of the weight matrix w3 and the bias b3 is to learn how to "translate" these high-level features into a specific, continuous control value, namely the target frequency.

[0112] Because the frequency is a continuous value, the output layer uses a linear activation function (i.e., no activation function is added) and outputs directly.

[0113] The loss function is the mean squared error (MSE):

[0114] ;

[0115] Formula explanation: It is the predicted target frequency of the model for the i-th sample; It is the actual target frequency (true value) of the i-th sample. This refers to the amount of training data. The loss function quantifies the difference between the model's predictions and the true values. By minimizing the loss function, the optimization algorithm can adjust the model parameters to make the predictions closer to the true values. Performance evaluation: The smaller the loss function value, the closer the model's predictions are to the true values, and the better the model's performance.

[0116] After training the model, the weight matrix and the bias term vector are obtained, thus obtaining the neural network model of the variable frequency fan.

[0117] When the variable frequency fan is running, obtain its real-time relevant parameters, such as ambient temperature. The real-time relevant parameters, including fan current I(t), fan voltage U(t), actual output air volume Q(t), and actual fan frequency f(t), are preprocessed and combined into an input vector x(t) = [ [(t), I(t), U(t), Q(t), f(t)], input vector x(t) is input into the variable frequency wind turbine neural network model, and the output is the target frequency. (t), based on the target frequency (t) Generates control signals for the frequency converter, which changes the fan speed by adjusting the output voltage and frequency.

[0118] Fan PWM signal generation: The duty cycle D of the PWM (Pulse Width Modulation) signal determines the output power of the frequency converter. ;

[0119] in, This is the maximum operating frequency of the fan.

[0120] Inverter control: The inverter adjusts the output voltage and frequency according to the PWM signal, thereby changing the fan speed and controlling the air volume and pressure.

[0121] like Figure 1 As shown, the control method for a variable frequency fan includes the following steps:

[0122] S1, Begin.

[0123] S2. Obtain the relevant parameters of the variable frequency fan.

[0124] The relevant parameters include operating parameters and environmental parameters.

[0125] The operating parameters should include at least the actual fan speed n(t) / frequency f(t), the actual fan output air volume Q(t) / air pressure P(t), the fan current, and the fan voltage.

[0126] Environmental parameters include at least one of ambient temperature, ambient humidity, air pressure, and gas concentration.

[0127] S3. Input the relevant parameters into the variable frequency fan neural network model after preprocessing.

[0128] S4, the target frequency is output by the neural network model of the variable frequency fan.

[0129] S5. Generate a fan control signal based on the target frequency to control the fan.

[0130] In some embodiments, a reinforcement learning model with air volume demand and energy consumption as the core reward function is designed to enable the system to adapt to various needs such as ventilation, air exchange, and pressure regulation, thereby improving the universality of the solution.

[0131] The target frequency is corrected through reinforcement learning to obtain the actual output air volume of the variable frequency fan. / wind pressure Determine the load power of the variable frequency fan. Determine the current state vector , the current state vector Input a Q-network, which is designed for several possible frequency variations. Calculate the corresponding Q value Select the frequency change corresponding to the highest Q value Calculate the corrected target frequency ;

[0132] Based on the corrected target frequency Control the operation of the variable frequency fan.

[0133] Action selection: An ε-greedy strategy is used to ultimately determine which frequency change Δf to take: Exploitation is employed – the action with the highest current Q value is selected with a high probability (1-ε), i.e., Δf = argmax_aQ( This is about making the optimal decision based on existing knowledge. Exploration – randomly selecting an action Δf with a small probability ε. This is to explore unknown possibilities and avoid getting trapped in local optima.

[0134] The Q network is updated periodically based on historical data. The updated Q network is as follows:

[0135] ;

[0136] The target Q value is derived from the actual reward obtained and the best estimate of the next state, and is a more reliable "true value" estimate.

[0137] It is the temporal difference error, which represents the difference between the predicted value and the target value of the Q network.

[0138] Updating the Q network means adjusting the current Q value a small step toward the "target Q value" (controlled by the learning rate α) to gradually reduce the prediction error.

[0139] Among them, the reward function , and These are weighting coefficients. For learning rate, As a discount factor, ∈[0, 1], For target air volume / air pressure, This represents the maximum power of the fan.

[0140] That is, the reward function .

[0141] The current air volume was measured. With expected air volume The degree of similarity. Measured current power Maximum power The proportion.

[0142] Alternatively, reward function .

[0143] The current wind pressure P was measured. With expected wind pressure The degree of similarity.

[0144] R(t+1): In the state vector Execution frequency variation Then, an immediate reward is given based on environmental feedback.

[0145] Changes in execution frequency After that, the system transitions to the next state.

[0146] : represents the frequency change at time t+1.

[0147] like Figure 2 As shown, the control method for a variable frequency fan includes the following steps:

[0148] S1, Begin.

[0149] S2. Obtain the relevant parameters of the variable frequency fan.

[0150] The relevant parameters include operating parameters and environmental parameters.

[0151] The operating parameters should include at least the actual fan speed n(t) / frequency f(t), the actual fan output air volume Q(t) / air pressure P(t), the fan current, and the fan voltage.

[0152] Environmental parameters include at least one of ambient temperature, ambient humidity, air pressure, and gas concentration.

[0153] S3. Input the relevant parameters into the variable frequency fan neural network model after preprocessing.

[0154] S4, the target frequency is output by the neural network model of the variable frequency fan.

[0155] S5. Correct the output target frequency using a Q network to obtain the corrected target frequency.

[0156] S6. Generate a fan control signal based on the corrected target frequency to control the fan.

[0157] like Figure 3 As shown, this embodiment also proposes a variable frequency fan control system, including:

[0158] The parameter acquisition module is used to acquire relevant parameters of the variable frequency fan. The parameter acquisition module includes an operating parameter acquisition module and an environmental parameter acquisition module.

[0159] A neural network model for variable frequency fans is used to receive relevant parameters and output the target frequency. ;

[0160] The neural network model for variable frequency fans includes:

[0161] Hidden layer 1: , This is a vector of relevant parameters for the variable frequency fan; This is the weight matrix. This is the deviation term;

[0162] Hidden layer 2: , This is the weight matrix. This is the deviation term;

[0163] Output layer: , This is the weight matrix. This is the deviation term;

[0164] The weight matrix and bias term vector are obtained by minimizing the loss function. Obtained through training. It is the predicted target frequency of the model for the i-th sample. It is the actual target frequency of the i-th sample;

[0165] The control module, based on the target frequency Control the operation of the variable frequency fan.

[0166] like Figure 4 As shown, the variable frequency fan control system also includes:

[0167] The module for acquiring the actual output air volume / pressure of a variable frequency fan is used to obtain the actual output air volume of the variable frequency fan. / wind pressure ;

[0168] The load power determination module is used to determine the load power of the variable frequency fan. ;

[0169] Current state vector The module determines the output target frequency. Actual output air volume Q(t) / air pressure P(t) of variable frequency fan, load power Determine the current state vector with relevant parameters. ;

[0170] The Q network receives the current state vector. For several possible frequency variations Calculate the corresponding Q value Select the frequency change corresponding to the highest Q value ;

[0171] The target frequency correction module calculates the corrected target frequency. ;

[0172] The control module, based on the corrected target frequency Control the variable frequency fan.

[0173] The system includes a Q network timed update module, which updates the Q network as follows:

[0174] ;

[0175] Among them, the reward function , and These are weighting coefficients. For learning rate, As a discount factor, ∈[0, 1], For target air volume / air pressure, This represents the maximum power of the fan.

[0176] In some embodiments, the operating parameter acquisition module includes a fan actual speed n(t) / frequency f(t) acquisition module, a fan actual output air volume Q(t) / air pressure P(t) acquisition module, and a fan current and fan voltage acquisition module;

[0177] The environmental parameter acquisition module includes at least one of the following: an environmental temperature acquisition module, an environmental humidity acquisition module, an air pressure acquisition module, and a gas concentration acquisition module.

[0178] In some embodiments, the operating parameter acquisition module includes at least one of a motor temperature detection module and a motor vibration detection module.

[0179] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions claimed by the present invention.

Claims

1. A variable frequency fan control method, characterized in that, The control method includes: Obtain relevant parameters of the variable frequency fan, including operating parameters and environmental parameters; Input the relevant parameters of the variable frequency fan into the neural network model of the variable frequency fan, and output the target frequency. ; According to the target frequency Control the operation of the variable frequency fan; The neural network model for variable frequency fans includes: Hidden layer 1: , This is a vector of relevant parameters for the variable frequency fan; This is the weight matrix. The vector of deviation terms; Hidden layer 2: , This is the weight matrix. The vector of deviation terms; Output layer: , This is the weight matrix. The vector of deviation terms; The weight matrix and bias term vector are obtained by minimizing the loss function. Obtained through training. It is the predicted target frequency of the model for the i-th sample. It is the actual target frequency of the i-th sample.

2. The variable frequency fan control method according to claim 1, characterized in that, Obtain the actual output air volume of the variable frequency fan / wind pressure Determine the load power of the variable frequency fan. Determine the current state vector , the current state vector Input a Q-network, which is designed for several possible frequency variations. Calculate the corresponding Q value Select the frequency change corresponding to the highest Q value Calculate the corrected target frequency ; Based on the corrected target frequency Control the operation of the variable frequency fan.

3. The variable frequency fan control method according to claim 2, characterized in that, The Q-network is updated periodically based on historical data. The updated Q-network is as follows: ； Among them, the reward function , and These are weighting coefficients. For learning rate, As a discount factor, ∈[0, 1], For target air volume / air pressure, This represents the maximum power of the fan.

4. The variable frequency fan control method according to claim 1, characterized in that, The operating parameters include the actual fan speed n(t) / frequency f(t), the actual fan output air volume Q(t) / air pressure P(t), the fan current, and the fan voltage; the environmental parameters include at least one of ambient temperature, ambient humidity, air pressure, and gas concentration.

5. The variable frequency fan control method according to claim 1, characterized in that, The operating parameters include at least one of motor temperature and motor vibration signal.

6. A variable frequency fan control system, characterized in that, The control system includes: The parameter acquisition module is used to acquire relevant parameters of the variable frequency fan. The parameter acquisition module includes an operating parameter acquisition module and an environmental parameter acquisition module. A neural network model for variable frequency fans is used to receive the relevant parameters and output the target frequency. ; The neural network model for variable frequency fans includes: Hidden layer 1: , This is a vector of relevant parameters for the variable frequency fan; This is the weight matrix. This is the deviation term; Hidden layer 2: , This is the weight matrix. This is the deviation term; Output layer: , This is the weight matrix. This is the deviation term; The weight matrix and bias term vector are obtained by minimizing the loss function. Obtained through training. It is the predicted target frequency of the model for the i-th sample. It is the actual target frequency of the i-th sample; The control module, based on the target frequency Control the operation of the variable frequency fan.

7. The variable frequency fan control system according to claim 6, characterized in that, The system includes: The module for acquiring the actual output air volume / pressure of a variable frequency fan is used to obtain the actual output air volume of the variable frequency fan. / wind pressure ; The load power determination module is used to determine the load power of the variable frequency fan. ; Current state vector The module determines the output target frequency. Actual output air volume Q(t) / air pressure P(t) of variable frequency fan, load power Determine the current state vector with relevant parameters. ; The Q network receives the current state vector. For several possible frequency variations Calculate the corresponding Q value Select the frequency change corresponding to the highest Q value ; The target frequency correction module calculates the corrected target frequency. ; The control module, based on the corrected target frequency Control the variable frequency fan.

8. The variable frequency fan control system according to claim 7, characterized in that, The system includes a Q network timed update module, which updates the Q network as follows: ； Among them, the reward function , and These are weighting coefficients. For learning rate, As a discount factor, ∈[0, 1], For target air volume / air pressure, This represents the maximum power of the fan.

9. The variable frequency fan control system according to claim 6, characterized in that, The operating parameter acquisition module includes a fan actual speed n(t) / frequency f(t) acquisition module, a fan actual output air volume Q(t) / air pressure P(t) acquisition module, and a fan current and fan voltage acquisition module; The environmental parameter acquisition module includes at least one of an environmental temperature acquisition module, an environmental humidity acquisition module, an air pressure acquisition module, and a gas concentration acquisition module.

10. The variable frequency fan control system according to claim 9, characterized in that, The operating parameter acquisition module includes at least one of a motor temperature detection module and a motor vibration detection module.

Images