Constant temperature and humidity chamber and artificial intelligence control system

Abstract

The application relates to a constant-temperature and constant-humidity box and an artificial intelligence control system, and belongs to the technical field of automatic control. In the artificial intelligence control system, a first subtracter generates a sequence first error signal according to a measured temperature provided by a temperature sensor and a set temperature, and a second subtracter generates a sequence second error signal according to a measured humidity provided by a humidity sensor and a set humidity; a deep reinforcement learning module generates a first state according to the sequence first error signal, generates a second state according to the sequence second error signal, respectively generates a control strategy of a first and second PID controller according to the first state and the second state, and generates a control strategy of an electronic change-over switch according to the first error signal; the first PID controller controls the working state of a variable-frequency compressor refrigeration, semiconductor refrigeration or heating subsystem selected through the electronic change-over switch; and the second PID controller controls the working state of a humidification subsystem. The application does not need to pre-learn the prior setting of the PID control parameter.

Description

TECHNICAL FIELD

[0001] The present application relates to a constant temperature and humidity chamber and an artificial intelligence control system, and belongs to the technical field of automatic control. BACKGROUND

[0002] Chinese patent application with publication number CN120371049A discloses a control method for a precision constant temperature test chamber. The method uses an advanced PID controller through a temperature control module and combines temperature sensor data for closed-loop adjustment. The specific implementation includes: acquiring temperature data at different positions in the test chamber in real time through a high-precision temperature sensor and transmitting the data to a central controller; the central controller calculates the output power required for heating or cooling according to the set target temperature and temperature change curve using the PID formula , t -T c , 、 、 , t , c , , , . SUMMARY

[0003] To overcome the shortcomings of the prior art, the present application provides a constant temperature and humidity chamber and an artificial intelligence control system, which has the advantages of simple structure and does not require prior learning of PID parameter settings.

[0004] To achieve the object of the present application, the present application provides a constant temperature and humidity chamber, which comprises: a chamber body, a variable frequency compressor refrigeration subsystem, a semiconductor refrigeration subsystem, a heating subsystem, a humidification subsystem, and an artificial intelligence control system. Temperature and humidity sensors are arranged in the chamber body. The artificial intelligence control system comprises a deep reinforcement learning module, first and second PID controllers, first and second subtractors, and an electronic switch. The first subtractor generates a first error signal based on the measured temperature provided by the temperature sensor and the set temperature. The deep reinforcement learning module generates a first state ; The second subtractor generates a sequence second error signal according to the measured humidity provided by the humidity sensor and the set humidity, and the deep reinforcement learning module generates a second state according to the first state and the second state to generate a control strategy of the first PID controller , , are respectively proportional, integral and differential coefficients of the first PID controller, and a control strategy of the second PID controller is generated ; , are respectively proportional, integral and differential coefficients of the second PID controller; and a control strategy of the electronic switching switch is generated according to the first error signal , B t is the switching action; the first PID controller selects the variable frequency compressor refrigeration subsystem, the semiconductor refrigeration subsystem or the heating subsystem through the electronic switching switch to further control the working state of the selected subsystem; and the second PID controller controls the working state of the humidification subsystem.

[0005] To achieve the object of the application, the application further provides an artificial intelligence control system of a constant temperature and humidity box, which comprises a deep reinforcement learning module, first and second PID controllers, first and second subtractors and an electronic switching switch, the first subtractor generates a sequence first error signal according to the measured temperature provided by the temperature sensor and the set temperature, the deep reinforcement learning module generates a first state according to the sequence first error signal; the second subtractor generates a sequence second error signal according to the measured humidity provided by the humidity sensor and the set humidity, and the deep reinforcement learning module generates a second state according to the first state and the second state to generate a control strategy of the first PID controller , , are respectively proportional, integral and differential coefficients of the first PID controller, and a control strategy of the second PID controller is generated ; , are respectively proportional, integral and differential coefficients of the second PID controller; and a control strategy of the electronic switching switch is generated according to the first error signal , B tThe switch is used for switching; the first PID controller selects the variable frequency compressor refrigeration subsystem, the semiconductor refrigeration subsystem or the heating subsystem through the electronic switch to further control the working state of the selected subsystem; and the second PID controller controls the working state of the humidification subsystem.

[0006] Compared with the prior art, the constant temperature and humidity box and the artificial intelligence control system have the following beneficial effects: (1) the first state is formed according to the difference between the measured temperature and the set temperature, the second state is generated according to the difference between the measured humidity and the set humidity, the proportional, integral and differential coefficients of the first PID controller for controlling the variable frequency compressor refrigeration subsystem, the semiconductor refrigeration subsystem or the heating subsystem are generated according to the first state and the second state to further control the working state of the selected subsystem, the proportional, integral and differential coefficients of the second PID controller for controlling the humidification subsystem are generated according to the first state and the second state to further control the working state of the humidification subsystem, and the control strategy of the electronic switch is generated according to the first error signal, so that the beneficial effects of simple structure and no need for prior setting of pre-learning PID parameters are achieved; (2) the control strategy of the electronic switch is generated according to the first error signal by the deep reinforcement learning module, if the set temperature is less than the measured temperature and the absolute value of the difference is less than or equal to the refrigeration threshold of the semiconductor refrigeration subsystem, the first PID controller is connected to the semiconductor refrigeration subsystem through the switch, and the semiconductor refrigeration subsystem provides cold energy to the inner container of the box; and when the absolute value of the difference is greater than the refrigeration threshold of the semiconductor refrigeration subsystem, the first PID controller is connected to the variable frequency compressor refrigeration subsystem through the electronic switch, and the variable frequency compressor refrigeration subsystem provides cold energy to the inner container of the box to reduce the temperature of the inner container of the box, so that the advantages of large compressor refrigeration capacity and adjustable semiconductor refrigeration capacity to 0 are fully utilized, and the beneficial effect of flexible adjustment between 0 refrigeration capacity and maximum refrigeration capacity is achieved. BRIEF DESCRIPTION OF DRAWINGS

[0007] Figure 1 is a schematic diagram of the constant temperature and humidity box provided by the first embodiment of the present application.

[0008] Figure 2 is a schematic diagram of the variable frequency compressor refrigeration subsystem provided by the first embodiment of the present application.

[0009] Figure 3 is a schematic diagram of the semiconductor refrigeration subsystem provided by the first embodiment of the present application.

[0010] Figure 4 is a block diagram of the artificial intelligence control system of the constant temperature and humidity box provided by the first embodiment of the present application. DETAILED DESCRIPTION

[0011] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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.

[0012] First Embodiment

[0013] Figure 1 This is a schematic diagram of the composition of the constant temperature and humidity chamber provided in the first embodiment of the present invention, as shown below. Figure 1 As shown, the constant temperature and humidity chamber includes an upper chamber 1, a lower chamber 16, a variable frequency compressor refrigeration subsystem, a semiconductor refrigeration subsystem, a heating subsystem, and a humidification subsystem. The upper chamber contains an inner liner 2, which is equipped with an air duct plate 5 that divides the inner liner into a working chamber 3 and a mixing chamber 4. The mixing chamber contains an isolation plate 6 that divides the mixing chamber into upper and lower chambers. The upper chamber contains a fan blade, which is driven by a fan 7. The fan shaft is inserted into a pre-drilled hole from the outer side of the rear wall of the inner liner. The fan is fixed to the outer side of the rear wall of the inner liner with M4×12 Phillips head screws, and the fan blades are installed on the inner fan shaft. The isolation plate is installed on the front side of the fan with M4×12 Phillips head screws.

[0014] The lower chamber houses the evaporator 10 of the variable frequency compressor refrigeration subsystem, the heating subsystem, and the humidification subsystem. A temperature sensor and a humidity sensor (not shown in the figure) are installed inside the chamber. The temperature sensor is used to obtain the measured temperature inside the inner liner, and the humidity sensor is used to obtain the measured humidity inside the inner liner. The temperature sensor and humidity sensor are respectively fixed to the isolation plate with plastic brackets.

[0015] The heating subsystem includes a heating element 8, which is used to heat the inner liner. The heating element 8 is clipped onto a heating element bracket 9; the heating element bracket is welded to the rear wall of the inner liner, and the heating element is clipped onto the heating element bracket.

[0016] The humidification subsystem includes a humidification heating element 12, which heats the water in the water tank 13 to convert the water into steam. The humidification heating element is mounted on a humidification heating element bracket 11, which is welded to the rear wall of the inner tank. The water tank and the inner tank are fully welded together.

[0017] The lower housing also houses a variable frequency compressor 14 and a condenser 15.

[0018] Figure 2 This is a schematic diagram of the composition of the variable frequency compressor refrigeration subsystem provided in the first embodiment of the present invention; as shown below. Figure 2As shown in the figure, the variable frequency compressor refrigeration subsystem includes an evaporator 10, a variable frequency compressor 14, a condenser 15 and a capillary tube, wherein the variable frequency compressor and the condenser are fixed on the bottom plate of the lower box body by M6*30 cylinder head hexagon screws, the evaporator is fixed on the rear wall of the inner container by M4*12 cross slot pan head screws, and the variable frequency compressor and the condenser are installed on the bottom fixing plate; the variable frequency compressor is driven by a variable frequency driver 18, the exhaust port of the variable frequency compressor is connected to the condenser, and the outlet of the condenser is connected to the capillary tube. The capillary tube is connected to the inlet of the evaporator, and the outlet of the evaporator is connected to the return air port of the variable frequency compressor.

[0019] Figure 3 is the composition schematic diagram of the semiconductor refrigeration subsystem provided by the first embodiment of the application; as Figure 3 shown, the semiconductor refrigeration subsystem includes a semiconductor sheet 22, a semiconductor cold end heat sink 21, a semiconductor hot end heat sink 23 and a semiconductor hot end heat sink fan 24, wherein the semiconductor hot end heat sink fan is fixed on the semiconductor hot end heat sink by screws, and the semiconductor cold end heat sink and the semiconductor hot end heat sink sandwich the semiconductor sheet, and the gap is filled with polyurethane foaming agent; the semiconductor cold end heat sink is arranged on the inner side of the inner container, and the semiconductor cold end heat sink is arranged on the outer side of the inner container.

[0020] In the first embodiment, the humidifying box further includes an artificial intelligence control system, which will be described in detail below. Figure 4 .

[0021] Figure 4 is the artificial intelligence control system provided by the first embodiment of the application, as Figure 4 shown, the artificial intelligence control system includes a deep reinforcement learning module, a first PID controller, a second PID controller, a first subtractor 31, a second subtractor 32 and an electronic switch, the first subtractor generates a sequence first error signal according to the measured temperature provided by the temperature sensor and the set temperature, the deep reinforcement learning module generates a first state according to the sequence first error signal; the second subtractor generates a sequence second error signal according to the measured humidity provided by the humidity sensor and the set humidity, the deep reinforcement learning module generates a second state according to the sequence second error signal, generates a control strategy of the first PID controller according to the first state and the second state , , are the proportional, integral and differential coefficients of the first PID controller respectively, and generates a control strategy of the second PID controller; , are the proportional, integral and differential coefficients of the second PID controller; the first error signal Control strategy of electronic switch The first PID controller selects the variable frequency compressor refrigeration subsystem, the semiconductor refrigeration subsystem or the heating subsystem through the electronic switch to further control the working state of the selected subsystem, mainly the working state of the variable frequency drive of the variable frequency compressor refrigeration subsystem, the semiconductor chip of the semiconductor refrigeration subsystem or the heating electric tube of the heating subsystem; the second PID controller controls the working state of the humidification subsystem, mainly the working state of the humidification electric tube. The variable frequency drive of the variable frequency compressor refrigeration subsystem, the semiconductor chip of the semiconductor refrigeration subsystem or the heater of the heating subsystem are connected to the output end of the adder 33. The adder can also be a connection node. For switch action, the static point of the electronic switch is connected to the output end of the first PID controller, and the dynamic point is connected to the heating subsystem, the variable frequency compressor refrigeration subsystem or the semiconductor refrigeration subsystem.

[0022] In the first embodiment, under the ideal condition that the temperature and humidity in the tank do not affect each other, the output of the first PID controller at time t is The error between the set temperature and the measured temperature of the tank at time t is Therefore, we have:

[0023] ,

[0024] In the formula, , ; ;

[0025] Written in matrix form:

[0026] ,

[0027] In the formula, , .

[0028] The output of the second PID controller at time t is The error between the set humidity and the measured humidity of the tank at time t is Therefore, we have:

[0029] ,

[0030] In the formula, , ; ;

[0031] Written in matrix form:

[0032] ,

[0033] In the formula, , .

[0034] However, humidification and heating or cooling are mutually influenced, so in the first embodiment, the control strategy of the working parameters of the first PID controller for controlling the working states of the variable frequency compressor of the variable frequency compressor refrigeration subsystem, the semiconductor sheet of the semiconductor refrigeration subsystem, and the electric heating tube of the heating subsystem is:

[0035] ,

[0036] In the formula, is the proportional coefficient of the first PID controller, is the integral coefficient of the first PID controller; is the differential coefficient of the first PID controller.

[0037] The control strategy of the working parameters of the second PID controller for controlling the humidification electric heating tube of the humidification subsystem is:

[0038] ,

[0039] In the formula, is the proportional coefficient of the second PID controller, is the integral coefficient of the second PID controller; is the differential coefficient of the second PID controller.

[0040] The neural network of the deep reinforcement learning module includes an input layer, a hidden layer, and an output layer, wherein the input layer includes 6 neurons, the hidden layer includes 6 neurons, and the output layer includes 8 neurons. The first to sixth neurons of the output layer respectively output the proportional coefficient, the integral coefficient, and the differential coefficient of the first and second PID controllers; the seventh and eighth neurons respectively output the temperature state value function and the humidity state value function.

[0041] In the first embodiment, the deep reinforcement learning module generates the control strategy of the first PID controller and the control strategy of the second PID controller according to the first state and the second state , which includes:

[0042] S1-1: Determine the control coefficient of the KID controller according to the following formula:

[0043] ,

[0044] In the formula, are the center and width of the jth Gaussian function of the neural network of the deep reinforcement learning module; are the weights between the jth neuron of the hidden layer and the nth neuron of the output layer of the neural network of the deep reinforcement learning module; j = 1, …, 6; n = 1, …, 6 denotes the two-norm.

[0045] The deep reinforcement learning module generates a control policy of the first PID controller and a control policy of the second PID controller according to the first state and the second state The deep reinforcement learning module further comprises:

[0046] S1-2: The deep reinforcement learning module generates a temperature state value function and a humidity state value function according to the first state and the second state The deep reinforcement learning module further comprises:

[0047]

[0048]

[0049] wherein, are the weights between the jth neuron of the hidden layer and the 7th neuron of the output layer of the neural network of the deep reinforcement learning module, are the weights between the jth neuron of the hidden layer and the 8th neuron of the output layer of the neural network of the deep reinforcement learning module.

[0050] The deep reinforcement learning module generates a control policy of the first PID controller and a control policy of the second PID controller according to the first state and the second state The deep reinforcement learning module further comprises:

[0051] S1-3: The deep reinforcement learning module constructs a cost function according to the following formula

[0052]

[0053] wherein, is the first reward function, is the measured temperature, is the set temperature; is the second reward function, is the measured humidity,​​​​​​​ to set the humidity; temperature state value function at time t; temperature state value function at time t-1; humidity state value function at time t; humidity state value function at time t-1; , , , is a weight.

[0054] In the first embodiment, the parameters are updated according to the following formula:

[0055] , -,

[0056] ,

[0057] wherein, , , , is a learning coefficient; and are the weight before updating and the weight after updating, respectively; and are the weight before updating and the weight after updating, respectively; and are the center value before updating and the center value after updating of the jth Gaussian function, respectively; and are the bandwidth before updating and the bandwidth after updating of the jth Gaussian function, respectively; n = 1, …, 6; N = 7, 8.

[0058] In the first embodiment,

[0059] ,

[0060] ,

[0061] ,

[0062] ,

[0063] ,

[0064] ,

[0065] wherein, is the temperature error at time t; is the measured temperature of the inner container of the tank at time t; is the measured temperature of the inner container of the cabinet at time t-1; is the output of the first PID controller at time t; ; is the humidity error at time t; is the measured humidity of the inner container of the cabinet at time t; is the measured humidity of the inner container of the cabinet at time t-1; is the output of the second PID controller at time t; represents splicing.

[0066] In the first embodiment,

[0067] ,

[0068] ,

[0069] N=7,8.

[0070] The deep reinforcement learning module generates a control strategy of the first PID controller and a control strategy of the second PID controller according to the first state and the second state . Further comprising:

[0071] S1-4: judging whether the cost function is minimum, if not, , , , , returning to step S1-1; if yes, output , , , as the optimal parameters to calculate , and assigning to the proportional control coefficient, the integral control coefficient and the differential control coefficient of the first PID controller respectively, and assigning to the proportional control coefficient, the integral control coefficient and the differential control coefficient of the second PID controller respectively.

[0072] In the present application, the control strategy of the electronic switch is generated according to the first error signal comprising: judging whether is positive or negative, i.e. if the set temperature is greater than the measured temperature, if is positive, then the first PID controller is connected to the heating subsystem through the switch, and the heating subsystem provides heat energy to the inner container of the cabinet to increase the temperature; if the set temperature is less than the measured temperature, then is negative, ​​If the temperature exceeds the cooling threshold of the semiconductor refrigeration subsystem, the first PID controller is connected to the variable frequency compressor refrigeration subsystem via a switch. The variable frequency compressor refrigeration subsystem then provides cooling energy to the inner liner of the cabinet to lower its temperature. Negative value If the temperature is less than or equal to the cooling threshold of the semiconductor refrigeration subsystem, the first PID controller is connected to the semiconductor refrigeration subsystem via a switch. The semiconductor refrigeration subsystem provides cooling energy to the inner liner of the cabinet to lower the temperature of the inner liner.

[0073] Compared with existing technologies, the artificial intelligence control system provided in the first embodiment has the following beneficial effects: It utilizes a deep reinforcement learning module to determine the temperature and humidity composition of the inner chamber of the enclosure. and ,according to and Control strategy for generating the first PID controller Control strategy of the second PID controller According to the first error signal Control strategy for generating electronic switching This invention achieves the advantage of automatically controlling the temperature and humidity of the cabinet without the need for pre-learning and prior setting of PID parameters. The invention uses a deep reinforcement learning module to generate a control strategy for an electronic switch based on the first error signal. If the set temperature is lower than the measured temperature, and the absolute value of the difference is less than or equal to the cooling threshold of the semiconductor refrigeration subsystem, then the first PID controller is connected to the semiconductor refrigeration subsystem via a switch, and the semiconductor refrigeration subsystem provides cooling energy to the inner liner of the cabinet. When the absolute value of the difference is greater than the cooling threshold of the semiconductor refrigeration subsystem, then the first PID controller is connected to the variable frequency compressor refrigeration subsystem via an electronic switch, and the variable frequency compressor refrigeration subsystem provides cooling energy to the inner liner of the cabinet to lower its temperature. This invention fully utilizes the advantages of the large cooling capacity of the compressor and the ability to adjust the semiconductor refrigeration capacity to zero, achieving the beneficial effect of flexible adjustment between zero cooling capacity and maximum cooling capacity.

[0074] Furthermore, although the parameters of the neural network in the deep reinforcement learning module are dynamically updated over time using gradient descent, the initial selection of these parameters is crucial for achieving the desired results.

[0075] Second Embodiment

[0076] The second embodiment of the present invention only describes the contents that are different from those of the first embodiment; the contents that are the same will not be described again.

[0077] The second embodiment of the present application provides a computer program product, which uses a computer language to compile the method provided by the first embodiment into a computer program, the computer program can be stored in a storage medium and called by one or more processors to implement the method of the plurality of steps in the first embodiment.

[0078] The disclosed preferred embodiments of the present application are only used to help explain the present application, the preferred embodiments do not describe all the details and limit the present application to the specific embodiments. Obviously, many modifications and variations can be made in light of the contents of the specification. The specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present application, so that those skilled in the art can well understand and utilize the present application. The present application is limited only by the claims and their full scope and equivalents.

Claims

1. A constant temperature and humidity chamber, characterized in that, include: The system comprises a cabinet, a variable frequency compressor refrigeration subsystem, a semiconductor refrigeration subsystem, a heating subsystem, a humidification subsystem, and an artificial intelligence control system. Temperature and humidity sensors are installed inside the cabinet. The artificial intelligence control system includes a deep reinforcement learning module, first and second PID controllers, first and second subtractors, and an electronic transfer switch. The first subtractor generates a first sequence error signal based on the measured temperature and the set temperature provided by the temperature sensor. The deep reinforcement learning module generates a first state based on the first sequence error signal. The second subtractor generates a second sequence error signal based on the measured humidity and the set humidity provided by the humidity sensor. The deep reinforcement learning module then generates a second state based on the second sequence error signal. According to the first state Second state Control strategy for generating the first PID controller Control strategy of the second PID controller According to the first error signal Control strategy for generating electronic switching B t For switching operations; the first PID controller selects the variable frequency compressor refrigeration subsystem, semiconductor refrigeration subsystem, or heating subsystem via an electronic transfer switch to further control the operating state of the selected subsystem; the second PID controller controls the operating state of the humidification subsystem, wherein, In the formula, , and These are the proportional, integral, and derivative coefficients of the first PID controller, respectively. In the formula, , and These are the proportional, integral, and derivative coefficients of the second PID controller, respectively. ， In the formula, , and These are the center and width of the j-th Gaussian function in the neural network of the deep reinforcement learning module; These are the weights between the j-th neuron in the hidden layer and the n-th neuron in the output layer of the deep reinforcement learning module's neural network; j=1,…,6; n=1,…,6. This represents the L2 norm.

2. The constant temperature and humidity chamber according to claim 1, characterized in that, The deep reinforcement learning module also considers the first state. Second state Generating temperature state value function Humidity state value function : ， , In the formula, These are the weights between the j-th neuron in the hidden layer and the 7th neuron in the output layer of a deep reinforcement learning neural network. It represents the weights between the j-th neuron in the hidden layer and the 8th neuron in the output layer of a deep reinforcement learning neural network.

3. The constant temperature and humidity chamber according to claim 2, characterized in that, Deep reinforcement learning models also construct a cost function based on the following formula. : ， In the formula, , For the first reward function, For actual measured temperature, To set the temperature; For the second reward function, For actual humidity measurement, To set the humidity; The temperature state value function at time t; This is a function of the temperature state value at time t-1; The humidity state value function at time t; The humidity state value function at time t-1; , , , As weight; Update the parameters according to the following formula: ， ， ， In the formula, , , , The learning coefficient; and These are the weights before and after the update, respectively; and These are the weights before and after the update, respectively; and These are the center values ​​of the j-th Gaussian function before and after the update, respectively; and Let $\mathbf{j}$ be the bandwidth before and after the update of the $j$-th Gaussian function, respectively; $n = 1, ..., 6$; $N = 7, 8$.

4. An artificial intelligence control system for a constant temperature and humidity chamber, characterized in that, The system includes a deep reinforcement learning module, first and second PID controllers, first and second subtractors, and an electronic transfer switch. The first subtractor generates a first sequence error signal based on the measured temperature provided by the temperature sensor and the set temperature. The deep reinforcement learning module generates a first state based on the first sequence error signal. The second subtractor generates a second sequence error signal based on the measured humidity and the set humidity provided by the humidity sensor. The deep reinforcement learning module then generates a second state based on the second sequence error signal. According to the first state Second state Control strategy for generating the first PID controller Control strategy of the second PID controller According to the first error signal Control strategy for generating electronic switching B t For switching operations; the first PID controller selects the variable frequency compressor refrigeration subsystem, semiconductor refrigeration subsystem, or heating subsystem via an electronic transfer switch to further control the operating state of the selected subsystem; the second PID controller controls the operating state of the humidification subsystem, wherein, In the formula, , and These are the proportional, integral, and derivative coefficients of the first PID controller, respectively. In the formula, , and These are the proportional, integral, and derivative coefficients of the second PID controller, respectively. ， In the formula, , and These are the center and width of the j-th Gaussian function in the neural network of the deep reinforcement learning module; These are the weights between the j-th neuron in the hidden layer and the n-th neuron in the output layer of the deep reinforcement learning module's neural network; j=1,…,6; n=1,…,6. This represents the L2 norm.

5. The artificial intelligence control system for the constant temperature and humidity chamber according to claim 4, characterized in that, Deep reinforcement learning models also rely on the first state Second state Generating temperature state value function Humidity state value function : ， , In the formula, These are the weights between the j-th neuron in the hidden layer and the 7th neuron in the output layer of a deep reinforcement learning neural network. It represents the weights between the j-th neuron in the hidden layer and the 8th neuron in the output layer of a deep reinforcement learning neural network.

6. The artificial intelligence control system for the constant temperature and humidity chamber according to claim 5, characterized in that, Deep reinforcement learning models also construct a cost function based on the following formula. : ， In the formula, , For the first reward function, For actual measured temperature, To set the temperature; For the second reward function, For actual humidity measurement, To set the humidity; The temperature state value function at time t; This is a function of the temperature state value at time t-1; The humidity state value function at time t; The humidity state value function at time t-1; , , , As weight; Update the parameters according to the following formula: ， ， ， In the formula, , , , The learning coefficient; and These are the weights before and after the update, respectively; and These are the weights before and after the update, respectively; and These are the center values ​​of the j-th Gaussian function before and after the update, respectively; and Let $\mathbf{j}$ be the bandwidth before and after the update of the $j$-th Gaussian function, respectively; $n = 1, ..., 6$; $N = 7, 8$.

Images