Indoor temperature dynamic matrix prediction control method of central air conditioner for hospital

Abstract

The invention relates to the technical field of central air conditioner control, and discloses an indoor temperature dynamic matrix prediction control method of a central air conditioner for a hospital, which comprises the following steps: step S101, constructing a feature matrix as sample data for training a first model; s102, acquiring the required temperature as a sample label for training the first model; s103, constructing a feature sequence as sample data for training a second model; s104, taking the wind speed corresponding to the minimum value of the comprehensive factors as a sample label for training a second model; step S105, repeating the steps S101 to S104 until the U first training samples and the U second training samples are obtained; according to the method, the temperature and the wind speed of the central air conditioner in the future time period are controlled by comprehensively considering the relevance among the sensor data of the data acquisition points through the first model and the second model, and the training data set is constructed by taking comfort maximization and energy consumption minimization as targets, so that the comfort of medical staff and patients is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of central air-conditioning control, and more specifically, to a method for dynamic matrix prediction control of indoor temperature of a central air-conditioning system for a hospital. Background Art

[0002] The central air conditioning system in a hospital is one of the key facilities to ensure a comfortable and hygienic medical environment. The traditional control method of central air conditioning in hospitals mainly relies on the user to manually set the temperature, wind speed and other parameters of the air conditioner. However, manual adjustment cannot respond to the dynamic changes of the indoor environment in real time, which may cause large fluctuations in temperature and humidity, affecting the comfort of medical staff and patients.

[0003] With the rapid development of artificial intelligence and the Internet of Things, modern hospital central air-conditioning systems are becoming increasingly intelligent. Currently, sensors are used to collect indoor temperature, humidity, and carbon dioxide concentration in real time, and the sensor data is transmitted to the central control system through wireless communication. The temperature of the air conditioner is automatically adjusted through the timing analysis model of the central control system, thereby reducing the energy consumption of the central air conditioner. However, the above schemes often ignore the uneven distribution of temperature and humidity in the central air-conditioning control area. Analyzing the sensor data of a single collection point alone may not be sufficient to meet the usage requirements of the entire control area. For example, different areas have different population densities and different temperature requirements in different departments and wards, resulting in low accuracy in temperature adjustment, which affects the comfort of medical staff and patients.

[0004] Therefore, a control method for central air conditioning in hospitals is needed to solve the above problems. Summary of the invention

[0005] The present invention provides a dynamic matrix predictive control method for indoor temperature of a central air conditioner for hospital use, which solves the technical problems in the above-mentioned background technology.

[0006] The present invention provides a method for dynamic matrix prediction control of indoor temperature of a central air conditioner for a hospital, comprising the following steps: Step S101, within a first preset time period T1, sensor data of M data collection points in a control area of ​​a central air conditioner are obtained at a preset time interval t, and a feature matrix is ​​constructed as sample data of a first training sample for training a first model; The feature matrix includes M rows and N columns, and the element value of the mth row and nth column represents the sensor data of the mth data collection point at the nth time point, where 1≤m≤M, 1≤n≤N, and N=T1 / t; Sensor data include: temperature value, humidity value and carbon dioxide concentration; The sample data of the first training sample corresponds to the input of the first model; Step S102, obtaining the required temperature after the second preset time period T2 through a simulation experiment as a sample label of a first training sample for training the first model; The sample label of the first training sample corresponds to the output of the first model; Step S103, within the first preset time period T1, the temperature values ​​of the M data collection points acquired in the last time interval minus the required temperature are constructed into a feature sequence as sample data of a second training sample for training a second model; The characteristic sequence includes M sequence units, and the mth sequence unit represents the temperature difference of the mth data collection point, that is, the temperature value minus the required temperature, where 1≤m≤M; The sample data of the second training sample corresponds to the input of the second model; Step S104, simulating K wind speeds at the required temperature through a simulation experiment, and calculating a comprehensive factor according to the energy consumption value of the central air conditioner and the temperature values ​​of the M data collection points after the third preset time period T3, and taking the wind speed corresponding to the minimum value of the comprehensive factor as the sample label of the second training sample for training the second model; The sample labels of the second training samples correspond to the outputs of the second model; Step S105, repeating steps S101 to S104 until U first training samples and second training samples are obtained, and using the first training samples and the second training samples to train the first model and the second model respectively.

[0007] Furthermore, the first preset time period T1, the preset time interval t, the number of data collection points M, the second preset time period T2, the number of simulated wind speeds K, the third preset time period T3, the number of first training samples and the number of second training samples U are all custom parameters.

[0008] Further, the first model includes N hidden layers, 1 extraction layer and 1 first classifier, each hidden layer includes a first unit and a second unit; The first unit of the nth hidden layer inputs the element values ​​of the nth column, the 1st row to the Mth row of the feature matrix, and outputs the first update vector; The second unit of the nth hidden layer inputs the first vector output by the first unit of the nth hidden layer, and outputs a second update vector; The extraction layer is used to extract the second update vector output by the second unit of the Nth hidden layer, and input it into the first classifier; The classification space of the first classifier represents the required temperature in the second preset time period T2.

[0009] Furthermore, the calculation formula of the first unit of the nth hidden layer includes: ; ; ; ; in represents the first update vector for the output of the first unit of the nth hidden layer, Represents the element values ​​of the first row to the Mth row of the nth column of the feature matrix input to the first unit of the nth hidden layer, , and denote the first intermediate vector, the second intermediate vector and the third intermediate vector of the first unit of the hidden layer respectively, , , and denote the first weight parameter, the second weight parameter and the third weight parameter of the first unit of the hidden layer respectively, , and They respectively represent the first bias parameter, the second bias parameter and the third bias parameter of the first unit of the nth hidden layer, tanh represents the hyperbolic tangent activation function, softmax represents the softmax activation function, T represents the transpose operation, and pile represents the stacking operation.

[0010] Furthermore, the calculation formula of the second unit of the nth hidden layer includes: ; ; ; in represents the second update vector for the output of the second unit of the nth hidden layer, and denote the first update vector of the first unit output of the nth and n-1th hidden layers, respectively, and denote the first gating value and the second gating value of the second unit of the nth hidden layer, respectively. and denote the fifth and sixth weight parameters of the second unit of the nth hidden layer, respectively. and They respectively denote the fourth bias parameter and the fifth bias parameter of the second unit of the nth hidden layer, and sigmoid denotes the sigmoid activation function.

[0011] Furthermore, an experimental environment with the same sensor data as the M data collection points in step S101 is constructed through simulation experiments, and 10 groups of required temperatures are generated by floating above and below the reference temperature value at a temperature interval of 0.5°. After the second preset time period T2, the comfort levels of 10 users are obtained through a questionnaire survey, and the average value is taken after eliminating one maximum value and one minimum value. The required temperature with the largest average value is taken as the sample label of the first training sample for training the first model, where the reference temperature value is a custom parameter.

[0012] Furthermore, the calculation formula of the second model is as follows: ; Where Wind represents the wind speed output by the second model, Represents the temperature difference corresponding to the mth sequence unit of the characteristic sequence, and The weight parameters and bias parameters corresponding to the mth sequence unit of the feature sequence respectively.

[0013] Furthermore, a second model is obtained by least squares fitting.

[0014] Furthermore, the calculation formula of the comprehensive factor Score is as follows: ; Energy represents the energy consumption value of central air conditioning. represents the temperature value of the mth data collection point, represents the required temperature, represents the duration of the third preset time period, and denote the first coefficient and the second coefficient respectively, and are custom parameters, and and The sum of is 1.

[0015] Furthermore, the indoor temperature dynamic matrix predictive control method of a central air conditioner for a hospital provided by the present invention further comprises the following steps: Step S106, in the first preset time period T1, sensor data of M data collection points in the control area of ​​the central air conditioner are obtained according to the preset time interval t, and input into the trained first model, and the required temperature in the second preset time period T2 is output; Step S107, obtaining the temperature value of the data collection point closest to the air outlet of the central air conditioner, and judging that the difference between the temperature value and the required temperature is less than the temperature threshold, the central air conditioner is controlled to stop air outlet, otherwise, the process proceeds to step S108; The temperature threshold is a custom parameter; Step S108, within the first preset time period T1, the temperature values ​​of the M data collection points obtained in the last time interval minus the required temperature are constructed into a feature sequence, and input into the trained second model to output the wind speed in the third preset time period T3.

[0016] The beneficial effects of the present invention are as follows: the present invention comprehensively considers the correlation between the sensor data of multiple data collection points through the first model and the second model to control the temperature and wind speed of the central air-conditioning in the future time period, and constructs a training data set with the goal of maximizing comfort and minimizing energy consumption, thereby ensuring the comfort of medical staff and patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The present invention is a flow chart of a method for dynamic matrix prediction and control of indoor temperature of a central air conditioner for a hospital. DETAILED DESCRIPTION

[0018] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0019] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0020] like Figure 1 As shown, a dynamic matrix predictive control method for indoor temperature of a central air conditioner for a hospital comprises the following steps: Step S101, within a first preset time period T1, sensor data of M data collection points in a control area of ​​a central air conditioner are obtained at a preset time interval t, and a feature matrix is ​​constructed as sample data of a first training sample for training a first model; The feature matrix includes M rows and N columns, and the element value of the mth row and nth column represents the sensor data of the mth data collection point at the nth time point, where 1≤m≤M, 1≤n≤N, and N=T1 / t; Sensor data include: temperature value, humidity value and carbon dioxide concentration; The sample data of the first training sample corresponds to the input of the first model; Step S102, obtaining the required temperature after the second preset time period T2 through a simulation experiment as a sample label of a first training sample for training the first model; The sample label of the first training sample corresponds to the output of the first model; Step S103, within the first preset time period T1, the temperature values ​​of the M data collection points acquired in the last time interval minus the required temperature are constructed into a feature sequence as sample data of a second training sample for training a second model; The characteristic sequence includes M sequence units, and the mth sequence unit represents the temperature difference of the mth data collection point, that is, the temperature value minus the required temperature, where 1≤m≤M; The sample data of the second training sample corresponds to the input of the second model; Step S104, simulating K wind speeds at the required temperature through a simulation experiment, and calculating a comprehensive factor according to the energy consumption value of the central air conditioner and the temperature values ​​of the M data collection points after the third preset time period T3, and taking the wind speed corresponding to the minimum value of the comprehensive factor as the sample label of the second training sample for training the second model; The sample labels of the second training samples correspond to the outputs of the second model; Step S105, repeating steps S101 to S104 until U first training samples and second training samples are obtained, and using the first training samples and the second training samples to train the first model and the second model respectively.

[0021] It should be noted that the feature matrix includes the changes in the sensor data of multiple data collection points in the time dimension, that is, the nonlinear mapping relationship between the feature matrix and the required temperature is established through the first model, and the wind speed control of the central air-conditioning is also one of the key factors affecting the energy consumption of the central air-conditioning. Therefore, the nonlinear mapping relationship between the feature sequence and the wind speed is constructed through the second model, and the idea of ​​the present invention can also be used to complete the prediction of the control time of the wind speed. For example, through simulation experiments, multiple control times of K wind speeds are simulated at the required temperature. After the control time, the wind speed and control time corresponding to the maximum value of the temperature closest to the required temperature or the average value of the comfort of multiple medical staff and patients are taken as training labels, which will not be elaborated here.

[0022] It should be noted that the present invention does not exclude the user from manually adjusting the temperature and wind speed of the central air-conditioning, and the manual adjustment instructions have the highest priority. The training of the first model and the second model provided by the present invention can be completed on the cloud server. After the training, the first model and the second model can be deployed in the local controller of the central air-conditioning system, so as to ensure the real-time and reliability of the control.

[0023] In one embodiment of the present invention, the first preset time period T1, the preset time interval t, the number of data collection points M, the second preset time period T2, the number of simulated wind speeds K, the third preset time period T3, the number of first training samples and the number of second training samples U are all custom parameters. Preferably, the default value of T1 is set to 1 minute, the default value of t is set to 10 seconds, the default value of M is set to 6, the default value of T2 is set to 10 minutes, the default value of K is set to 3, the default value of T3 is set to 5 minutes, and U is set to 2000.

[0024] It should be noted that M is set according to the volume of the control area of ​​the central air conditioner. For example, when M is set to 6, 2 data collection points are set outdoors, and the remaining 4 data collection points can be distributed in the 4 corners of the control area. The number of simulated wind speeds K can be set according to the air outlet gear of the central air conditioner, such as low, medium and high gears. Before constructing the feature matrix, the sensor data of multiple data collection points can also be normalized, such as the Min-Max normalization method, which will not be elaborated here.

[0025] In one embodiment of the present invention, the sensor data may also include: human flow and target temperature value, wherein the human flow can be obtained by a camera-integrated target detection algorithm, the target detection algorithm may be a YOLO model or an SSD model, etc., and the target temperature value is manually set according to the requirements of different departments or wards. For example, the ward temperature is set between 22° and 24°, and the operating room temperature is set between 20° and 22°.

[0026] In one embodiment of the present invention, the first model includes N hidden layers, 1 extraction layer and 1 first classifier, each hidden layer includes a first unit and a second unit; The first unit of the nth hidden layer inputs the element values ​​of the nth column, the 1st row to the Mth row of the feature matrix, and outputs the first update vector; The second unit of the nth hidden layer inputs the first vector output by the first unit of the nth hidden layer, and outputs a second update vector; The extraction layer is used to extract the second update vector output by the second unit of the Nth hidden layer, and input it into the first classifier; The classification space of the first classifier represents the required temperature in the second preset time period T2.

[0027] In one embodiment of the present invention, the calculation formula of the first unit of the nth hidden layer includes: ; ; ; ; in represents the first update vector for the output of the first unit of the nth hidden layer, Represents the element values ​​of the first row to the Mth row of the nth column of the feature matrix input to the first unit of the nth hidden layer, , and denote the first intermediate vector, the second intermediate vector and the third intermediate vector of the first unit of the hidden layer respectively, , , and denote the first weight parameter, the second weight parameter and the third weight parameter of the first unit of the hidden layer respectively, , and They respectively represent the first bias parameter, the second bias parameter and the third bias parameter of the first unit of the nth hidden layer, tanh represents the hyperbolic tangent activation function, softmax represents the softmax activation function, T represents the transpose operation, and pile represents the stacking operation.

[0028] It should be noted that the element values ​​from the 1st row to the Mth row of the nth column of the feature matrix, that is, If it is a matrix of size M×3, the second weight parameter, the third weight parameter and the fourth weight parameter can all be designed as vectors of size 3×1, that is, the sizes of the first intermediate vector, the second intermediate vector and the third intermediate vector are all M×1, and the result of stacking the first intermediate vector and the second intermediate vector is a matrix of size M×2. The first weight parameter can be designed as a vector of size 2×1, and the two are multiplied to obtain a vector of size M×1, and then transposed to obtain a vector of size 1×M. The first intermediate vector is multiplied with it to obtain a matrix of size M×M. After the softmax activation function, the element values ​​in the matrix of size M×M are controlled between 0 and 1, and finally multiplied with the third intermediate vector to obtain the first update vector of size M×1.

[0029] In one embodiment of the present invention, during the training process of the first model, the first model can also be pre-trained, and the second update vector output by the second unit of the Nth hidden layer is extracted and input into the second classifier. The classification space of the second classifier represents the temperature value of the second preset time period T2. The weight parameters and bias parameters of the pre-trained first model are retained for further training, which can greatly improve the training speed of the first model.

[0030] In one embodiment of the present invention, the calculation formula of the second unit of the nth hidden layer includes: ; ; ; in represents the second update vector for the output of the second unit of the nth hidden layer, and denote the first update vector of the first unit output of the nth and n-1th hidden layers, respectively, Assign a value of 0, and They represent the first gating value and the second gating value of the second unit of the nth hidden layer, respectively, and are both real values ​​between 0 and 1. and denote the fifth and sixth weight parameters of the second unit of the nth hidden layer, respectively. and They respectively denote the fourth bias parameter and the fifth bias parameter of the second unit of the nth hidden layer, and sigmoid denotes the sigmoid activation function.

[0031] It should be noted that the size of the first update vector is M×1, so the fifth weight parameter and the sixth weight parameter can be designed as a vector of size 1×M. The first gating value and the second gating value obtained by multiplying the two are both real values ​​between 0 and 1, so the size of the final second update vector is M×1.

[0032] In one embodiment of the present invention, the second unit of the nth hidden layer can also be constructed based on RNN (recurrent neural network model), LSTM (long short-term memory neural network model) or GRU (gated neural network model), which will not be elaborated here.

[0033] In one embodiment of the present invention, an experimental environment identical to the sensor data of the M data collection points in step S101 is constructed through a simulation experiment, and 10 groups of required temperatures are generated at a temperature interval of 0.5° above and below the reference temperature value. After a second preset time period T2, the comfort levels of 10 users are obtained through a questionnaire survey, and the average value is taken after eliminating one maximum value and one minimum value. The required temperature with the largest average value is taken as the sample label of the first training sample for training the first model, wherein the reference temperature value is a custom parameter.

[0034] It should be noted that the reference temperature value is set artificially according to the indoor and outdoor temperatures. For example, if the reference temperature value is set to 30°, the 10 groups of required temperatures are 28°, 28.5°, 29°, 29.5°, 30°, 30.5°, 31°, 31.5°, 32°, and 32.5° respectively.

[0035] In one embodiment of the present invention, the calculation formula of the second model is as follows: ; Where Wind represents the wind speed output by the second model, Represents the temperature difference corresponding to the mth sequence unit of the characteristic sequence, and The weight parameters and bias parameters corresponding to the mth sequence unit of the feature sequence respectively.

[0036] In one embodiment of the present invention, the second model is obtained by least squares fitting.

[0037] It should be noted that the maximum number of iterations of the first model and the second model during training is set, and the root mean square error is specified as the loss function. The gradient information of the weight parameters and bias parameters in the first model and the second model is calculated by the chain rule, and the weight parameters and bias parameters in the first model and the second model are reversely updated according to the gradient information, so that the loss value calculated by the loss function is minimized.

[0038] In one embodiment of the present invention, the calculation formula of the comprehensive factor Score is as follows: ; Energy represents the energy consumption value of the central air conditioner, which can be expressed by power consumption. represents the temperature value of the mth data collection point, represents the required temperature, represents the duration of the third preset time period, and denote the first coefficient and the second coefficient respectively, and are custom parameters, and and The sum of the values ​​is 1, preferably, Set to 0.8, Set to 0.2.

[0039] In one embodiment of the present invention, a method for dynamic matrix prediction control of indoor temperature of a central air conditioner for a hospital provided by the present invention further includes the following steps: Step S106, in the first preset time period T1, sensor data of M data collection points in the control area of ​​the central air conditioner are obtained according to the preset time interval t, and input into the trained first model, and the required temperature in the second preset time period T2 is output; Step S107, obtaining the temperature value of the data collection point closest to the air outlet of the central air conditioner, and judging that the difference between the temperature value and the required temperature is less than the temperature threshold, the central air conditioner is controlled to stop air outlet, otherwise, the process proceeds to step S108; The temperature threshold is a custom parameter. Preferably, the temperature threshold is set to 1°; Step S108, within the first preset time period T1, the temperature values ​​of the M data collection points obtained in the last time interval minus the required temperature are constructed into a feature sequence, and input into the trained second model to output the wind speed in the third preset time period T3.

[0040] It should be noted that when the difference between the indoor temperature and the required temperature is too small, the central air conditioner can be directly controlled to stop blowing air, thereby reducing the energy consumption of the central air conditioner.

[0041] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. A dynamic matrix predictive control method for indoor temperature of a central air conditioner for a hospital, characterized in that: The following steps are involved: Step S101, within a first preset time period T1, sensor data of M data collection points in a control area of ​​a central air conditioner are obtained at a preset time interval t, and a feature matrix is ​​constructed as sample data of a first training sample for training a first model; The feature matrix includes M rows and N columns, and the element value of the mth row and nth column represents the sensor data of the mth data collection point at the nth time point, where 1≤m≤M, 1≤n≤N, and N=T1 / t; Sensor data include: temperature value, humidity value and carbon dioxide concentration; The sample data of the first training sample corresponds to the input of the first model; Step S102, obtaining the required temperature after the second preset time period T2 through a simulation experiment as a sample label of a first training sample for training the first model; The sample label of the first training sample corresponds to the output of the first model; Step S103, within the first preset time period T1, the temperature values ​​of the M data collection points acquired in the last time interval minus the required temperature are constructed into a feature sequence as sample data of a second training sample for training a second model; The characteristic sequence includes M sequence units, and the mth sequence unit represents the temperature difference of the mth data collection point, that is, the temperature value minus the required temperature, where 1≤m≤M; The sample data of the second training sample corresponds to the input of the second model; Step S104, simulating K wind speeds at the required temperature through a simulation experiment, and calculating a comprehensive factor according to the energy consumption value of the central air conditioner and the temperature values ​​of the M data collection points after the third preset time period T3, and taking the wind speed corresponding to the minimum value of the comprehensive factor as the sample label of the second training sample for training the second model; The sample labels of the second training samples correspond to the outputs of the second model; Step S105, repeating steps S101 to S104 until U first training samples and second training samples are obtained, and using the first training samples and the second training samples to train the first model and the second model respectively.

2. The indoor temperature dynamic matrix predictive control method of a central air conditioner for a hospital according to claim 1 is characterized in that: The first preset time period T1, the preset time interval t, the number of data collection points M, the second preset time period T2, the number of simulated wind speeds K, the third preset time period T3, the number of first training samples and the number of second training samples U are all custom parameters.

3. The indoor temperature dynamic matrix predictive control method of a central air conditioner for a hospital according to claim 1 is characterized in that: The first model includes N hidden layers, 1 extraction layer and 1 first classifier, each hidden layer includes a first unit and a second unit; The first unit of the nth hidden layer inputs the element values ​​of the nth column, the 1st row to the Mth row of the feature matrix, and outputs the first update vector; The second unit of the nth hidden layer inputs the first vector output by the first unit of the nth hidden layer, and outputs a second update vector; The extraction layer is used to extract the second update vector output by the second unit of the Nth hidden layer, and input it into the first classifier; The classification space of the first classifier represents the required temperature in the second preset time period T2.

4. The indoor temperature dynamic matrix predictive control method of a central air conditioner for a hospital according to claim 3 is characterized in that: The calculation formula for the first unit of the nth hidden layer includes: ； ； ； ； in represents the first update vector for the output of the first unit of the nth hidden layer, Represents the element values ​​of the first row to the Mth row of the nth column of the feature matrix input to the first unit of the nth hidden layer, , and denote the first intermediate vector, the second intermediate vector and the third intermediate vector of the first unit of the hidden layer respectively, , , and denote the first weight parameter, the second weight parameter and the third weight parameter of the first unit of the hidden layer respectively, , and They respectively represent the first bias parameter, the second bias parameter and the third bias parameter of the first unit of the nth hidden layer, tanh represents the hyperbolic tangent activation function, softmax represents the softmax activation function, T represents the transpose operation, and pile represents the stacking operation.

5. The indoor temperature dynamic matrix predictive control method of a central air conditioner for a hospital according to claim 3 is characterized in that: The calculation formula for the second unit of the nth hidden layer includes: ； ； ； in represents the second update vector for the output of the second unit of the nth hidden layer, and denote the first update vector of the first unit output of the nth and n-1th hidden layers, respectively, and denote the first gating value and the second gating value of the second unit of the nth hidden layer, respectively. and denote the fifth and sixth weight parameters of the second unit of the nth hidden layer, respectively. and They respectively denote the fourth bias parameter and the fifth bias parameter of the second unit of the nth hidden layer, and sigmoid denotes the sigmoid activation function.

6. The indoor temperature dynamic matrix predictive control method of a central air conditioner for a hospital according to claim 1, characterized in that: Through simulation experiments, an experimental environment with the same sensor data as the M data collection points in step S101 is constructed, and 10 groups of required temperatures are generated with a temperature interval of 0.5° floating above and below the reference temperature value. After the second preset time period T2, the comfort levels of 10 users are obtained through a questionnaire survey, and the average value is taken after eliminating one maximum value and one minimum value. The required temperature with the largest average value is taken as the sample label of the first training sample for training the first model, where the reference temperature value is a custom parameter.

7. The indoor temperature dynamic matrix predictive control method of a central air conditioner for a hospital according to claim 1, characterized in that: The calculation formula of the second model is as follows: ； Where Wind represents the wind speed output by the second model, Represents the temperature difference corresponding to the mth sequence unit of the characteristic sequence, and The weight parameters and bias parameters corresponding to the mth sequence unit of the feature sequence respectively.

8. A method for dynamic matrix prediction control of indoor temperature of central air conditioner for hospital use according to claim 7, characterized in that: The second model was obtained by least squares fitting.

9. The indoor temperature dynamic matrix predictive control method of a central air conditioner for a hospital according to claim 1, characterized in that: The calculation formula of the comprehensive factor Score is as follows: ； Energy represents the energy consumption value of central air conditioning. represents the temperature value of the mth data collection point, represents the required temperature, represents the duration of the third preset time period, and denote the first coefficient and the second coefficient respectively, and are custom parameters, and and The sum of is 1.

10. The indoor temperature dynamic matrix predictive control method of a central air conditioner for a hospital according to claim 1, characterized in that: The following steps are also included: Step S106, in the first preset time period T1, sensor data of M data collection points in the control area of ​​the central air conditioner are obtained according to the preset time interval t, and input into the trained first model, and the required temperature in the second preset time period T2 is output; Step S107, obtaining the temperature value of the data collection point closest to the air outlet of the central air conditioner, and judging that the difference between the temperature value and the required temperature is less than the temperature threshold, the central air conditioner is controlled to stop air outlet, otherwise, the process proceeds to step S108; The temperature threshold is a custom parameter; Step S108, within the first preset time period T1, the temperature values ​​of the M data collection points obtained in the last time interval minus the required temperature are constructed into a feature sequence, and input into the trained second model to output the wind speed in the third preset time period T3.

Images