CONTROL SYSTEM OF DAMPER OF VARIABLE-AIR-VOLUME AIR DISTRIBUTOR AND CONTROL METHOD THEREOF

Abstract

Disclosed is a control method of a damper of a variable-air-volume air distributor and a control system thereof, mainly comprising a volume adjusting valve for the variable-air-volume air distributor, a valve actuator, a BP neural network-based main controller, a data collector and a user operation panel. The control method thereof is to use, mainly based on the collected room temperature and static pressure at the inlet of the air distributor, the BP neural network predictive model to establish a nonlinear model of temperature & static pressure and valve opening, perform training and optimization, and optimize the output valve opening to obtain and label an optimal value as the set value of the controller to control the action of the valve actuator, thereby realizing an automatic variable air volume.

Claims

1. A control system of a damper of a variable-air-volume air distributor, comprising a volume adjusting valve for the variable-air-volume air distributor, a valve actuator, a BP neural network-based main controller, a data collector and a user operation panel, wherein a signal of a set temperature of the user operation panel is connected to an input of the main controller, the data collector collects room temperature and a static pressure at an inlet of the air distributor, the collected data is connected to the input of the main controller through different sensors, and an output of the main controller is connected to an input of the valve actuator.

2. The control system of a damper of a variable-air-volume air distributor according to claim 1, wherein the BP neural network-based main controller comprises a BP neural network predictive control module and a proportional adjustment control module.

3. The control system of a damper of a variable-air-volume air distributor according to claim 2, wherein the data collector comprises a temperature sensor and a pressure sensor.

4. The control system of a damper of a variable-air-volume air distributor according to claim 3, wherein the BP neural network predictive control module is a dual-input single-output module; and by setting two inputs of the neural network model for the opening action of the damper as a static pressure u and a demanded air volume v within an air duct, respectively, and one output as a valve opening y, the established mathematical model of the opening action of the damper is as follows: $\begin{array}{cc}\left\{\begin{array}{c}{\mathrm{net}}_j^{\left(2\right)}\left(k\right)=\underset{i=1}{\overset{m}{.Math.}}{w}_{\mathrm{ji}}^{\left(2\right)}{O}_i^{\left(1\right)}-{\theta }_j\\ O_j^{\left(2\right)}\left(k\right)=f\left[{\mathrm{net}}_j^{\left(2\right)}\left(k\right)\right]\end{array}\right\}\left\{\begin{array}{c}{\mathrm{net}}_l^{\left(3\right)}\left(k\right)=\underset{j=1}{\overset{k}{.Math.}}{w}_{\mathrm{jl}}^{\left(3\right)}{O}_j^{\left(1\right)}-{\theta }_l\\ O_l^{\left(3\right)}\left(k\right)=g\left[{\mathrm{net}}_l^{\left(3\right)}\left(k\right)\right]\end{array}\right\}y=f\left(f\left(\left[u,v\right]\left[w_1\left(i\right),w_2\left(j\right),w_3\left(k\right)\right]\right)\right)& \left(1\right)\end{array}$ in formula (1), m is the number of neurons in the input layer, and k is the number of neurons in the output layer, $\tanh \left(x\right)=\frac{\sinh \left(x\right)}{\cosh \left(x\right)}=\frac{e^x-e^{-x}}{e^x+e^{-x}};g\left(x\right)=\frac{1}{1+e^{-x}},$ the output value of the above model is continuously corrected by a negative gradient descent method, and the error function calculation formula is:
E(n)=Σ.sub.k=1.sup.L½[r.sub.k(n)−y.sub.k(n)].sup.2   (2) in formula (2), r.sub.k(n) is an expected output value, and y.sub.k(n) is an actual output value.

5. A control method for a control system of a damper of a variable-air-volume air distributor, comprising the specific steps: collecting a data collector in real-time room temperature, a set temperature of a user operation panel and a static pressure at an inlet of the air distributor, and inputting same to a main controller; converting the input temperature signal into an air volume signal within the main controller; passing the data to the main controller, wherein operating conditions is matched firstly in a storage unit, proceed to step 5 if the matching is successful, otherwise proceed to step 4 for modeling on-line learning; modeling and training a neural network model of the damper opening action through a BP neural network predictive model module according to the input data, and establishing a corresponding input-output mapping; and outputting the control parameters optimized by the BP neural network model module to a proportional adjustment control module through the main controller, and then controlling the action of a valve actuator to realize an automatic variable-air-volume.

6. The control method for a control system of a damper of a variable-air-volume air distributor according to claim 5, wherein the temperature signal comprises the collected room temperature and set temperature of the user operation panel.

7. The control method for a control system of a damper of a variable-air-volume air distributor according to claim 5, wherein the control method comprises two functions: on-line modeling learning and off-line matching operating conditions, wherein the matched operating conditions are the prediction models constructed from sample data of the factory test, which are stored within the main controller after training and learning.

8. The control method for a control system of a damper of a variable-air-volume air distributor according to claim 5, wherein the step of modeling and training a neural network model of the damper opening action through a BP neural network predictive model module according to the input data, and establishing a corresponding input-output mapping comprises: the calculation of the BP neural network predictive model module is to establish a mathematical model based on the data obtained by the data collector, and the learning is to verify the effectiveness of the modeling based on the set temperature of the user operation panel; the BP neural network prediction model module corrects and optimizes a nonlinear model established by the module according to the verification results; and the BP neural network prediction model module stores the optimized results in the storage unit of the main controller.

9. The control method for a control system of a damper of a variable-air-volume air distributor according to claim 5, wherein the damper opening action comprises a damper forward action and a damper reverse action.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0028] FIG. 1 is a block diagram of a control system in the present invention.

[0029] FIG. 2 is a schematic diagram of the control method in the present invention.

[0030] FIG. 3 is a diagram of the structure of a BP neural network of the control method in the present invention.

[0031] FIG. 4 is a diagram of the prediction result of a network model valve opening degree of 40% v.s. an actual air volume in the present invention.

[0032] FIG. 5 is a diagram of the prediction error of a network model valve opening degree of 40% v.s. an actual air volume in the present invention.

[0033] FIG. 6 is a diagram of the prediction result of a network model valve opening degree of 65% v.s. an actual air volume in the present invention.

[0034] FIG. 7 is a diagram of the prediction error of a network model valve opening degree of 65% v.s. an actual air volume in the present invention.

DETAILED DESCRIPTION

[0035] In order to deepen the understanding of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and an embodiment. The embodiment is only used to explain the present invention and does not limit the protection scope of the present invention.

[0036] The present invention is directed to a control system of a damper of a variable-air-volume air distributor, and the adjustment of the variable-air-volume damper is the key of the control system. Traditional feedback control has slow response speed and frequent valve actuator actions. By analyzing the principle of variable-air-volume, the present invention uses a BP neural network prediction model to establish a nonlinear model of temperature & static pressure and valve opening according to a collected room temperature and a static pressure at an inlet of the air distributor. According to the established neural network model, an optimization algorithm is used to optimize the output valve opening, to obtain the optimal value under this operating condition, which is labeled as the set value of the controller to control the action of the valve actuator, to realize the automatic variable-air-volume.

[0037] FIG. 1 is a block diagram of the control system of the damper of the variable-air-volume air distributor in the present invention, including a volume adjusting valve, a valve actuator, a user operation panel, a data collector, and a BP neural network-based main controller (a BP neural network predictive model module and a proportional adjustment control module). The data collector is connected to an input of the main controller, and an output of the main controller is connected to the valve actuator and the user operation panel, respectively.

[0038] The working process of the system is that: room temperature is measured by the data collector, and is converted into air volume in the main controller and the air volume is passed to the BP neural network predictive model module; a static pressure in the pipeline is measured by a pressure sensor at an inlet of the air distributor, and a static pressure signal is passed to the BP neural network prediction model module. After neural network modeling and calculation with the demanded air volume and the static pressure signal as two inputs, the main controller first judges the valve adjustment direction to select the forward stroke or reverse stroke model, predicts the valve opening, and feeds same back to the main controller and sends instructions to the valve actuator. After receiving the instruction of the controller, the valve actuator directly moves to a position of the predicted valve opening, to finally achieve the purpose of automatic variable air volume.

[0039] FIG. 2 is a schematic diagram of a control method of the damper of the variable-air-volume air distributor of the present invention. The working principle thereof is that: the proportional adjustment control module is mainly used to receive data from the data collector, and establish the action process and correction of the valve actuator. The BP neural network predictive model module uses the received air volume and static pressure sample data to establish a nonlinear mathematical model with a neural network learning algorithm and perform simulation; complete the calculation and learning of the neural network by comparing to the set temperature of the user operation panel, and then feed the sample data back to the proportional adjustment control module for correction.

[0040] Further, the BP neural network predictive model is a dual-input single-output model;

[0041] two inputs of the neural network model of the damper opening action are a static pressure u and a demanded air volume v within an air duct, respectively, and one output is a valve opening y. The established mathematical model of the damper opening action is as follows:

[00003]$\begin{array}{cc}\left\{\begin{array}{c}{\mathrm{net}}_j^{\left(2\right)}\left(k\right)=\underset{i=1}{\overset{m}{.Math.}}{w}_{\mathrm{ji}}^{\left(2\right)}{O}_i^{\left(1\right)}-{\theta }_j\\ O_j^{\left(2\right)}\left(k\right)=f\left[{\mathrm{net}}_j^{\left(2\right)}\left(k\right)\right]\end{array}\right\}\left\{\begin{array}{c}{\mathrm{net}}_l^{\left(3\right)}\left(k\right)=\underset{j=1}{\overset{k}{.Math.}}{w}_{\mathrm{jl}}^{\left(3\right)}{O}_j^{\left(1\right)}-{\theta }_l\\ O_l^{\left(3\right)}\left(k\right)=g\left[{\mathrm{net}}_l^{\left(3\right)}\left(k\right)\right]\end{array}\right\}y=f\left(f\left(\left[u,v\right]\left[w_1\left(i\right),w_2\left(j\right),w_3\left(k\right)\right]\right)\right)& \left(1\right)\end{array}$

[0042] In formula (1), m is the number of neurons in the input layer, and k is the number of neurons in the output layer,

[00004]$\tanh \left(x\right)=\frac{\sinh \left(x\right)}{\cosh \left(x\right)}=\frac{e^x-e^{-x}}{e^x+e^{-x}};g\left(x\right)=\frac{1}{1+e^{-x}};$

[0043] the output value of the above model is continuously corrected by a negative gradient descent method, and the error function calculation formula is:

E(n)=Σ.sub.k=1.sup.L½[r.sub.k(n)−y.sub.k(n)].sup.2   (2)

[0044] In formula (2), r.sub.k(n) is an expected output value, and y.sub.k(n) is an actual output value.

[0045] The specific steps of a control method for automatic variable air volume of the air distributor comprise: 1. a data collector collects in real-time room temperature, a set temperature of a user operation panel and a static pressure at an inlet of the air distributor, and inputs same to a main controller;

[0046] 2. the input temperature signal is converted into an air volume signal within the main controller;

[0047] 3. the data is passed to the main controller, and first is matched for preset operating conditions in a storage unit, and if the matching is successful, proceed to step 5; otherwise, proceed to step 4, on-line modeling learning;

[0048] 4. a neural network model of the damper opening action is modeled and trained by a BP neural network predictive model module according to the input data, and a corresponding input-output mapping is established;

[0049] 5. the main controller outputs the control parameters optimized by the BP neural network model module to a proportional adjustment control module, and then controls the action of a valve actuator to realize an automatic variable-air-volume.

[0050] Further, the temperature signal in the step 2 above includes the collected room temperature and set temperature of the user operation panel.

[0051] Further, the present control method has two functions: on-line modeling learning and off-line matching operating conditions. The matched operating conditions in the step 3 are the prediction models constructed from sample data of the factory test, which are stored within the main controller after training and learning. Further, the step 4 comprises:

[0052] A1. the calculation of the BP neural network predictive model module is to establish a mathematical model based on the data obtained by the data collector, and the learning is to verify the effectiveness of the modeling based on the set temperature of the user operation panel;

[0053] A2. the BP neural network prediction model module corrects and optimizes a nonlinear model established by the module according to the verification results;

[0054] A3. the BP neural network prediction model module stores the optimized results in the storage unit of the main controller.

[0055] Further, in the step 4, the damper opening action comprises a damper forward action and a damper reverse action.

Embodiment:

[0056] By means of the functions of formula (1) and formula (2) and the data collected by experiments, the established dual-input single-output neural network predictive model is trained for the forward and reverse actions of the damper, and the calculation results of connection weights and thresholds of the trained network structure are shown in the table below.

[0057] Neuron connection weights and thresholds of the trained neural network

TABLE-US-00001 Adjustment direction w.sub.1(i) a(i) w.sub.2(j) b(j) w.sub.3(k) c Forward stroke 1.875 7.281 0.688 −1.097 −1.043 1.071 −1.037 −1.318 −2.545 1.120 0.480 4.561 2.908 1.002 −3.044 −1.291 −1.950 −0.411 −0.094 4.464 0.360 1.523 −6.446 −2.109 −1.932 4.680 1.009 −1.202 −0.934 7.258 0.162 0.310 −0.181 −3.516 −2.483 1.513 −2.304 −0.560 −1.005 −0.025 1.935 −3.741 0.253 −1.588 3.632 −2.692 0.678 −1.135 −0.360 1.071 0.030 −0.527 1.037 5.397 Reverse stroke 3.839 −3.444 3.570 0.057 −0.795 −2.319 −2.774 −0.581 −0.545 0.158 2.038 2.103 1.308 −1.547 −1.649 1.060 −0.527 −0.084 −0.301 −2.637 4.850 5.736 −0.868 0.014 −1.176 1.130 1.370 0.327 −0.466 0.392 0.901 −0.408 −0.807 1.003 4.049 4.271 −1.153 2.248 −0.139 0.818 1.779 −4.349 −3.257 −2.283 −1.689 −0.078 −0.426 −1.144 1.419 −0.778 −2.050 0.656 0.412 −0.646 −1.541 −0.023 2.205 −0.129 −0.312 2.466 1.459

[0058] The above neuron connection weights and thresholds are used depending on different given static pressure commands. When the input demanded air volume assumes a step change, the prediction results of the actual air volume by the network model are as shown in FIG. 4-7.

[0059] For a specific opening of the damper, the larger the static pressure is, the larger the prediction error is; the smaller the static pressure, the smaller the prediction error. In the daily use of marine air distributors, the use at the operating conditions of small static pressure and small flow accounts for a large proportion, and the model will have a good predictive effect in most cases. The error calculation results show that there is a good follow effect between the actual neural network output and the expected output, with an average error being within 5%.

[0060] The control system of the damper of the variable-air-volume air distributor and the control method thereof provided by the present invention can realize design optimization and transformation of the variable-air-volume control system of the existing variable-air-volume air distributor on the market, enhance the degree of automation of the air distributor, and are of great significance for improving the comfortableness of living in the cabin and reducing equipment energy consumption.

[0061] The basic principles, main features and advantages of the present invention have been shown and described above. Those skilled in the art should understand that the present invention is not limited to the foregoing embodiment. The foregoing embodiment and the description in the specification are only intended to illustrate the principles of the present invention. Various variations and improvements may further be made to the present invention without departing from the spirit and scope of the present invention. Such variations and improvements all fall within the scope claimed by the present invention. The scope claimed by the present invention is bounded by claims and their equivalents.