CN114488811A

CN114488811A - Greenhouse environment energy-saving control method based on second-order Voltalla model prediction

Info

Publication number: CN114488811A
Application number: CN202210087874.XA
Authority: CN
Inventors: 蔚瑞华; 徐立鸿; 蔡文韬
Original assignee: Tongji University
Current assignee: Tongji University
Priority date: 2022-01-25
Filing date: 2022-01-25
Publication date: 2022-05-13
Anticipated expiration: 2042-01-25
Also published as: CN114488811B

Abstract

The invention relates to a greenhouse environment energy-saving control method based on second-order Voltalla model prediction, which comprises the following steps: 1) constructing a second-order Voltara greenhouse environment model, and identifying model parameters by combining historical greenhouse environment data; 2) constructing a loss function which gives consideration to the tracking error of the control target and the energy consumption index, and taking the loss function as a target function of the optimal problem; 3) adjusting the size of a weight factor in the loss function according to actual requirements; 4) and optimizing the loss function in the feasible region range of the actuator by adopting a traversal method, and acquiring the output value of the actuator meeting the requirement as the system input at the next moment. Compared with the prior art, the method has the characteristics of accurate environment prediction, good control effect and less energy consumption requirement, can effectively improve the economic benefit of greenhouse control of an automation department, has the advantages of strong universality and expansibility and the like, provides theoretical guidance and decision support for improving the production benefit of the greenhouse, and ensures the reliability and energy conservation of greenhouse environment control.

Description

Greenhouse environment energy-saving control method based on second-order Voltalla model prediction

Technical Field

The invention relates to the technical field of agricultural facility environment control, in particular to a greenhouse environment energy-saving control method based on second-order Voltara model prediction.

Background

In the aspect of greenhouse control, a control technology (MPC) based on model prediction is commonly adopted in the current intelligent greenhouse, and compared with PID control only suitable for a single-loop stable system, the model prediction control can obtain a better control effect in multivariable system control, and is a common control method in the current control field. For obtaining the prediction model, common modeling methods are: mechanism modeling method and data-driven modeling method. The mechanism model has high accuracy, good interpretability and abundant theoretical support, so the mechanism model is widely applied, but the model parameters are extremely many, the identification difficulty is extremely high, and an available prediction model is difficult to establish in a short time; although the data-driven modeling methods such as the neural network model and the decision tree model can efficiently model, the prediction effect is limited by the quality of the data set, and the machine learning method is lack of interpretability at present, so the credibility of the machine learning method applied to prediction control is not as good as the mechanism method.

In addition, most model prediction control methods need to solve the optimal solution of a non-convex optimization problem to obtain an optimal control strategy, in order to prevent the situation of falling into a local optimal solution, an evolutionary method is generally adopted for solving, the method needs to obtain a global optimal solution through a large number of iterative computations, and then a series of problems such as high computation cost and high program complexity are brought.

Therefore, in the current greenhouse development situation and the actual production requirement in China, the problems widely existing in the existing prediction control method are considered, the practicability and the economic benefit are taken as the middle targets, and a novel greenhouse environment prediction control method which is good in control performance, convenient and fast in model establishment, low in calculation cost and easy to expand is required to be established.

Disclosure of Invention

The invention aims to overcome the defects in the prior art and provides a greenhouse environment energy-saving control method based on second-order Volterra model prediction, which is applied to the field of agricultural greenhouse environment control.

The purpose of the invention can be realized by the following technical scheme:

a greenhouse environment energy-saving control method based on second-order Voltara model prediction comprises the following steps:

1) constructing a second-order Voltara greenhouse environment model, and identifying model parameters by combining historical greenhouse environment data;

2) constructing a loss function which gives consideration to the tracking error of the control target and the energy consumption index, and taking the loss function as a target function of the optimal problem;

3) adjusting the size of a weight factor in the loss function according to actual requirements;

4) and optimizing the loss function in the feasible region range of the actuator by adopting a traversal method, and acquiring the output value of the actuator meeting the requirement as the system input at the next moment.

In the step 1), the second-order Voltalla greenhouse environment model specifically adopts a second-order Voltalla series with a multi-input single-output form, and the expression is as follows:

wherein y (k) is the model output at time k, i.e. the system controlled variable, h₀Is a fixed offset of the model, n_uAnd n_dRespectively representing the control variables u as model inputs_m1And disturbance d_m2And m1 is 1, …, n_u，m2＝1,…,n_d，

Are respectively a controlled variable u_m1And disturbance d_m2Represents the order of the linear part and the non-linear part,

are respectively a controlled variable u_m1And disturbance d_m2The weights of the linear and nonlinear parts are acted on, i and j being the current iteration values of the linear and nonlinear action truncation orders, respectively.

In the step 1), parameter identification is carried out on the second-order Voltalla greenhouse environment model through a least square method.

The specific process of parameter identification is as follows:

in the identification process of the first-order truncation order, initializing the value of each truncation order to 1, uniformly taking the value of the second-order truncation order to 0, identifying and obtaining a preliminary model parameter under the current condition by adopting a least square method, recording a fitting error, sequentially increasing the first-order truncation order value of each model input variable by 1, identifying by adopting the least square method again, stopping iteration when the error reduction of the series fitting under the truncation order is less than 5% than that of the last error, and taking the identified truncation order as the final truncation order value of the model input variable;

after the identification of the first-order truncation order is finished, the same method is applied to the second-order truncation order, the previously identified first-order truncation order is brought into the same for calculation, and the truncation order value which enables the error between the model output value and the acquired data to be minimum and the order value to be minimum can be identified by traversing the error under each truncation order;

and after the first-order truncation order and the second-order truncation order of all the model input variables are determined, identifying the weight of each model input variable by adopting a least square method again to form a complete second-order Volterra greenhouse environment model.

The model output is specifically the temperature or humidity of the greenhouse, the control variable is specifically the opening degree of an actuator, the actuator comprises a skylight, a sunshade net and a fan, and the disturbance is specifically greenhouse environment state factors including outdoor temperature, sunlight irradiation and wind speed.

In step 2), the expression of the loss function J is:

J＝(1-λ)(y(k+1|k)-r(k+1|k))²+λ(Δu(k+1|k)²+u(k+1|k)²)

wherein r (k +1| k) is a set value of the controlled variable at the time of k +1, λ is a weight factor for balancing the tracking error of the system and the variation of the controller, y (k +1| k) is a model output of the controlled variable at the time of k +1, and Δ u (k +1| k) is a control variable increment.

The control variable increment Δ u (k +1| k) is calculated by the following formula:

Δu(k+1|k)＝u(k+1|k)-u(k)

where u (k +1| k) and u (k) represent the actuator output values of the system at time k +1 and time k, respectively.

In the step 3), the sensitivity of the control error and the energy consumption is adjusted by adjusting the weight factor in the loss function.

In the step 4), a loss function is used as an optimization objective function, and a control variable increment Δ u (k +1| k) at the following moment is used as a variable to be solved to form a minimum optimization problem with constraints, which includes:

Δu^*＝argminJ(Δu(k+1|k),u(k+1|k))

wherein, Δ u^*For optimum control variable increment, u_maxAnd u_minRespectively, the maximum and minimum values of the control variable range.

And 4) carrying out the step once every 5 minutes to solve the optimal control input value of the greenhouse environment at the current moment.

Compared with the prior art, the invention has the following advantages:

firstly, the model establishment cost is reasonable: compared with a pure mechanical model, the greenhouse environment model based on the second-order Voltara series has fewer to-be-identified parameters, can realize the fitting of data without using a parameter identification method with high time complexity, and has less time spent in the fitting process and relatively less time consumption for solving a prediction result compared with a machine learning method which is used for data-driven modeling.

Secondly, the method has strong interpretability: compared with a direct data-driven prediction method adopting machine learning, the control method based on the second-order Volterra series model takes the current values of each environment disturbance action and the actuator as the input of the model, takes the environment variables as the output values of the model, and has a similar mathematical principle with a mechanism model, so that the prediction output of the second-order Volterra series model has stronger interpretability for the machine learning method and has higher reliability in actual control.

Thirdly, the control method has good economic benefits: compared with the existing greenhouse environment control method, the method provided by the invention takes the energy consumption as a comprehensive consideration index, can improve the overall economic benefit of the greenhouse by reducing the energy consumption of the actuator on the premise of ensuring controllable greenhouse environment, and provides an adjustable weight factor, so that a user can adjust the method according to the actual requirement to control the error and optimize the energy consumption.

Fourthly, the universality and the expansibility are strong: the method has simple principle, is easy to program and modify, can work as an independent control method and can also be integrated into the existing control method as a plug-in to realize function expansion, and has certain universality.

Drawings

FIG. 1 is a schematic flow diagram of the process of the present invention.

FIG. 2 is a diagram of the effect of a second-order Voltara series model on the greenhouse environment temperature.

Detailed Description

The technical solutions in the technical embodiments of the present invention will be described in detail, clearly and completely with reference to the accompanying drawings in the embodiments of the present invention.

Examples

As shown in FIG. 1, the invention provides a greenhouse environment energy-saving control method based on second-order Waltala model prediction, which comprises the following steps:

s1: according to a general expression of a second-order Voltalla series, combining with historical environmental data of the greenhouse, identifying the truncation order and the coefficient of each parameter of the Voltalla series by using a least square method, and establishing a second-order Voltalla greenhouse environmental model;

s2: constructing a loss function which simultaneously gives consideration to the tracking error of a control target and the energy consumption index, and adopting a weight factor to adjust the balance between the error and the energy consumption as a target function of an optimal problem;

s3: according to actual requirements, the size of a weight factor in the loss function is adjusted, so that the control method is adjusted in an 'aggressive' and 'smooth' control style;

s4: and optimizing the loss function in the feasible region range of the controller by adopting a traversal method every 5 minutes, and acquiring the output value of the controller or the actuator meeting the requirement as the system input of the next moment.

In step S1, a general expression of a second-order volterra series is written, and the weight coefficients of each parameter in the volterra series are identified, which includes the following steps:

s101: according to the basic mathematical form of the second-order Volterra series model, the expression of the second-order Volterra series of the multi-input single-output form is as follows:

where y (k) is the model output at time k, h₀Is a fixed offset of the model, n_uAnd n_dRespectively representing the control variables u as model inputs_m1And disturbance d_m2And m1 is 1, …, n_u，m2＝1,…,n_d，

are respectively a controlled variable u_m1And disturbance d_m2The invention takes greenhouse environment parameters (such as temperature and humidity) as model output, opening information of an actuator as model input and environmental state factor information as disturbance.

S102: in order to obtain the truncation order and the weight coefficient of each variable of the second-order Woltalant series, the system identification is needed, in the identification process of the first-order truncation order, the value of each truncation order is initialized to a smaller value (1 is taken in the invention, at the moment, the second-order truncation order is uniformly taken as 0), a least square method is applied to identify a preliminary model parameter under the current condition and record the fitting error, then the first-order truncation order value of each variable is sequentially increased by 1 and identified by the least square method again, when the reduction of the number fitting error under the truncation order is less than 5 percent than the last error reduction, the iteration is stopped, the last identified truncation order is taken as the final truncation order value of the variable, after the identification of the first-order truncation order, the same method is applied to the second-order truncation order, but at the moment, the previously identified first-order truncation order is required to be brought into the calculation together, by traversing the errors under each truncation order by the method, the truncation order numerical value which enables the error between the model output value and the acquired data to be smaller and the order numerical value to be smaller can be identified.

And after the first-order and second-order truncation orders of all variables are determined, identifying the weight of each variable by using a least square method again, and forming a complete second-order Volterra greenhouse environment model.

Fig. 2 shows the fitting result of the above identification steps applied to experimental greenhouse data (where the model output is indoor temperature, the actuator is skylight and fan, disturbance factors are outdoor temperature, sunlight illumination and wind speed), and it can be seen that the second-order volterra series model can more accurately fit greenhouse environment data, and has the capability of being applied to subsequent model predictive control.

In step S2, in order to simultaneously take into account the control tracking error and the energy consumption of the system, the loss function of the control method is defined as follows:

J＝(1-λ)(y(k+1|k)-r(k+1|k))²+λ(Δu(k+1|k)²+u(k+1|k)²)

wherein r (k +1| k) represents a set value of a controlled variable (indoor temperature or humidity) at the moment t ═ k +1, λ is a weighting factor for balancing a system tracking error and a controller variation, a method emphasis point between control accuracy and energy consumption optimization can be adjusted by adjusting the value, Δ u (k +1| k) represents a controlled variable increment, and a specific calculation mode of the method can be defined by the following formula:

Δu(k+1|k)＝k(k+1|k)-u(k)

where u (k +1| k) and u (k) represent actuator output values of the system at time t + k +1 and time t k, respectively.

Because the energy consumption calculation modes of actuators in the actual greenhouse are different, the energy consumption of partial actuators comes from the change of states (such as skylights and sunshade nets), the energy consumption of partial actuators comes from the continuation of the states (such as fans), and the increment or the output value of the controller is respectively brought into calculation according to the attributes of the actuators in the actual calculation.

In step S3, the sensitivity of the method to control errors and energy consumption can be adjusted by adjusting the weight factor λ in step S2, so as to adjust the method according to actual requirements;

in step S4, the loss function in step S2 is used as an optimization objective function, and the control variable increment (actuator output change value) Δ u (k +1| k) at the next time is used as a variable to be solved, so as to form a minimum optimization problem with constraints:

Δu^*＝argminJ(Δu(k+1|k),u(k+1|k))

the greenhouse environment output in the loss function, i.e., the y (k +1| k) term, can be obtained by substituting k (k +1| k) ═ Δ u (k +1| k) + k (k) into the second-order volterra series model obtained in the above step.

Because the number of actuators is limited in the actual greenhouse control, and the two-state actuators only have 2 states, the searching range is greatly reduced, and the traversal method has greater advantages compared with the conventional optimization method. Therefore, the optimization problem is solved by using a traversal method in practical application, and the three-state actuator needs to be graded in advance (for example, the opening degree of the skylight is divided into 6 grades: 0%, 20%, 40%, 60%, 80%, 100% in the greenhouse applied by the invention). And then, taking values of each actuator in the feasible region of the actuator to form a series of control strategies, and substituting the control strategies into the loss function to calculate so as to obtain the control strategy which enables the output value of the loss function to be minimum.

In the subsequent process, step S4 will be executed every 5 minutes to solve the optimal control input value of the greenhouse environment at the present time.

When the weight coefficient of the loss function is 0.15, the control method of the invention is adopted, the range of the indoor temperature exceeding the upper temperature limit is controllable, and the action range and the frequency of the skylight are also within the acceptable range, thereby realizing the balance of the control model between the control error and the energy consumption and simultaneously obtaining better control effect.

Claims

1. A greenhouse environment energy-saving control method based on second-order Voltara model prediction is characterized by comprising the following steps:

2. The method for controlling energy conservation of greenhouse environment based on second-order Waltala model prediction as claimed in claim 1, wherein in step 1), the second-order Waltala greenhouse environment model specifically adopts a second-order Waltala series of multiple inputs and single outputs, and the expression is as follows:

wherein y (k) is the model output at time k, i.e. the system controlled variable, h₀Is a fixed offset of the model, n_uAnd n_dRespectively representing the control variables u as model inputs_m1And disturbance d_m2And m1 ═1,…,n_u，m2＝1,…,n_d，

3. The energy-saving control method for greenhouse environment based on second-order Waltala model prediction as claimed in claim 2, wherein in step 1), the second-order Waltala greenhouse environment model is subjected to parameter identification by a least square method.

4. The second-order Volterra model prediction-based greenhouse environment energy-saving control method as claimed in claim 3, wherein the specific process of parameter identification is as follows:

5. The energy-saving control method for greenhouse environment based on second-order Volterra model prediction as claimed in claim 2, wherein the model output is specifically the temperature or humidity of the greenhouse, the control variable is specifically the opening degree of an actuator, the actuator comprises a skylight, a sunshade net and a fan, and the disturbance is specifically the state factor of the greenhouse environment, including outdoor temperature, sunlight and wind speed.

6. The energy-saving control method for greenhouse environment based on second-order Volterra model prediction as claimed in claim 2, wherein in step 2), the expression of the loss function J is:

J＝(1-λ)(y(k+1|k)-r(k+1|k))²+λ(Δu(k+1|k)²+u(k+1|k)²)

7. The energy-saving control method for greenhouse environment based on second-order Volterra model prediction as claimed in claim 6, wherein the calculation formula of the control variable increment Δ u (k +1| k) is:

Δu(k+1|k)＝u(k+1|k)-u(k)

8. The method for controlling energy conservation of greenhouse environment based on second-order Volterra model prediction as claimed in claim 1, wherein in the step 3), the sensitivity of control error and energy consumption is adjusted by adjusting the weight factor of the loss function.

9. The method for controlling energy conservation of greenhouse environment based on second-order Volterra model prediction as claimed in claim 6, wherein in the step 4), a loss function is used as an optimization objective function, and a control variable increment Δ u (k +1| k) at the following moment is used as a candidate variable to form a constrained minimum optimization problem, and then:

10. The energy-saving control method for greenhouse environment based on second-order Volterra model prediction as claimed in claim 1, wherein step 4) is performed every 5 minutes to solve the optimal control input value for greenhouse environment at the current moment.