A predictive control method for energy systems based on dynamic Bayesian network model

Abstract

The invention discloses a predictive control method for an energy system based on a dynamic Bayesian network model, and relates to the technical field of predictive control for energy systems. A dynamic Bayesian network is constructed based on the physical information of the energy system, and the dynamic Bayesian network is used to predict the state quantity nodes and posterior probability distribution of the next time step according to the state quantity nodes, control quantity nodes and environmental parameter nodes of the current time step; the energy system is predictively controlled based on the dynamic Bayesian network, and the state quantity nodes and posterior probability distribution of each time step in the next N time steps are gradually predicted by using the dynamic Bayesian network according to the state quantity nodes, control quantity nodes and environmental parameter nodes of the current time step; the target equation and constraint conditions are set, and the decision variables, i.e., the control quantity nodes, are optimized and solved in combination with the prediction results of the next N time steps, and the control quantity nodes of the next time step are finally output, so as to obtain the system control quantity of the next time step. The invention realizes the uncertainty quantification and optimization control of the energy system.

Description

Technical Field

[0001] The present invention relates to the technical field of energy system predictive control, and in particular to an energy system predictive control method based on a dynamic Bayesian network model. Background Art

[0002] With the continuous integration of new energy, energy storage and traditional energy systems, the thermal management of energy systems has become increasingly complex. Traditional control methods (such as proportional-integral-differential control and rule control) have difficulty in dealing with nonlinear and uncertain issues in multi-energy systems. With the development of network information interaction technology and the deep integration of big data and energy systems, the demand for dynamic regulation of system energy consumption (such as refrigeration and air-conditioning systems in smart buildings) has increased significantly. Among them, data-driven predictive control technology has been widely studied in complex energy systems because it can dynamically predict system behavior and optimize control inputs.

[0003] However, current data-driven predictive control technology still has the following limitations, which seriously hinder its practical application:

[0004] Data dependence: Highly dependent on historical data, but insufficient energy data may lead to model prediction errors;

[0005] Computational complexity: The optimization process requires high computing power, which limits its real-time application;

[0006] Physical consistency: Data-driven models lack the ability to describe the physical laws of the system, resulting in insufficient credibility of prediction results;

[0007] Control reliability: Intelligent control based on model prediction cannot effectively respond to system and environmental disturbances, resulting in poor control results. Summary of the invention

[0008] In order to overcome the defects in the above-mentioned prior art, the present invention provides an energy system predictive control method based on a dynamic Bayesian network model, which directly converts the physical information of the energy system into a dynamic Bayesian network structure, and combines predictive control to achieve uncertainty quantification and optimal control of the energy system.

[0009] To achieve the above object, the present invention adopts the following technical solutions, including:

[0010] A method for predictive control of an energy system based on a dynamic Bayesian network model comprises the following steps:

[0011] S1, constructing a dynamic Bayesian network based on the physical information of the energy system;

[0012] The dynamic Bayesian network includes a state quantity node, a control quantity node and an environmental parameter node; the dynamic Bayesian network is used to predict the state quantity node and the posterior probability distribution of the next time step according to the state quantity node, the control quantity node and the environmental parameter node of the current time step;

[0013] S2, predictive control of energy systems based on dynamic Bayesian networks, as shown below:

[0014] The dynamic Bayesian network is used to gradually predict the state quantity nodes and posterior probability distribution of each time step in the next N time steps according to the state quantity nodes, control quantity nodes and environmental parameter nodes of the current time step; among which, the control quantity nodes of each time step in the next N time steps are decision variables, and the environmental parameter nodes are known quantities;

[0015] The objective equation and constraints are set, and the decision variables are optimized and solved in combination with the prediction results of the next N time steps. After the solution, the control quantity node of the next time step is output, so as to obtain the system control quantity of the next time step.

[0016] Preferably, in step S1, the energy system is a building refrigeration and air conditioning system; the physical information includes system state quantity, control quantity and environmental parameters; wherein the state quantity includes room temperature T z , wall temperature T wall , return water temperature T rw and supply air temperature T sa ; Controlled quantity includes cold water flow m sw , air flow rate m sa , cold water supply temperature T sw and the ratio of fresh air flow to supply air flow β; environmental parameters include the external ambient temperature T oa , solar irradiance Q rad and the heat generated inside the building Q ig ;

[0017] The dynamic Bayesian network is used to input the system state quantity, control quantity and environmental parameter of the current time step, i.e., the kth time step, as observation values ​​into the state quantity node x k , control quantity node u k and environment parameter node d k , predict the state node x of the next time step, i.e. the k+1th time step k+1 And the posterior probability distribution P(x k+1 |x k ,u k ,d k );

[0018] Among them, x k =[T z ,T wall ,Trw ,T sa ] k ,u k =[m sw ,m sa ,T sw ,β] k , d k =[T oa ,Q rad ,Q ig ] k , x k+1 =[T z ,T wall ,T rw ,T sa ] k+1 , the subscripts k and k+1 represent the kth and k+1th time steps respectively.

[0019] Preferably, in step S2, the target equation A is:

[0020] ;

[0021] Among them, w1 and w2 are weight factors; E(Q k+j ) is the mathematical expectation of the total energy consumption of the system at the k+jth time step; E(T z,k+j ) is the mathematical expectation of the room temperature at the k+jth time step; T z,s is the set room temperature value.

[0022] Preferably, the mathematical expectation of the total energy consumption of the system E(Q) is calculated by the expected power consumption of the pump E(Q p ), the expected power consumption of the fan E(Q f ) and the expected power consumption E(Q c ) is composed as follows:

[0023] E(Q) = E(Q p )+E(Q f )+E(Q c );

[0024] E(Q p ) = E(p1×(m sw ) 3 +p2×(m sw ) 2 +p3×m sw +p4);

[0025] E(Q f ) = E(f1×(m sa ) 3 +f2×(m sa ) 2 +f3×msa +f4);

[0026] E(Q c ) = E(m sw ×c p,w ×(T rw -T sw ))×EIR;

[0027] Among them, p1, p2, p3, p4 are all coefficients; E(·) is the expected function; f1, f2, f3, f4 are all coefficients; c p,w is the specific heat capacity of water; EIR is the ratio of the electrical energy input to the cooling output of the chiller.

[0028] Preferably, in step S2, the constraints include:

[0029] P(x k+j |x k+j-1 ,u k+j-1 ,d k+j-1 ) = Norm(x k+j ,σ 2 );

[0030] x min <x k+j <x max ;

[0031] u min k+j max ;

[0032] Among them, P(x k+j |x k+j-1 ,u k+j-1 ,d k+j-1 ) is the state node x of the k+jth time step k+j The posterior probability distribution of , j = 1, 2, ..., N; x k+j-1 、u k+j-1 d k+j-1 are the state quantity node, control quantity node, and environmental parameter node of the k+j-1th time step respectively; Norm(x k+j , σ 2 ) means the average is x k+j , a Gaussian probability distribution function with a standard deviation of σ; x k+j is the state node of the k+jth time step; x min 、x max are the minimum and maximum values ​​of the system state quantity respectively; u k+j is the control node of the k+jth time step; u min 、u max are the minimum and maximum values ​​of the system control quantity respectively.​​

[0033] Preferably, the dynamic Bayesian network is based on the current time step, that is, the kth time step x k ,u k ,d k , predict the state node x of the next time step, i.e. the k+1th time step k+1 , calculated as:

[0034] ;

[0035] Among them, a 11 、a 12 、a 21 、a 22 、a 33 、a 34 、a 43 、a 44 、b 31 、b 42 、b 43 、c 11 、c 14 、c 44 d 11 d 12 d 13 d 21 d 22 All are coefficients; m sa,k is the air supply flow rate at the kth time step.

[0036] Preferably, the state node x of the next time step, i.e., the k+1th time step, is k+1 The posterior probability distribution P(x k+1 |x k ,u k ,d k )for:

[0037] P(x k+1 |x k ,u k ,d k ) = P(x k ,u k ,d k |x k+1 )×P(x k+1 ) / P(x k ,u k ,d k );

[0038] Among them, P(x k+1 |x k ,u k ,d k ) represents the state node x of the next time step k+1The posterior probability distribution of k ,u k ,d k Under specific numerical conditions, x k+1 is the probability distribution of possible values; P(x k ,u k ,d k |x k+1 ) indicates that x is known k+1 Under specific numerical conditions, x k ,u k ,d k is the probability distribution of possible values; P(x k+1 ) represents x k+1 The prior probability distribution of P(x k ,u k ,d k ) = P(x k )P(u k )P(d k ); P(x k )、P(u k )、P(d k ) are x k 、u k d k The prior probability distribution of .

[0039] Preferably, in step S2, the minimax algorithm is used to optimize the objective equation, solve the decision variables, and output the control quantity node of the next time step in the decision variables as the system control quantity of the next time step.

[0040] The present invention also provides a computer program product, which includes a computer program / instruction, and when the computer program / instruction is executed by a processor, it implements the above-mentioned energy system predictive control method based on the dynamic Bayesian network model.

[0041] The present invention also provides an electronic device, which includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above-mentioned energy system predictive control method based on the dynamic Bayesian network model is implemented.

[0042] The advantages of the present invention are:

[0043] (1) The present invention directly converts the physical information of the energy system into a dynamic Bayesian network structure, and then predicts and controls the energy system based on the dynamic Bayesian network, thereby achieving uncertainty quantification and optimal control of the energy system.

[0044] (2) The present invention can improve the dynamic performance of complex energy systems, enhance control stability and improve system operation efficiency, and overcome the limitations of existing predictive control methods in terms of data dependence, computational complexity and physical consistency, and can significantly expand the practical application of predictive control technology.

[0045] (3) By embedding physical information, dynamic Bayesian networks can accurately describe the operating laws of energy systems, reduce dependence on large-scale data, and improve the interpretability of control results.

[0046] (4) Combined with the probability model, the control performance of the energy system under uncertain conditions is improved, ensuring the stable operation of the energy system.

[0047] (5) The present invention is applicable to the optimization control of energy systems under different working conditions (such as refrigeration and air-conditioning systems of smart buildings).

[0048] (6) Through multi-objective optimization, the energy efficiency of the energy system (building refrigeration and air-conditioning system) can be maximized while ensuring thermal comfort.

[0049] (7) The computer program product and electronic device disclosed in the present invention can produce the same beneficial effects as the above method by applying the above method, and will not be described in detail. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 The present invention is a flow chart of an energy system predictive control method based on a dynamic Bayesian network model.

[0051] Figure 2 It is a schematic diagram of the structure of the dynamic Bayesian network of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] Example 1

[0054] In this embodiment, the energy system takes the building refrigeration and air conditioning system as an example. Figure 1 As shown, a predictive control method for an energy system based on a dynamic Bayesian network model comprises the following steps:

[0055] S1, construct a dynamic Bayesian network based on the physical information of the building refrigeration and air-conditioning system.

[0056] The physical information of the building refrigeration and air conditioning system includes system state quantity, control quantity and environmental parameters. Among them, the state quantity includes room temperature T z , wall temperature T wall , return water temperature T rw and supply air temperature T sa ; Controlled quantity includes cold water flow m sw , air flow rate m sa , cold water supply temperature T sw and the ratio of fresh air flow to supply air flow β; environmental parameters include the external ambient temperature T oa , solar irradiance Q rad and the heat generated inside the building Q ig .

[0057] Depend on Figure 2 As shown in Figure 1, the dynamic Bayesian network includes state quantity nodes, control quantity nodes and environmental parameter nodes. k It is used to describe the operating status of the building refrigeration and air conditioning system at the current time step, i.e., the kth time step, x k =[T z ,T wall ,T rw ,T sa ] k ; Control quantity node u k It is used to describe the control input of the building refrigeration and air conditioning system at the current time step k, u k =[m sw ,m sa ,T sw ,β] k ; Environmental parameter node d k It is used to describe the internal and external environmental conditions of the building refrigeration and air conditioning system at the current time step k, d k =[T oa ,Q rad ,Q ig ] k .

[0058] The dynamic Bayesian network is used to input the system state quantity, control quantity and environmental parameter measured at the current time step k as observation values ​​into the state quantity node x k , control quantity node u k and environment parameter node d k , and predict through Bayesian inference to get the state node x of the next time step, i.e. the k+1th time step k+1 And the corresponding posterior probability distribution P(x k+1 |x k ,u k ,d k ):

[0059] P(xk+1 |x k ,u k ,d k ) = P(x k ,u k ,d k |x k+1 )×P(x k+1 ) / P(x k ,u k ,d k );

[0060] Among them, P(x k+1 |x k ,u k ,d k ) represents the state node x of the next time step k+1 k+1 The posterior probability distribution of k ,u k ,d k Under specific numerical conditions, x k+1 is the probability distribution of possible values; P(x k ,u k ,d k |x k+1 ) indicates that x is known k+1 Under specific numerical conditions, x k ,u k ,d k is the probability distribution of possible values; P(x k+1 ) represents x k+1 The prior probability distribution of P(x k ,u k ,d k ) means (x k ,u k ,d k ), and since x k 、u k ,d k are independent of each other, so P(x k ,u k ,d k ) = P(x k )P(u k )P(d k ), P(x k )、P(u k )、P(d k ) are x k 、u k ,d k The prior probability distribution of .

[0061] The state node x of the next time step k+1k+1 The system physical equation obtained by combining the mass and energy conservation equations is obtained, that is, x k+1 =f(x k ,u k .d k ), this physical equation can be understood as the state quantity of the system at the next time step k+1 is determined by the state quantity, control quantity and environmental parameters of the system at the current time step k. For building refrigeration and air conditioning systems, x k+1 =f(x k ,u k .d k ) is as follows:

[0062] ;

[0063] Among them, a 11 、a 12 、a 21 、a 22 、a 33 、a 34 、a 43 、a 44 、b 31 、b 42 、b 43 、c 11 、c 14 、c 44 d 11 d 12 d 13 d 21 d 22 These are coefficients in the physical equations of the system and can be learned from the historical data of the building refrigeration and air-conditioning system.

[0064] Anchor point x of the dynamic Bayesian network i Used to store the system state quantity of the current time step; the end point x of the dynamic Bayesian network f Used to store the system state at the next time step.

[0065] S2, predictive control of building refrigeration and air conditioning system based on dynamic Bayesian network, the specific method is as follows:

[0066] Set the prediction calculation step number N of the dynamic Bayesian network, and use the dynamic Bayesian network to perform N-step loop calculations. During the loop calculation process, the data stored at the end point x are gradually f The system state is passed to the anchor point x i , and then carry out the next cycle calculation, and gradually output the system state quantity of the next time step.

[0067] Using dynamic Bayesian network, we can find the state of node x according to the current time step k. k, control quantity node u k and environment parameter node d k , gradually predict the state node x of each time step in the next N time steps k+j And the corresponding posterior probability distribution P(x k+j |x k+j-1 ,u k+j-1 ,d k+j-1 ), j = 1, 2, .., N. Among them, the control quantity node of each time step in the next N time steps is the decision variable; the environmental parameter node of each time step in the next N time steps is a known quantity, which can be predicted based on the weather data of the day and the historical weather data.

[0068] Set the target equation, constraints and optimization algorithm, and optimize the decision variables based on the prediction results of the next N time steps, and finally output the optimized control node u of the next time step k+1. k+1 , thus obtaining the system control quantity for the next time step.

[0069] For building refrigeration and air conditioning systems, the optimization goal is to minimize the mathematical expectation of the total energy consumption of the system E(Q) in the next N time steps, while maintaining thermal comfort as much as possible (i.e. the mathematical expectation of room temperature E(T z ) as close as possible to the set room temperature value T z,s ), the target equation A is as follows:

[0070]

[0071] Among them, w1 and w2 are pre-set weight factors; E(Q k+j ) is the mathematical expectation of the total energy consumption of the system at the k+jth time step; E(T z,k+j ) is the mathematical expectation of the room temperature at the k+jth time step; T z,s is the set room temperature value.

[0072] The mathematical expectation of the total energy consumption of the system E(Q) is calculated by the expected power consumption of the pump E(Q p ), the expected power consumption of the fan E(Q f ) and the expected power consumption E(Q c )constitute:

[0073] E(Q) = E(Q p )+E(Q f )+E(Q c );

[0074] E(Q p ) = E(p1×(m sw ) 3 +p2×(m sw )2 +p3×m sw +p4);

[0075] E(Q f ) = E(f1×(m sa ) 3 +f2×(m sa ) 2 +f3×m sa +f4);

[0076] E(Q c ) = E(m sw ×c p,w ×(T rw -T sw ))×EIR;

[0077] Among them, p1, p2, p3, p4 are all coefficients; E(·) is the expected function; f1, f2, f3, f4 are all coefficients; c p,w is the specific heat capacity of water, and EIR is the ratio of the electrical energy input to the cooling capacity output of the chiller.

[0078] Constraints are based on the physical constraints of the energy system, physical equations and professional knowledge to standardize and rationalize the output of predictive control. For building refrigeration and air conditioning systems, the constraints specifically include:

[0079] P(x k+j |x k+j-1 ,u k+j-1 ,d k+j-1 )=Norm(x k+j ,σ 2 );

[0080] x min <x k+j <x max ;

[0081] u min k+j max ;

[0082] Among them, P(x k+j |x k+j-1 , u k+j-1 , d k+j-1 ) is the state node x of the k+jth time step k+j The posterior probability distribution of j=1,2,...,N; Norm(x k+j , σ 2 ) means the average is x k+j , a Gaussian probability distribution function with a standard deviation of σ; x min 、x​​max are the minimum and maximum values ​​of the system state quantity respectively; u min 、u max are the minimum and maximum values ​​of the system control quantity respectively.

[0083] The optimization algorithm is an algorithmic program used to achieve minimization or minimize the objective equation. In this embodiment, the minimization and maximization (Min-Max) algorithm is used to minimize the objective equation A of the building refrigeration and air-conditioning system, solve the decision variables, and output the control node u of the next time step k+1 in the decision variables. k+1 , as the system control quantity for the next time step.

[0084] Subsequently, the energy system is controlled based on the system control quantity of the next time step, and the system state quantity and environmental parameters of the next time step are measured; then, the algorithm of step S2 above is continued based on the system state quantity, control quantity and environmental parameters measured at the next time step as observation values, so as to obtain the system control quantity of the next time step and complete the dynamic predictive control of the system.

[0085] Example 2

[0086] An electronic device comprises a processor, a memory and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, an energy system predictive control method based on a dynamic Bayesian network model of the above-mentioned embodiment 1 is implemented.

[0087] The electronic device of the embodiment of the present application can be the mobile device itself, or a stand-alone device independent of the mobile device, which can communicate with the mobile device to receive the collected input signals from them and send the selected target decision behavior to them.

[0088] The electronic device includes one or more processors and memory. The processor may be a central processing unit (CPU) or other forms of processing units with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions. The memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor may run the program instructions to implement the decision behavior decision method and / or other desired functions of the various embodiments of the present application described above.

[0089] Electronic devices may also include input devices and output devices.

[0090] Example 3

[0091] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the decision-making behavior decision method according to various embodiments of the present application described in the above-mentioned embodiment 1 of this specification.

[0092] The computer program product may be written in any combination of one or more programming languages ​​to write program codes for performing the operations of the embodiments of the present application, including object-oriented programming languages, such as Java, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0093] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A predictive control method for an energy system based on a dynamic Bayesian network model, characterized in that: The following steps are involved: S1, constructing a dynamic Bayesian network based on the physical information of the energy system; The dynamic Bayesian network includes a state quantity node, a control quantity node and an environmental parameter node; the dynamic Bayesian network is used to predict the state quantity node and the posterior probability distribution of the next time step according to the state quantity node, the control quantity node and the environmental parameter node of the current time step; S2, predictive control of energy systems based on dynamic Bayesian networks, as shown below: The dynamic Bayesian network is used to gradually predict the state quantity nodes and posterior probability distribution of each time step in the next N time steps according to the state quantity nodes, control quantity nodes and environmental parameter nodes of the current time step; among which, the control quantity nodes of each time step in the next N time steps are decision variables, and the environmental parameter nodes are known quantities; Set the target equation and constraints, and optimize the decision variables based on the prediction results of the next N time steps. After solving, output the control quantity node of the next time step, so as to obtain the system control quantity of the next time step. In step S1, the energy system is a building refrigeration and air conditioning system; the physical information includes system state quantity, control quantity and environmental parameters; the state quantity includes room temperature T z , wall temperature T wall , return water temperature T rw and supply air temperature T sa ; Controlled quantity includes cold water flow m sw , air flow rate m sa , cold water supply temperature T sw and the ratio of fresh air flow to supply air flow β; environmental parameters include the external ambient temperature T oa , solar irradiance Q rad and the heat generated inside the building Q ig ; The dynamic Bayesian network is used to input the system state quantity, control quantity and environmental parameter of the current time step, i.e., the kth time step, as observation values ​​into the state quantity node x k , control quantity node u k and environment parameter node d k , predict the state node x of the next time step, i.e. the k+1th time step k+1 And the posterior probability distribution P(x k+1 |x k ,u k ,d k ); Among them, x k =[T z ,T wall ,T rw ,T sa ] k ,u k =[m sw ,m sa ,T sw ,β] k , d k =[T oa ,Q rad ,Q ig ] k , x k+1 =[T z ,T wall ,T rw ,T sa ] k+1 , subscripts k and k+1 represent the kth and k+1th time steps respectively; In step S2, the target equation A is: ； Among them, w1 and w2 are weight factors; E(Q k+j ) is the mathematical expectation of the total energy consumption of the system at the k+jth time step; E(T z,k+j ) is the mathematical expectation of the room temperature at the k+jth time step; T z,s is the set room temperature value.

2. The energy system predictive control method based on the dynamic Bayesian network model according to claim 1 is characterized in that: The mathematical expectation of the total energy consumption of the system E(Q) is calculated by the expected power consumption of the pump E(Q p ), the expected power consumption of the fan E(Q f ) and the expected power consumption E(Q c ) is composed as follows: E(Q) = E(Q p )+E(Q f )+E(Q c ); E(Q p ) = E(p1×(m sw ) 3 +p2×(m sw ) 2 +p3×m sw +p4); E(Q f ) = E(f1×(m sa ) 3 +f2×(m sa ) 2 +f3×m sa +f4); E(Q c ) = E(m sw ×c p,w ×(T rw -T sw ))×EIR; Among them, p1, p2, p3, p4 are all coefficients; E(·) is the expected function; f1, f2, f3, f4 are all coefficients; c p,w is the specific heat capacity of water; EIR is the ratio of the electrical energy input to the cooling capacity output of the chiller.

3. The energy system predictive control method based on the dynamic Bayesian network model according to claim 1 is characterized in that: In step S2, the constraints include: P(x k+j |x k+j-1 ,u k+j-1 ,d k+j-1 ) = norm(x k+j ,σ 2 ); x min <x k+j <x max ; in min k+j max ;​​ Among them, P(x k+j |x k+j-1 ,u k+j-1 ,d k+j-1 ) is the state node x of the k+jth time step k+j The posterior probability distribution of , j = 1, 2, ..., N; x k+j-1 、u k+j-1 ,d k+j-1 They are the state quantity node, control quantity node, and environmental parameter node of the k+j-1th time step respectively; Norm(x k+j , σ 2 ) means the average is x k+j , a Gaussian probability distribution function with a standard deviation of σ; x k+j is the state node of the k+jth time step; x min 、x max are the minimum and maximum values ​​of the system state quantity respectively; u k+j is the control node of the k+jth time step; u min 、u max They are the minimum and maximum values ​​of the system control quantity respectively.

4. The energy system predictive control method based on the dynamic Bayesian network model according to claim 1 is characterized in that: The dynamic Bayesian network is based on the current time step, that is, the kth time step x k ,u k ,d k , predict the state node x of the next time step, i.e. the k+1th time step k+1 , calculated as: ； Among them, a 11 、a 12 、a 21 、a 22 、a 33 、a 34 、a 43 、a 44 , b 31 , b 42 , b 43 、c 11 、c 14 、c 44 ,d 11 ,d 12 ,d 13 ,d 21 ,d 22 All are coefficients; m sa,k is the air supply flow rate at the kth time step.

5. The energy system predictive control method based on the dynamic Bayesian network model according to claim 1 is characterized in that: The state node x of the next time step, i.e. the k+1th time step k+1 The posterior probability distribution P(x k+1 |x k ,u k ,d k )for: P(x k+1 |x k ,u k ,d k ) = P(x k ,u k ,d k |x k+1 )×P(x k+1 ) / P(x k ,u k ,d k ); Among them, P(x k+1 |x k ,u k ,d k ) represents the state node x of the next time step k+1 The posterior probability distribution of x k ,u k ,d k Under specific numerical conditions, x k+1 is the probability distribution of possible values; P(x k ,u k ,d k |x k+1 ) indicates that x is known k+1 Under specific numerical conditions, x k ,u k ,d k is the probability distribution of possible values; P(x k+1 ) represents x k+1 The prior probability distribution of P(x k ,u k ,d k ) = P(x k )P(u k )P(d k ); P(x k )、P(u k )、P(d k ) are x k 、u k ,d k The prior probability distribution of .

6. The energy system predictive control method based on a dynamic Bayesian network model according to any one of claims 1 to 5, characterized in that: In step S2, the minimax algorithm is used to optimize the objective equation, solve the decision variables, and output the control quantity node of the next time step in the decision variables as the system control quantity of the next time step.

7. A computer program product, characterized in that It includes a computer program / instruction, which, when executed by a processor, implements an energy system predictive control method based on a dynamic Bayesian network model as described in any one of claims 1-6.

8. An electronic device, characterized in that: It includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements an energy system predictive control method based on a dynamic Bayesian network model as described in any one of claims 1 to 6.

Images