MPC-based two-layer optimization scheduling method and device for heating system with heat storage

Abstract

The present invention discloses a two-layer optimization scheduling method for a heating system with heat storage based on MPC, comprising: configuring a heat storage device in the heating system, setting a scheduling instruction based on a predicted output value of a renewable energy heating unit, tracking the scheduling instruction with the combined output of the renewable energy heating unit and the heat storage device and minimizing the action amount of the heat storage device as control optimization objectives, optimizing the heat storage device through a model predictive control strategy, optimizing and solving the model predictive control through rolling optimization and feedback correction links to obtain the action instruction of the heat storage device; reducing the load of a traditional heating unit through the combined output of the renewable energy heating unit and the heat storage device, optimizing the traditional heating unit using the residual load, optimizing the output of the traditional heating unit with the minimization of the operating cost of the traditional heating unit, the operating cost of the renewable energy heating unit and the operating cost of the heat storage device within the scheduling period as the objective function, and obtaining the action instruction of the traditional heating unit.

Description

Technical Field

[0001] The present invention belongs to the field of smart heating technology, and specifically relates to a two-layer optimization scheduling method for a heating system with heat storage based on MPC. Background Art

[0002] The autonomous optimization operation of the smart heating system means that the system constructs a digital twin model of the heating system through mechanism modeling and data identification, and combines artificial intelligence technologies such as model predictive control and real-time optimization to coordinate and optimize the heating system from multiple levels. The optimized operation strategy is automatically issued to the control system to achieve automatic and safe closed-loop control of the entire process, so that the heating system has the ability of self-perception, self-learning, self-adaptation, and self-regulation.

[0003] With the development trend towards low-carbon, clean, and intelligent heating, a high proportion of renewable energy needs to be integrated into heating systems. However, renewable energy is characterized by randomness, volatility, and uncertainty. When using renewable energy for heating, it is bound to cause fluctuations in the heating system, affecting the safe and stable operation of the heating system and the heating quality, greatly increasing the difficulty of heating system scheduling. Therefore, how to reduce the adverse effects of renewable energy uncertainty on heating scheduling, smooth out the volatility of renewable energy output, and improve the economic and reliability of heating system operation are currently urgent issues that need to be addressed.

[0004] Based on the above technical problems, it is necessary to design a new MPC-based two-layer optimization scheduling method for heating systems with heat storage. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the shortcomings of the existing technology and provide a two-layer optimization scheduling method for a heating system with heat storage based on MPC. By establishing a two-layer optimization scheduling model, optimizing the output of the heat storage device can effectively compensate for the output power of renewable energy, reduce the adverse effects and uncertainty and volatility of the uncertainty of renewable energy on the scheduling of the heating system, and fill the remaining load shortfall with the output of traditional heating units. The output of traditional heating units is optimized by an intelligent optimization algorithm as the second-layer optimization of scheduling, which can improve the economy and reliability of the heating system.

[0006] In order to solve the above technical problems, the technical solution of the present invention is:

[0007] The present invention provides a two-layer optimization scheduling method for a heating system with heat storage based on MPC, which includes:

[0008] A digital twin model of a heating system is established by using a mechanism modeling and data identification method; the heating units in the heating system include at least a renewable energy heating unit and a traditional heating unit;

[0009] Based on the digital twin model of the heating system, historical heating operation data and outdoor meteorological data are obtained. A machine learning algorithm is used to establish a heating system load forecasting model and a renewable energy heating unit output forecasting model to obtain the total load demand of the heating system and the output forecast values ​​of the renewable energy heating units.

[0010] Establish a first-level optimization scheduling model: Introduce a model predictive control strategy into a heating system equipped with a heat storage device. Set the scheduling instructions based on the output forecast of the renewable energy heating unit. Track the scheduling instructions based on the combined output of the renewable energy heating unit and the heat storage device, and minimize the amount of heat storage device movement as the control optimization objectives. Optimize the heat storage device using the model predictive control strategy. The model predictive control is optimized and solved through rolling optimization and feedback correction links to obtain the action instructions for the heat storage device.

[0011] Establish a second-level optimization scheduling model: reduce the load of traditional heating units through the joint output of renewable energy heating units and heat storage devices, use the remaining load to optimize traditional heating units, and take the minimization of traditional heating unit operating costs, renewable energy operating costs and heat storage device operating costs during the scheduling period as the objective function. Optimize the output of traditional heating units through intelligent optimization algorithms to obtain action instructions for traditional heating units.

[0012] Furthermore, the method of using mechanism modeling and data identification to establish a digital twin model of the heating system includes:

[0013] Establish physical models, logical models, simulation models, and data models for renewable energy heating units, traditional heating units, primary networks, secondary networks, thermal power stations, and terminal buildings in the heating system. Couple the physical models, logical models, simulation models, and data models, and integrate them at multiple levels and scales. Establish a digital twin model of the heating system after mapping and reconstructing physical entities in the physical space in the virtual space.

[0014] An improved adaptive inertia weighted particle swarm optimization algorithm is used to optimize the parameters of the digital twin model of the heating system: the inertia weight nonlinear decreasing update strategy and mutation operation are combined and introduced into the PSO algorithm to form an improved adaptive inertia weighted particle swarm optimization algorithm; the parameters to be identified in the digital twin model of the heating system are determined, the value range of each parameter is set, and each parameter optimization problem is converted into a particle position optimization problem; the position variable is introduced, and the deviation of the digital twin model of the heating system is solved with the help of the training sample data obtained in the heating system. The improved adaptive inertia weighted particle swarm optimization algorithm is used to select the optimal position of the particle according to the size of the deviation value to obtain the optimal parameters of the digital twin model of the heating system; the root mean square error and mean absolute percentage error are selected as measurement indicators to verify the performance of the digital twin model of the heating system.

[0015] Furthermore, the heating system digital twin model is used to obtain historical heating operation data and outdoor meteorological data, and a heating system load forecasting model is established using a machine learning algorithm to obtain the total load demand of the heating system and the output forecast value of the renewable energy heating unit, including:

[0016] Based on the digital twin model of the heating system, the historical output of renewable energy units, the output of traditional heating units, the heating system load, the operating parameters of the heating station, the outdoor temperature, humidity, and the indoor temperature of the terminal building are obtained as training samples for the heating system load prediction model; the historical output of renewable energy units, the operating parameters of the heating station, the outdoor temperature, humidity, and the indoor temperature of the terminal building are obtained as training samples for the renewable energy heating unit output prediction model;

[0017] After optimizing the CNN-REGST model parameters using the POA Pelican optimization algorithm, the optimized CNN-REGST model was used to establish a heating system load forecasting model and a renewable energy heating unit output forecasting model: after extracting features from the model training samples using the CNN convolutional neural network model, the REGST stacked regression model theory was used to input the feature-extracted data into multiple basic learners respectively to obtain multiple heating system load forecast values ​​and multiple renewable energy heating unit output forecast values. The SVM model was then used as a meta-regressor to calculate the multiple forecast values ​​to obtain the heating system load forecast value and the renewable energy heating unit output forecast value.

[0018] Furthermore, the POA Pelican optimization algorithm is used to optimize the CNN-REGST model parameters, including:

[0019] Initialize the CNN-REGST model parameters, including the number of neurons, learning rate, and the number of nodes and filters in the fully connected layer;

[0020] The CNN-REGST model parameters are optimized using the POA Pelican optimization algorithm: the number of Pelican population members and the maximum number of iterations are initialized; the initial population is generated and the objective function is calculated; the objective function is updated in the exploration and mining phases until the optimal candidate parameters are output;

[0021] Among them, the population matrix of pelicans is expressed as:

[0022]

[0023] X is the pelican population matrix, which is used to identify the pelican population members. Each row of the matrix represents a candidate solution, and each column represents the recommended value of the problem variable. iis the i-th pelican; i = 1, 2, ..., N; j = 1, 2, ..., m; N is the number of population members; m is the number of problem variables;

[0024] The value of the objective function is expressed as:

[0025]

[0026] B is the objective function vector; B i is the objective function value of the i-th candidate solution;

[0027] During the exploration phase, the pelican's movement towards the prey location is represented by:

[0028]

[0029] is the new state of the i-th pelican in the j-th dimension; a i,j is the value of the jth variable of the i-th candidate solution; I is a random number equal to 1 or 2; p j is the position of the target in the jth dimension; B p is the target function value; rand is the random number interval [0,1];

[0030] If the value of the objective function improves at that location, then the new position of the pelican is accepted, expressed as:

[0031]

[0032] is the new state of the i-th pelican; is the objective function value based on the exploration phase;

[0033] In the mining phase, the pelican's hunting process is represented as:

[0034]

[0035] is the new state of the i-th pelican in the mining phase; R = 0.2; for a i,j The neighborhood radius; t is the iteration counter; T is the maximum number of iterations;

[0036] During the mining phase, the new position of the pelican is represented by:

[0037]

[0038] is the new state of i pelicans; is the objective function value based on the mining stage.

[0039] Furthermore, the REGST stacked regression model theory includes:

[0040] Assume there are k base learners v1(x),v1(x),…,v k (x), use the same model training sample to train the model, expressed as: L = {(y n ,x n ),n=1,2,...,N};x n is the input vector of the nth base learner; y n is the output vector of the nth base learner;

[0041] The combination of n basic learners is expressed as: w k is the proportion of each basic learner in the combination; v k (x) is the k-th learner.

[0042] Furthermore, the first-level optimization scheduling model is established: the model predictive control strategy is introduced into the heating system equipped with a heat storage device, the scheduling instructions are set based on the output forecast value of the renewable energy heating unit, the combined output of the renewable energy heating unit and the heat storage device is used to track the scheduling instructions and the minimum amount of heat storage device action is used as the control optimization goal, the heat storage device is optimized through the model predictive control strategy, and the model predictive control is optimized and solved through rolling optimization and feedback correction links to obtain the action instructions of the heat storage device, including:

[0043] Configure a heat storage device in the heating system, and then introduce the model predictive control strategy into the heating system configured with the heat storage device;

[0044] Based on the output forecast value of renewable energy heating units, the average value of the forecast value in different time periods is used as the dispatch instruction of the dispatch cycle;

[0045] The control optimization objectives are to track the dispatching instructions of the combined output of the renewable energy heating unit and the heat storage device and to minimize the operation of the heat storage device;

[0046] The heat storage device is optimized through the model predictive control strategy. The model predictive control is optimized and solved through rolling optimization and feedback correction links to obtain the action instructions of the heat storage device.

[0047] Before optimizing the heat storage device through the model predictive control strategy, the power and energy balance of the heating system integrated with the renewable energy heating unit and the heat storage device is converted into a state space model, which can be expressed as:

[0048]

[0049] The state variable x(k) includes the output power of the renewable energy heating unit with heat storage and the remaining heat of the heat storage device at time k; the control variable u(k) includes the heat storage and release power increment of the heat storage device at time k; the system output variable y(k) includes the output power of the renewable energy heating unit with heat storage at time k; the uncontrollable variable d(k) for power calculation includes the output power increment of the renewable energy heating unit at time k; A, B1, B2, and C are known coefficient matrices; the output power of the renewable energy heating unit with heat storage includes the heat storage and release power of the heat storage device and the output power of the renewable energy heating unit;

[0050] The control optimization objectives are to track the dispatching instructions of the combined output of the renewable energy heating unit and the heat storage device and to minimize the operation of the heat storage device, which can be expressed as:

[0051]

[0052] y(k+j) is the predicted output; r(k+j) is the scheduling instruction of the system at time k+j in the future, j = 1, 2, ..., p; t w , t u are the error output weight coefficient and the incremental weight coefficient of the control variable respectively; m is the control time domain, m≤p;

[0053] The constraints are the operation constraints of the heat storage device, including the capacity constraints of the heat storage device and the power constraints of the heat storage and heat release of the heat storage device;

[0054] The heat storage device is optimized through the model predictive control strategy. The model predictive control is optimized and solved through rolling optimization and feedback correction links to obtain the action instructions of the heat storage device, including: according to the idea of ​​model predictive control, the system output variable value y(k) is measured at time k, the state variable x(k) at time k is inferred, and p time periods after time k are predicted to obtain y(k+1|k), y(k+2|k),..., y(k+p|k) in sequence, where y(k+p|k) is the system output response at time k+p predicted at time k, y(k+p|k) is solved to obtain the control variable u(k) at time k, and then u(k+1|k) is applied to the next moment. At time k+1, the solved u(k+1|k) and the measured y(k+1) are used for cyclic rolling optimization and feedback correction to obtain the optimal action instructions of the heat storage device;

[0055]

[0056] j=0,1,...,m-1; when m≤j≤p, u(k+j|k)=u(k+m-1|k); when 1≤j≤p, d(k+j|k)=d(k).

[0057] Furthermore, the second-level optimization scheduling model is established: the load of the traditional heating unit is reduced by the combined output of the renewable energy heating unit and the heat storage device, and the traditional heating unit is optimized by using the residual load. The objective function is to minimize the operating cost of the traditional heating unit, the operating cost of the renewable energy, and the operating cost of the heat storage device during the scheduling period. The output of the traditional heating unit is optimized through an intelligent optimization algorithm to obtain the action instructions of the traditional heating unit, including:

[0058] Taking the combined output of the renewable energy heating unit and the heat storage device as a known condition, in order to reduce the load of the traditional heating unit, the residual load is obtained by subtracting the combined output of the renewable energy heating unit and the heat storage device from the total load demand of the heating system, and the residual load is used to optimize the traditional heating unit;

[0059] The objective function is to minimize the operating costs of traditional heating units, renewable energy, and heat storage devices during the scheduling period, which can be expressed as:

[0060]

[0061] F i (t) is the operating cost of i different traditional heating units; F w (t) is the operating cost of w different renewable energy heating units; F b (t) is the operating cost of the heat storage device b; T is the optimal scheduling period;

[0062] Set power balance constraints for the heating system, output constraints for traditional heating units, operating constraints for heat storage devices, and operating constraints for renewable energy heating units;

[0063] The improved Harris Eagle optimization algorithm is used to optimize the output of the traditional heating unit and obtain the action instructions of the traditional heating unit.

[0064] Furthermore, the improved Harris Hawk optimization algorithm includes:

[0065] The tent chaotic map is used to improve the initial population, which is expressed as:

[0066]

[0067]

[0068] X i =lb+Y i (ub-lb);

[0069] i=1,2,3,…,N-1, the first individual of the population is randomly generated and mapped to the (0,1) space, denoted as Y1; the remaining N-1 individuals in the mapping space are denoted as Yi+1 ; Y i is the i-th individual in the mapping space; X i is the i-th individual in the initial population; ub and lb are the upper and lower limits of the population position;

[0070] Using nonlinear escape energy, it is expressed as:

[0071]

[0072] E0 is a random number (-1,1); t is the number of iterations; T is the maximum number of iterations;

[0073] The golden sine algorithm is used to update the position of the population, which is expressed as:

[0074]

[0075] R1 is a random number in [0,2π]; R2 is a random number in [0,π]; τ is the golden section number; R3 and R4 determine the global optimization ability of the algorithm; X(t) is the position of the t-th generation population.

[0076] Furthermore, when establishing the second-level optimization scheduling model, in addition to minimizing the operating cost of the traditional heating unit, the operating cost of renewable energy and the operating cost of the heat storage device during the scheduling period as the objective function, the minimization of the carbon emissions of the heating system can also be set as the objective function, and the minimization of the operating cost and carbon emissions of the heating system can be used as a multi-objective function. The multi-objective function is then converted into a single objective function, and an intelligent optimization algorithm is used to optimize the output of the traditional heating unit to obtain the action instructions of the traditional heating unit.

[0077] The present invention also provides a dual-layer optimization scheduling device for a heating system with MPC and heat storage, comprising:

[0078] A digital twin model establishment unit is used to establish a digital twin model of the heating system using a mechanism modeling and data identification method; the heating units in the heating system include at least a renewable energy heating unit and a traditional heating unit;

[0079] The prediction model training unit is used to obtain historical heating operation data and outdoor meteorological data based on the digital twin model of the heating system, and use machine learning algorithms to establish a heating system load prediction model and a renewable energy heating unit output prediction model to obtain the total load demand of the heating system and the output prediction values ​​of the renewable energy heating units;

[0080] A two-layer optimization scheduling model establishment unit is used to establish a first-layer optimization scheduling model: introduce a model predictive control strategy into a heating system equipped with a heat storage device, set a scheduling instruction based on the output forecast value of the renewable energy heating unit, track the scheduling instruction based on the combined output of the renewable energy heating unit and the heat storage device, and minimize the action of the heat storage device as the control optimization target, optimize the heat storage device through the model predictive control strategy, and optimize and solve the model predictive control through rolling optimization and feedback correction links to obtain the action instruction of the heat storage device; it is also used to establish a second-layer optimization scheduling model: reduce the load of the traditional heating unit through the combined output of the renewable energy heating unit and the heat storage device, optimize the traditional heating unit using the residual load, and minimize the operating cost of the traditional heating unit, the renewable energy operating cost and the heat storage device operating cost during the scheduling period as the objective function, optimize the output of the traditional heating unit through the intelligent optimization algorithm, and obtain the action instruction of the traditional heating unit.

[0081] The beneficial effects of the present invention are:

[0082] The present invention introduces the model predictive control strategy into the renewable energy heating unit equipped with a certain capacity heat storage device in the first-level optimization scheduling model, optimizes the heat storage device in the renewable energy heating unit through the model predictive control strategy, and takes the combined output of renewable energy and heat storage device to be close to the expected output set in advance and the minimum action amount of the heat storage device as the control optimization goal. The model predictive control is optimized and solved through rolling optimization and feedback correction links to obtain the action instructions of the heat storage device. The output of the heat storage device can effectively compensate for the output power of renewable energy, reducing the adverse effects, uncertainty and volatility brought by the uncertainty of renewable energy to the scheduling of the heating system. The combined output is schedulable, and the optimized combined output is equivalent to a benign schedulable heat source for the heating system. ; In the second-level optimization scheduling model, since the first-level optimization scheduling model obtains the output of renewable energy heating units and heat storage devices, when optimizing the output of traditional heating units, this part is taken as a known part, which is equivalent to subtracting an equal amount of load. The remaining load shortfall is filled by the output of traditional heating units. The output of traditional heating units is optimized using intelligent optimization algorithms as the second-level optimization of scheduling, which can improve the economy and reliability of the heating system. In addition, a digital twin model of the heating system is established, and a heating system load forecasting model and a renewable energy heating unit output forecasting model are established using machine learning algorithms to obtain accurate total load demand of the heating system and output forecast values ​​of renewable energy heating units, which establishes a data foundation for the two-level optimization scheduling of the heating system and improves the accuracy and effectiveness of the two-level optimization scheduling of the entire heating system.

[0083] Other features and advantages will be described in the following description, and in part will become apparent from the description, or understood by practicing the invention. The purpose and other advantages of the invention are realized and obtained by the structures particularly pointed out in the description and the drawings.

[0084] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0086] Figure 1 This is a flow chart of a two-layer optimization scheduling method for a heating system with heat storage based on MPC according to the present invention;

[0087] Figure 2 Schematic diagram of the structure of the double-layer optimization scheduling device of the heating system with heat storage based on MPC of the present invention. DETAILED DESCRIPTION

[0088] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0089] Example 1

[0090] Figure 1 This is a flow chart of a two-layer optimization scheduling method for a heating system with heat storage based on MPC involved in the present invention.

[0091] like Figure 1 As shown, this embodiment 1 provides a two-layer optimization scheduling method for a heating system with heat storage based on MPC, which includes:

[0092] A digital twin model of a heating system is established by using a mechanism modeling and data identification method; the heating units in the heating system include at least a renewable energy heating unit and a traditional heating unit;

[0093] Based on the digital twin model of the heating system, historical heating operation data and outdoor meteorological data are obtained. A machine learning algorithm is used to establish a heating system load forecasting model and a renewable energy heating unit output forecasting model to obtain the total load demand of the heating system and the output forecast values ​​of the renewable energy heating units.

[0094] Establish a first-level optimization scheduling model: Introduce a model predictive control strategy into a heating system equipped with a heat storage device. Set the scheduling instructions based on the output forecast of the renewable energy heating unit. Track the scheduling instructions based on the combined output of the renewable energy heating unit and the heat storage device, and minimize the amount of heat storage device movement as the control optimization objectives. Optimize the heat storage device using the model predictive control strategy. The model predictive control is optimized and solved through rolling optimization and feedback correction links to obtain the action instructions for the heat storage device.

[0095] Establish a second-level optimization scheduling model: reduce the load of traditional heating units through the joint output of renewable energy heating units and heat storage devices, use the remaining load to optimize traditional heating units, and take the minimization of traditional heating unit operating costs, renewable energy operating costs and heat storage device operating costs during the scheduling period as the objective function. Optimize the output of traditional heating units through intelligent optimization algorithms to obtain action instructions for traditional heating units.

[0096] It should be noted that in the first-level optimization scheduling model, the model predictive control strategy is introduced into the renewable energy heating unit equipped with a certain capacity heat storage device. The heat storage device in the renewable energy heating unit is optimized through the model predictive control strategy. The control optimization goal is to ensure that the combined output of renewable energy and heat storage device is close to the expected output set in advance and the action amount of heat storage device is minimized. The model predictive control is optimized and solved through rolling optimization and feedback correction links to obtain the action instructions of the heat storage device. The heat storage device processing can effectively compensate for the output power of renewable energy and reduce the uncertainty of renewable energy to the supply. The adverse effects, uncertainty and volatility brought by the scheduling of the heat system are eliminated, and the common output has the dispatchability. The optimized common output is equivalent to a benign dispatchable heat source for the heating system; in the second-level optimization scheduling model, since the first-level optimization scheduling model obtains the output of the renewable energy heating unit and the heat storage device, when optimizing the output of the traditional heating unit, this part is taken as a known part, which is equivalent to subtracting an equal amount of load. The remaining load shortfall is filled by the output of the traditional heating unit. The output of the traditional heating unit is optimized using an intelligent optimization algorithm as the second-level optimization of scheduling, which can improve the economy and reliability of the heating system.

[0097] In this embodiment, the method of using mechanism modeling and data identification to establish a digital twin model of the heating system includes:

[0098] Establish physical models, logical models, simulation models, and data models for renewable energy heating units, traditional heating units, primary networks, secondary networks, thermal power stations, and terminal buildings in the heating system. Couple the physical models, logical models, simulation models, and data models, and integrate them at multiple levels and scales. Establish a digital twin model of the heating system after mapping and reconstructing physical entities in the physical space in the virtual space.

[0099] An improved adaptive inertia weighted particle swarm optimization algorithm is used to optimize the parameters of the digital twin model of the heating system: the inertia weight nonlinear decreasing update strategy and mutation operation are combined and introduced into the PSO algorithm to form an improved adaptive inertia weighted particle swarm optimization algorithm; the parameters to be identified in the digital twin model of the heating system are determined, the value range of each parameter is set, and each parameter optimization problem is converted into a particle position optimization problem; the position variable is introduced, and the deviation of the digital twin model of the heating system is solved with the help of the training sample data obtained in the heating system. The improved adaptive inertia weighted particle swarm optimization algorithm is used to select the optimal position of the particle according to the size of the deviation value to obtain the optimal parameters of the digital twin model of the heating system; the root mean square error and mean absolute percentage error are selected as measurement indicators to verify the performance of the digital twin model of the heating system.

[0100] In this embodiment, the heating system digital twin model is used to obtain historical heating operation data and outdoor meteorological data, and a machine learning algorithm is used to establish a heating system load forecasting model to obtain the total load demand of the heating system and the output forecast value of the renewable energy heating unit, including:

[0101] Based on the digital twin model of the heating system, the historical output of renewable energy units, the output of traditional heating units, the heating system load, the operating parameters of the heating station, the outdoor temperature, humidity, and the indoor temperature of the terminal building are obtained as training samples for the heating system load prediction model; the historical output of renewable energy units, the operating parameters of the heating station, the outdoor temperature, humidity, and the indoor temperature of the terminal building are obtained as training samples for the renewable energy heating unit output prediction model;

[0102] After optimizing the CNN-REGST model parameters using the POA Pelican optimization algorithm, the optimized CNN-REGST model was used to establish a heating system load forecasting model and a renewable energy heating unit output forecasting model: after extracting features from the model training samples using the CNN convolutional neural network model, the REGST stacked regression model theory was used to input the feature-extracted data into multiple basic learners respectively to obtain multiple heating system load forecast values ​​and multiple renewable energy heating unit output forecast values. The SVM model was then used as a meta-regressor to calculate the multiple forecast values ​​to obtain the heating system load forecast value and the renewable energy heating unit output forecast value.

[0103] In this embodiment, the CNN-REGST model parameters are optimized using the POA Pelican optimization algorithm, including:

[0104] Initialize the CNN-REGST model parameters, including the number of neurons, learning rate, and the number of nodes and filters in the fully connected layer;

[0105] The CNN-REGST model parameters are optimized using the POA Pelican optimization algorithm: the number of Pelican population members and the maximum number of iterations are initialized; the initial population is generated and the objective function is calculated; the objective function is updated in the exploration and mining phases until the optimal candidate parameters are output;

[0106] Among them, the population matrix of pelicans is expressed as:

[0107]

[0108] X is the pelican population matrix, which is used to identify the pelican population members. Each row of the matrix represents a candidate solution, and each column represents the recommended value of the problem variable. i is the i-th pelican; i = 1, 2, ..., N; j = 1, 2, ..., m; N is the number of population members; m is the number of problem variables;

[0109] The value of the objective function is expressed as:

[0110]

[0111] B is the objective function vector; B i is the objective function value of the i-th candidate solution;

[0112] During the exploration phase, the pelican's movement towards the prey location is represented by:

[0113]

[0114] is the new state of the i-th pelican in the j-th dimension; a i,j is the value of the jth variable of the i-th candidate solution; I is a random number equal to 1 or 2; p j is the position of the target in the jth dimension; B p is the target function value; rand is the random number interval [0,1];

[0115] If the value of the objective function improves at that location, then the new position of the pelican is accepted, expressed as:

[0116]

[0117] is the new state of the i-th pelican; is the objective function value based on the exploration phase;

[0118] In the mining phase, the pelican's hunting process is represented as:

[0119]

[0120] is the new state of the i-th pelican in the mining phase; R = 0.2; for a i,j The neighborhood radius; t is the iteration counter; T is the maximum number of iterations;

[0121] During the mining phase, the new position of the Pelican is represented by:

[0122]

[0123] is the new state of i pelicans; is the objective function value based on the mining stage.

[0124] It should be noted that CNN is trained to extract prediction features, and REGST is used as the regression operator algorithm. The hybrid model has more accurate prediction accuracy than a single model. The Pelican Optimization Algorithm (POA) is used to optimize the CNN-REGST model parameters, which improves the generalization ability and practical operability of the model.

[0125] In this embodiment, the REGST stacked regression model theory includes:

[0126] Assume there are k base learners v1(x),v1(x),...,v k (x), use the same model training sample to train the model, expressed as: L = {(y n ,x n ),n=1,2,...,N};x n is the input vector of the nth base learner; y n is the output vector of the nth base learner;

[0127] The combination of n basic learners is expressed as: w k is the proportion of each basic learner in the combination; v k (x) is the k-th learner.

[0128] In this embodiment, the first-level optimization scheduling model is established: a model predictive control strategy is introduced into a heating system equipped with a heat storage device, a scheduling instruction is set based on the output forecast value of the renewable energy heating unit, and the combined output of the renewable energy heating unit and the heat storage device is used to track the scheduling instruction and minimize the amount of heat storage device movement as the control optimization goal. The heat storage device is optimized using the model predictive control strategy. The model predictive control is optimized and solved through rolling optimization and feedback correction links to obtain the action instructions of the heat storage device, including:

[0129] Configure a heat storage device in the heating system, and then introduce the model predictive control strategy into the heating system configured with the heat storage device;

[0130] Based on the output forecast value of renewable energy heating units, the average value of the forecast value in different time periods is used as the dispatch instruction of the dispatch cycle;

[0131] The control optimization objectives are to track the dispatching instructions of the combined output of the renewable energy heating unit and the heat storage device and to minimize the operation of the heat storage device;

[0132] The heat storage device is optimized through the model predictive control strategy. The model predictive control is optimized and solved through rolling optimization and feedback correction links to obtain the action instructions of the heat storage device.

[0133] Before optimizing the heat storage device through the model predictive control strategy, the power and energy balance of the heating system integrated with the renewable energy heating unit and the heat storage device is converted into a state space model, which can be expressed as:

[0134]

[0135] The state variable x(k) includes the output power of the renewable energy heating unit with heat storage and the remaining heat of the heat storage device at time k; the control variable u(k) includes the heat storage and release power increment of the heat storage device at time k; the system output variable y(k) includes the output power of the renewable energy heating unit with heat storage at time k; the uncontrollable variable d(k) for power calculation includes the output power increment of the renewable energy heating unit at time k; A, B1, B2, and C are known coefficient matrices; the output power of the renewable energy heating unit with heat storage includes the heat storage and release power of the heat storage device and the output power of the renewable energy heating unit;

[0136] The control optimization objectives are to track the dispatching instructions of the combined output of the renewable energy heating unit and the heat storage device and to minimize the operation of the heat storage device, which can be expressed as:

[0137]

[0138] y(k+j) is the predicted output; r(k+j) is the scheduling instruction of the system at time k+j in the future, j = 1, 2, ..., p; t w , t u are the error output weight coefficient and the incremental weight coefficient of the control variable respectively; m is the control time domain, m≤p;

[0139] The constraints are the operation constraints of the heat storage device, including the capacity constraints of the heat storage device and the power constraints of the heat storage and heat release of the heat storage device;

[0140] The heat storage device is optimized through the model predictive control strategy. The model predictive control is optimized and solved through rolling optimization and feedback correction links to obtain the action instructions of the heat storage device, including: according to the idea of ​​model predictive control, the system output variable value y(k) is measured at time k, the state variable x(k) at time k is inferred, and p time periods after time k are predicted to obtain y(k+1|k), y(k+2|k),..., y(k+p|k) in sequence, where y(k+p|k) is the system output response at time k+p predicted at time k, y(k+p|k) is solved to obtain the control variable u(k) at time k, and then u(k+1|k) is applied to the next moment. At time k+1, the solved u(k+1|k) and the measured y(k+1) are used for cyclic rolling optimization and feedback correction to obtain the optimal action instructions of the heat storage device;

[0141]

[0142] j=0,1,...,m-1; when m≤j≤p, u(k+j|k)=u(k+m-1|k); when 1≤j≤p, d(k+j|k)=d(k).

[0143] It should be noted that MPC has an additional rolling optimization link compared to traditional control. The difference from traditional control is that MPC does not solve a fixed and unique global optimal solution, but the optimal solution at each moment. Its optimization process is not completed once, but is performed repeatedly online. The optimization of its performance indicators only considers the period from the current moment to a finite moment in the future, and at the next moment, this optimization process will be repeated, that is, the optimization period moves forward. By solving the optimal solution at each moment, it can make up for the errors caused by factors such as model mismatch and environmental interference, so that the final optimization control result can be closer to reality. Feedback correction: MPC has the characteristics of rolling optimization. At each moment, the optimal control sequence from the current moment to a certain finite time in the future is obtained. In order to avoid large errors in system control caused by some influencing factors, the forward rolling optimization is repeated at each moment. At each moment, only the optimal control solution at the current moment is used. That is, each time control is implemented, only the first control variable in the control sequence is used to predict the output value at the future moment. By introducing the real-time measurement information of the actual system as the feedback value, the rolling optimization process of model prediction is no longer just based on model optimization, but also utilizes the actual measurement information of the system to form a closed-loop rolling optimization feedback system.

[0144] In this embodiment, the second-level optimization scheduling model is established: the load of the traditional heating unit is reduced by the combined output of the renewable energy heating unit and the heat storage device, and the residual load is used to optimize the traditional heating unit. The objective function is to minimize the operating cost of the traditional heating unit, the operating cost of the renewable energy, and the operating cost of the heat storage device during the scheduling period. The output of the traditional heating unit is optimized through an intelligent optimization algorithm to obtain the action instructions of the traditional heating unit, including:

[0145] Taking the combined output of the renewable energy heating unit and the heat storage device as a known condition, in order to reduce the load of the traditional heating unit, the residual load is obtained by subtracting the combined output of the renewable energy heating unit and the heat storage device from the total load demand of the heating system, and the residual load is used to optimize the traditional heating unit;

[0146] The objective function is to minimize the operating costs of traditional heating units, renewable energy, and heat storage devices during the scheduling period, which can be expressed as:

[0147]

[0148] F i (t) is the operating cost of i different traditional heating units; F w (t) is the operating cost of w different renewable energy heating units; F b (t) is the operating cost of the heat storage device b; T is the optimal scheduling period;

[0149] Set power balance constraints for the heating system, output constraints for traditional heating units, operating constraints for heat storage devices, and operating constraints for renewable energy heating units;

[0150] The improved Harris Eagle optimization algorithm is used to optimize the output of the traditional heating unit and obtain the action instructions of the traditional heating unit.

[0151] In this embodiment, the improved Harris Hawk optimization algorithm includes:

[0152] The tent chaotic map is used to improve the initial population, which is expressed as:

[0153]

[0154]

[0155] X i =lb+Y i (ub-lb);

[0156] i=1,2,3,...,N-1, the first individual of the population is randomly generated and mapped to the (0,1) space, denoted as Y1; the remaining N-1 individuals in the mapping space are denoted as Y i+1 ; Y i is the i-th individual in the mapping space; X i is the i-th individual in the initial population; ub and lb are the upper and lower limits of the population position;

[0157] Using nonlinear escape energy, it is expressed as:

[0158]

[0159] E0 is a random number (-1,1); t is the number of iterations; T is the maximum number of iterations;

[0160] The golden sine algorithm is used to update the position of the population, which is expressed as:

[0161]

[0162] R1 is a random number in [0,2π]; R2 is a random number in [0,π]; τ is the golden section number; R3 and R4 determine the global optimization ability of the algorithm; X(t) is the position of the t-th generation population.

[0163] It should be noted that the Harris Hawk optimization algorithm includes three stages: global exploration stage, transition from exploration stage to exploitation stage, and local exploitation stage; global exploration stage: Harris Hawk is a raptor and a social animal, usually hunting in groups. When the absolute value of the prey's escape energy is greater than 1, the prey is physically strong and the distance between the Harris Hawk and the prey is far, and the Harris Hawk is in the stage of exploring the prey; transition from exploration stage to exploitation stage: Harris Hawk will transition between different hunting stages according to the different escape energies of the prey. When the absolute value of the escape energy is greater than or equal to 1, it enters the global exploration stage; otherwise, it enters the local exploitation stage; local exploitation stage: according to the prey's escape behavior and the Harris Hawk's chasing strategy There are four possible attack strategies. When the escape probability is less than 0.5, it indicates that the rabbit successfully escaped before the attack; when the escape probability is greater than or equal to 0.5, it indicates that the rabbit failed to escape; when the absolute value of the escape energy is greater than or equal to 0.5 and less than 1, and the escape probability is greater than or equal to 0.5, the Harris Hawk adopts a soft encirclement strategy; when the absolute value of the escape energy is less than 0.5 and the escape probability is greater than or equal to 0.5, the hard encirclement strategy is adopted; when the absolute value of the escape energy is greater than or equal to 0.5 and less than 1, and the escape probability is less than 0.5, the Harris Hawk adopts a gradual dive and rapid soft encirclement strategy; when the absolute value of the escape energy is less than 0.5 and the escape probability is less than 0.5, the Harris Hawk adopts a gradual dive and rapid hard encirclement strategy.

[0164] The soft encirclement strategy refers to the situation where the prey fails to escape the encirclement but still has enough energy to escape. At this stage, the prey will use jumping movements to confuse the Harris's Hawk; the hard encirclement strategy refers to the situation where the prey fails to escape the encirclement and does not have enough energy to escape. At this time, the Harris's Hawk adopts a hard encirclement strategy and carries out a surprise attack; the progressive dive and fast soft encirclement strategy refers to the situation where the prey successfully escapes the Harris's Hawk's encirclement and has enough energy to escape successfully. The Harris's Hawk will dive attack the prey. If the dive attack fails, the Harris's Hawk will conduct an irregular dive attack based on Levy flight on the prey; the progressive dive and fast hard encirclement strategy refers to the situation where the prey successfully escapes the encirclement but does not have enough energy to escape. The Harris's Hawk will first shorten the average distance with the prey, and then besiege the prey. If the siege fails, it will conduct an irregular attack.

[0165] The original Harris Hawk optimization algorithm generates the initial population randomly, which is not conducive to rapid convergence. However, by introducing the Tent chaos map, a more evenly distributed initial population is obtained, accelerating the convergence of the algorithm. In the original Harris Hawk optimization algorithm, the escape energy decreases linearly with the number of iterations. However, the size of the escape energy directly affects the global exploration and local development capabilities of the Harris Hawk optimization algorithm. Using a nonlinear escape energy is more conducive to the convergence of the algorithm. The golden sine algorithm, which has strong global optimization capabilities, is used to update the position of the population and escape from the local optimum. By improving the Harris Hawk optimization algorithm, the ability to escape from the local optimum is improved, the convergence speed of the algorithm is accelerated, the maximum number of iterations is significantly reduced, and the stability of the algorithm is enhanced.

[0166] In this embodiment, when establishing the second-level optimization scheduling model, in addition to minimizing the operating cost of the traditional heating unit, the operating cost of renewable energy, and the operating cost of the heat storage device during the scheduling period as the objective function, the minimization of the carbon emissions of the heating system can also be set as the objective function, and the minimization of the operating cost and carbon emissions of the heating system can be used as a multi-objective function. The multi-objective function is then converted into a single objective function, and an intelligent optimization algorithm is used to optimize the output of the traditional heating unit to obtain the action instructions of the traditional heating unit.

[0167] Example 2

[0168] Figure 2 This is a structural diagram of a double-layer optimization scheduling device for a heating system with heat storage based on MPC involved in the present invention.

[0169] like Figure 2 As shown, this embodiment 2 provides a two-layer optimization scheduling device for a heating system with MPC and heat storage, which includes:

[0170] A digital twin model establishment unit is used to establish a digital twin model of the heating system using a mechanism modeling and data identification method; the heating units in the heating system include at least a renewable energy heating unit and a traditional heating unit;

[0171] The prediction model training unit is used to obtain historical heating operation data and outdoor meteorological data based on the digital twin model of the heating system, and use machine learning algorithms to establish a heating system load prediction model and a renewable energy heating unit output prediction model to obtain the total load demand of the heating system and the output prediction values ​​of the renewable energy heating units;

[0172] A two-layer optimization scheduling model establishment unit is used to establish a first-layer optimization scheduling model: introduce a model predictive control strategy into a heating system equipped with a heat storage device, set a scheduling instruction based on the output forecast value of the renewable energy heating unit, track the scheduling instruction based on the combined output of the renewable energy heating unit and the heat storage device, and minimize the action of the heat storage device as the control optimization target, optimize the heat storage device through the model predictive control strategy, and optimize and solve the model predictive control through rolling optimization and feedback correction links to obtain the action instruction of the heat storage device; it is also used to establish a second-layer optimization scheduling model: reduce the load of the traditional heating unit through the combined output of the renewable energy heating unit and the heat storage device, optimize the traditional heating unit using the residual load, and minimize the operating cost of the traditional heating unit, the renewable energy operating cost and the heat storage device operating cost during the scheduling period as the objective function, optimize the output of the traditional heating unit through the intelligent optimization algorithm, and obtain the action instruction of the traditional heating unit.

[0173] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can also be implemented in other ways. The system embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions, and operations of the systems, methods, and computer program products according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, a program segment, or a portion of code, and the module, program segment, or a portion of code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the boxes can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified functions or actions, or can be implemented using a combination of dedicated hardware and computer instructions.

[0174] In addition, the functional modules in each embodiment of the present invention can be integrated together to form an independent part, or each module can exist separately, or two or more modules can be integrated to form an independent part. If the function is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0175] With the above-described preferred embodiments of the present invention as a guide, and with reference to the above description, relevant personnel are fully capable of making various changes and modifications without departing from the technical scope of this invention. The technical scope of this invention is not limited to the contents of the specification and must be determined according to the scope of the claims.

Claims

1. A two-layer optimization scheduling method for a heating system with heat storage based on MPC, characterized in that: It includes: A digital twin model of a heating system is established by using a mechanism modeling and data identification method; the heating units in the heating system include at least a renewable energy heating unit and a traditional heating unit; Based on the heating system digital twin model, historical heating operation data and outdoor meteorological data are obtained. A machine learning algorithm is used to establish a heating system load forecasting model and a renewable energy heating unit output forecasting model. The total heating system load demand and renewable energy heating unit output forecast values ​​are obtained, including: Based on the digital twin model of the heating system, the historical output of renewable energy units, the output of traditional heating units, the heating system load, the operating parameters of the heating station, the outdoor temperature, humidity, and the indoor temperature of the terminal building are obtained as training samples for the heating system load prediction model; the historical output of renewable energy units, the operating parameters of the heating station, the outdoor temperature, humidity, and the indoor temperature of the terminal building are obtained as training samples for the renewable energy heating unit output prediction model; After optimizing the CNN-REGST model parameters using the POA Pelican optimization algorithm, the optimized CNN-REGST model was used to establish a heating system load forecasting model and a renewable energy heating unit output forecasting model: after extracting features from the model training samples using the CNN convolutional neural network model, the REGST stacked regression model theory was used to input the feature-extracted data into multiple basic learners to obtain multiple heating system load forecast values ​​and multiple renewable energy heating unit output forecast values. The SVM model was then used as a meta-regressor to calculate the multiple forecast values ​​to obtain the heating system load forecast value and the renewable energy heating unit output forecast value; Establish a first-level optimization scheduling model: Introduce a model predictive control strategy into a heating system equipped with a heat storage device. Set the scheduling instructions based on the output forecast of the renewable energy heating unit. Track the scheduling instructions based on the combined output of the renewable energy heating unit and the heat storage device, and minimize the amount of heat storage device movement as the control optimization objectives. Optimize the heat storage device using the model predictive control strategy. The model predictive control is optimized and solved through rolling optimization and feedback correction links to obtain the action instructions for the heat storage device. Establish a second-level optimization scheduling model: reduce the load of traditional heating units through the joint output of renewable energy heating units and heat storage devices, use the remaining load to optimize traditional heating units, and take the minimization of traditional heating unit operating costs, renewable energy operating costs and heat storage device operating costs during the scheduling period as the objective function. Optimize the output of traditional heating units through intelligent optimization algorithms to obtain action instructions for traditional heating units.

2. The double-layer optimization scheduling method for a heating system according to claim 1 is characterized in that: The method of mechanism modeling and data identification is used to establish a digital twin model of the heating system, including: Establish physical models, logical models, simulation models, and data models for renewable energy heating units, traditional heating units, primary networks, secondary networks, thermal power stations, and terminal buildings in the heating system. Couple the physical models, logical models, simulation models, and data models, and integrate them at multiple levels and scales. Establish a digital twin model of the heating system after mapping and reconstructing physical entities in the physical space in the virtual space. An improved adaptive inertia weighted particle swarm optimization algorithm is used to optimize the parameters of the digital twin model of the heating system: the inertia weight nonlinear decreasing update strategy and mutation operation are combined and introduced into the PSO algorithm to form an improved adaptive inertia weighted particle swarm optimization algorithm; the parameters to be identified in the digital twin model of the heating system are determined, the value range of each parameter is set, and each parameter optimization problem is converted into a particle position optimization problem; the position variable is introduced, and the deviation value of the digital twin model of the heating system is solved with the help of the training sample data obtained in the heating system. The improved adaptive inertia weighted particle swarm optimization algorithm is used to select the optimal position of the particle according to the size of the deviation value to obtain the optimal parameters of the digital twin model of the heating system; the root mean square error and mean absolute percentage error are selected as measurement indicators to verify the performance of the digital twin model of the heating system.

3. The double-layer optimization scheduling method for a heating system according to claim 1 is characterized in that: The POA Pelican optimization algorithm is used to optimize the CNN-REGST model parameters, including: Initialize the CNN-REGST model parameters, including the number of neurons, learning rate, and the number of nodes and filters in the fully connected layer; The CNN-REGST model parameters are optimized using the POA Pelican optimization algorithm: the number of Pelican population members and the maximum number of iterations are initialized; the initial population is generated and the objective function is calculated; the objective function is updated in the exploration and mining phases until the optimal candidate parameters are output; Among them, the population matrix of pelicans is expressed as: X is the pelican population matrix, which is used to identify the pelican population members. Each row of the matrix represents a candidate solution, and each column represents the recommended value of the problem variable. i is the i-th pelican; i = 1, 2, ..., N; j = 1, 2, ..., m; N is the number of population members; m is the number of problem variables; The value of the objective function is expressed as: B is the objective function vector; B i is the objective function value of the i-th candidate solution; During the exploration phase, the pelican's movement towards the prey location is represented by: is the new state of the i-th pelican in the j-th dimension; a i,j is the value of the jth variable of the i-th candidate solution; I is a random number equal to 1 or 2; p j is the position of the target in the jth dimension; B p is the target function value; rand is the random number interval [0,1]; If the value of the objective function improves at that location, then the new position of the pelican is accepted, expressed as: is the new state of the i-th pelican; is the objective function value based on the exploration phase; In the mining phase, the pelican's hunting process is represented as: is the new state of the i-th pelican in the mining phase; R = 0.2; for a i,j The neighborhood radius; t is the iteration counter; T is the maximum number of iterations; During the mining phase, the new position of the pelican is represented by: is the new state of i pelicans; is the objective function value based on the mining stage.

4. The double-layer optimization scheduling method for a heating system according to claim 1, characterized in that: The REGST stacked regression model theory includes: Assume there are k base learners v1(x),v1(x),…,v k (x), use the same model training sample to train the model, expressed as: L = {(y n ,x n ),n=1,2,...,N};x n is the input vector of the nth base learner; y n is the output vector of the nth base learner; The combination of n basic learners is expressed as: w k is the proportion of each basic learner in the combination; v k (x) is the k-th learner.

5. The double-layer optimization scheduling method for a heating system according to claim 1 is characterized in that: The first-level optimization scheduling model is established: a model predictive control strategy is introduced into a heating system equipped with a heat storage device, a scheduling instruction is set based on the output forecast value of the renewable energy heating unit, and the combined output of the renewable energy heating unit and the heat storage device is used to track the scheduling instruction and minimize the amount of heat storage device movement as the control optimization goal. The heat storage device is optimized using the model predictive control strategy. The model predictive control is optimized and solved through rolling optimization and feedback correction links to obtain the action instruction of the heat storage device, including: Configure a heat storage device in the heating system, and then introduce the model predictive control strategy into the heating system configured with the heat storage device; Through the output forecast value of renewable energy heating units, the average value of the forecast value in different time periods is used as the dispatch instruction of the dispatch cycle; The control optimization objectives are to track the dispatching instructions of the combined output of the renewable energy heating unit and the heat storage device and to minimize the operation of the heat storage device; The heat storage device is optimized through the model predictive control strategy. The model predictive control is optimized and solved through rolling optimization and feedback correction links to obtain the action instructions of the heat storage device. Before optimizing the heat storage device through the model predictive control strategy, the power and energy balance of the heating system integrated with the renewable energy heating unit and the heat storage device is converted into a state space model, which can be expressed as: The state variable x(k) includes the output power of the renewable energy heating unit with heat storage and the remaining heat of the heat storage device at time k; the control variable u(k) includes the heat storage and release power increment of the heat storage device at time k; the system output variable y(k) includes the output power of the renewable energy heating unit with heat storage at time k; the uncontrollable variable d(k) for power calculation includes the output power increment of the renewable energy heating unit at time k; A, B1, B2, and C are known coefficient matrices; the output power of the renewable energy heating unit with heat storage includes the heat storage and release power of the heat storage device and the output power of the renewable energy heating unit; The control optimization objectives are to track the dispatching instructions of the combined output of the renewable energy heating unit and the heat storage device and to minimize the operation of the heat storage device, which can be expressed as: y(k+j) is the predicted output; r(k+j) is the scheduling instruction of the system at time k+j in the future, j = 1, 2, ..., p; t w , t u are the error output weight coefficient and the incremental weight coefficient of the control variable respectively; m is the control time domain, m≤p; The constraints are the operation constraints of the heat storage device, including the capacity constraints of the heat storage device and the power constraints of the heat storage and heat release of the heat storage device; The heat storage device is optimized through the model predictive control strategy. The model predictive control is optimized and solved through rolling optimization and feedback correction links to obtain the action instructions of the heat storage device, including: according to the idea of ​​model predictive control, the system output variable value y(k) is measured at time k, the state variable x(k) at time k is inferred, and p time periods after time k are predicted to obtain y(k+1|k), y(k+2|k),..., y(k+p|k) in sequence, where y(k+p|k) is the system output response at time k+p predicted at time k, y(k+p|k) is solved to obtain the control variable u(k) at time k, and then u(k+1|k) is applied to the next moment. At time k+1, the solved u(k+1|k) and the measured y(k+1) are used for cyclic rolling optimization and feedback correction to obtain the optimal action instructions of the heat storage device; j=0,1,...,m-1; when m≤j≤p, u(k+j|k)=u(k+m-1|k); when 1≤j≤p, d(k+j|k)=d(k).

6. The double-layer optimization scheduling method for a heating system according to claim 1 is characterized in that: The second-level optimization scheduling model is established: the load of the traditional heating unit is reduced by the combined output of the renewable energy heating unit and the heat storage device, and the residual load is used to optimize the traditional heating unit. The objective function is to minimize the operating cost of the traditional heating unit, the operating cost of the renewable energy, and the operating cost of the heat storage device during the scheduling period. The output of the traditional heating unit is optimized through an intelligent optimization algorithm to obtain the action instructions of the traditional heating unit, including: Taking the combined output of the renewable energy heating unit and the heat storage device as a known condition, in order to reduce the load of the traditional heating unit, the residual load is obtained by subtracting the combined output of the renewable energy heating unit and the heat storage device from the total load demand of the heating system, and the residual load is used to optimize the traditional heating unit; The objective function is to minimize the operating costs of traditional heating units, renewable energy, and heat storage devices during the scheduling period, which can be expressed as: F i (t) is the operating cost of i different traditional heating units; F w (t) is the operating cost of w different renewable energy heating units; F b (t) is the operating cost of the heat storage device b; T is the optimal scheduling period; Set power balance constraints for the heating system, output constraints for traditional heating units, operating constraints for heat storage devices, and operating constraints for renewable energy heating units; The improved Harris Eagle optimization algorithm is used to optimize the output of the traditional heating unit and obtain the action instructions of the traditional heating unit.

7. The double-layer optimization scheduling method for a heating system according to claim 6, characterized in that: The improved Harris Hawk optimization algorithm includes: The tent chaotic map is used to improve the initial population, which is expressed as: X i =lb+Y i (ub-lb); i=1,2,3,…,N-1, the first individual of the population is randomly generated and mapped to the (0,1) space, denoted as Y1; the remaining N-1 individuals in the mapping space are denoted as Y i+1 ; Y i is the i-th individual in the mapping space; X i is the i-th individual in the initial population; ub and lb are the upper and lower limits of the population position; Using nonlinear escape energy, it is expressed as: E0 is a random number (-1,1); t is the number of iterations; T is the maximum number of iterations; The golden sine algorithm is used to update the position of the population, which is expressed as: R1 is a random number in [0,2π]; R2 is a random number in [0,π]; τ is the golden section number; R3 and R4 determine the global optimization ability of the algorithm; X(t) is the position of the t-th generation population.

8. The double-layer optimization scheduling method for a heating system according to claim 1 is characterized in that: When establishing the second-level optimization scheduling model, in addition to setting the minimization of the operating cost of the traditional heating unit, the operating cost of renewable energy and the operating cost of the heat storage device during the scheduling period as the objective function, the minimization of the carbon emissions of the heating system can also be set as the objective function, and the minimization of the operating cost and carbon emissions of the heating system can be set as a multi-objective function. The multi-objective function is then converted into a single-objective function, and an intelligent optimization algorithm is used to optimize the output of the traditional heating unit to obtain the action instructions of the traditional heating unit.

9. A dual-layer optimization scheduling device for a heating system with MPC and heat storage, characterized in that: It includes: A digital twin model establishment unit is used to establish a digital twin model of the heating system using a mechanism modeling and data identification method; the heating units in the heating system include at least a renewable energy heating unit and a traditional heating unit; The prediction model training unit is used to obtain historical heating operation data and outdoor meteorological data based on the digital twin model of the heating system, and use machine learning algorithms to establish a heating system load prediction model and a renewable energy heating unit output prediction model to obtain the total load demand of the heating system and the output prediction values ​​of the renewable energy heating units, including: Based on the digital twin model of the heating system, the historical output of renewable energy units, the output of traditional heating units, the heating system load, the operating parameters of the heating station, the outdoor temperature, humidity, and the indoor temperature of the terminal building are obtained as training samples for the heating system load prediction model; the historical output of renewable energy units, the operating parameters of the heating station, the outdoor temperature, humidity, and the indoor temperature of the terminal building are obtained as training samples for the renewable energy heating unit output prediction model; After optimizing the CNN-REGST model parameters using the POA Pelican optimization algorithm, the optimized CNN-REGST model was used to establish a heating system load forecasting model and a renewable energy heating unit output forecasting model: after extracting features from the model training samples using the CNN convolutional neural network model, the REGST stacked regression model theory was used to input the feature-extracted data into multiple basic learners to obtain multiple heating system load forecast values ​​and multiple renewable energy heating unit output forecast values. The SVM model was then used as a meta-regressor to calculate the multiple forecast values ​​to obtain the heating system load forecast value and the renewable energy heating unit output forecast value; A two-layer optimization scheduling model establishment unit is used to establish a first-layer optimization scheduling model: introduce a model predictive control strategy into a heating system equipped with a heat storage device, set a scheduling instruction based on the output forecast value of the renewable energy heating unit, track the scheduling instruction based on the combined output of the renewable energy heating unit and the heat storage device, and minimize the action of the heat storage device as the control optimization target, optimize the heat storage device through the model predictive control strategy, and optimize and solve the model predictive control through rolling optimization and feedback correction links to obtain the action instruction of the heat storage device; it is also used to establish a second-layer optimization scheduling model: reduce the load of the traditional heating unit through the combined output of the renewable energy heating unit and the heat storage device, optimize the traditional heating unit using the residual load, and minimize the operating cost of the traditional heating unit, the renewable energy operating cost and the heat storage device operating cost during the scheduling period as the objective function, optimize the output of the traditional heating unit through the intelligent optimization algorithm, and obtain the action instruction of the traditional heating unit.

Images