ICES model prediction control method and device based on GRU driving dynamic energy hub

Abstract

The invention discloses an ICES model prediction control method and device based on a GRU driving dynamic energy hub, and the method comprises the steps: carrying out the modeling of a coupling relation between the input and output of an ICES, and obtaining a DEH model; constructing an equipment efficiency parameter correction model based on the GRU network and performing training; inputting the input matrix in a set time period into the equipment efficiency parameter correction model for correction processing, and outputting an equipment efficiency parameter prediction matrix; a G-DEH model of the ICES is constructed, and an ICES optimization scheduling model is constructed based on the G-DEH model; updating the efficiency parameters of the G-DEH model in the previous time period according to the equipment efficiency parameter prediction matrix; solving the ICES optimization scheduling model in a set time period to obtain a scheduling strategy; according to the scheduling strategy, the ICES performs scheduling execution; and after the scheduling execution is completed, the ICES updates the set parameters and optimizes the scheduling strategy in the next time period. According to the invention, an optimal scheduling strategy with real-time response capability is provided, and the operation economy and scheduling precision of the ICES are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of integrated energy system optimal scheduling, and particularly to a method and device for model predictive control of an ICES based on a GRU-driven dynamic energy hub. Background Art

[0002] With the intensification of energy shortage and ecological crisis problems, the high-carbon energy structure is difficult to meet the sustainable development needs, and the low-carbon transformation has become the core task of the current energy system development. The Integrated Community Energy System (ICES) can significantly improve the energy utilization efficiency and cleanliness through the coupling and coordination of multi-energy devices. However, at present, many ICES projects have not achieved the expected economic benefits in actual operation, and the energy use cost is still relatively high.

[0003] Existing research shows that the optimal scheduling technology is the key to improving the operation economy and energy efficiency of ICES. As a general modeling theory describing the coupling relationship between the supply, conversion, storage, and utilization processes of multiple energies, the Energy Hub (EH) has now been widely applied to the modeling and optimal scheduling of ICES. A large number of studies have shown that the matching degree between the device efficiency parameters and the actual operating conditions significantly affects the accuracy of the EH model. Due to this matching relationship, the energy conversion devices in ICES often exhibit variable operating condition characteristics, that is, the device efficiency parameters deviate from the rated efficiency as the actual operating conditions change; if this deviation in device efficiency is ignored, it may affect the entire ICES system through the energy coupling relationship, thereby reducing the overall accuracy of the EH model. From the perspective of dynamic time series, the change of device efficiency parameters with operating conditions shows significant temporal correlation, that is, the device efficiency parameters not only depend on the current operating conditions but also are affected by the previous system state, which poses a challenge to the efficient and economic operation of ICES. Traditional optimization methods usually assume that the device efficiency is constant or approximate it through piecewise linearization, without fully considering the variable operating condition characteristics and temporal correlation of EH, which restricts the effectiveness of optimal scheduling. In addition, the real-time fluctuations of load demand and renewable energy output also pose severe challenges to the stability and economy of ICES scheduling.

[0004] Therefore, how to invent a model predictive control method for an integrated community energy system to improve the operation economy and scheduling accuracy of ICES has become an urgent problem to be solved. Summary of the Invention

[0005] To this end, the present invention provides an ICES model predictive control method and device based on a GRU-driven dynamic energy hub, which combines data-driven and MPC strategies, and proposes an optimized scheduling strategy with real-time response ability on the basis of fully considering the temporal correlation between device efficiency parameters and operating conditions, improving the operating economy and scheduling accuracy of ICES.

[0006] To achieve the above object, the present invention provides the following technical solutions: An ICES model predictive control method based on a GRU-driven dynamic energy hub, comprising:

[0007] Modeling the coupling relationship between the input and output of ICES based on the coupling matrix of the multivariable nonlinear system to obtain the DEH model;

[0008] Constructing a device efficiency parameter correction model based on the GRU network; training the device efficiency parameter correction model through a training strategy to obtain the trained device efficiency parameter correction model;

[0009] Inputting the input matrix of the set time period into the trained device efficiency parameter correction model for correction processing, and outputting the device efficiency parameter prediction matrix of the set time period;

[0010] Constructing the G-DEH model of ICES based on the trained device efficiency parameter correction model and the DEH model;

[0011] Constructing an ICES optimized scheduling model based on the G-DEH model;

[0012] Updating the efficiency parameters of the previous time period of the G-DEH model according to the device efficiency parameter prediction matrix of the set time period to obtain the G-DEH model and the ICES optimized scheduling model of the set time period;

[0013] Solving the ICES optimized scheduling model of the set time period to obtain the scheduling strategy of the set time period;

[0014] According to the scheduling strategy of the set time period, ICES performs scheduling execution; after the scheduling execution is completed, ICES updates the set parameters and optimizes the scheduling strategy for the next time period.

[0015] As a preferred solution of the ICES model predictive control method based on a GRU-driven dynamic energy hub, the expression of the DEH model is:

[0016]

[0017] In the formula, L e,t and L h,tThe electrical and heating load demands of the ICES in period t are respectively; P e,t and P g,t The electricity purchase and gas purchase powers of the ICES in period t are respectively; is the correction value of the heating efficiency of the HP in period t; is the correction value of the heating efficiency of the GB in period t; is the correction value of the power generation efficiency of the CHP in period t; is the correction value of the heat - electricity ratio of the CHP in period t; ν ee,t is the distribution factor of the electricity purchase power, representing the proportion of the electricity purchase power directly supplied to the electrical load in period t; (1 - ν ee,t ) is the proportion of the heat generated by the conversion of the electricity purchase power by the HP in period t; ν gh,t is the distribution factor of the gas purchase power in period t, representing the proportion of natural gas supplied to the CHP; (1 - ν gh,t ) is the proportion of the heat generated by the conversion of the gas purchase power by the GB in period t; P BAT,t is the charge - discharge power of the BAT in period t, taking a positive value for charging and a negative value for discharging; P TST,t is the charge - heat and discharge - heat power of the TST in period t, taking a positive value for heat charging and a negative value for heat discharging.

[0018] As an optimal solution of the ICES model predictive control method based on GRU - driven dynamic energy hub, the device efficiency parameter correction model includes four - layer networks: an input layer, a hidden layer, a fully - connected layer, and an output layer; the input matrix enters the hidden layer through the input layer, is processed by the hidden layer, and outputs a hidden state matrix; the hidden state matrix is transformed and processed by the fully - connected layer to obtain a device efficiency parameter prediction matrix; the device efficiency parameter prediction matrix is output through the output layer;

[0019] The expression of the input matrix is:

[0020]

[0021] where, N GRU is the length of the prediction time domain of the device efficiency parameter correction model; k is the k - th GRU in the hidden layer (1 ≤ k ≤ N GRU ); is the input matrix of the k - th GRU;

[0022] The expression of the hidden state matrix is:

[0023]

[0024] where, is the hidden state value at time k + t; N GRU is the number of GRUs;

[0025] The expression of the device efficiency parameter prediction matrix is as follows:

[0026]

[0027] In the formula, is the prediction vector composed of device efficiency parameters at time t 0 + k.

[0028] As an optimal solution of the ICES model predictive control method based on GRU-driven dynamic energy hub, the ICES optimal scheduling model aims at the optimal operation economy of ICES within the scheduling time domain; the expression of the ICES optimal scheduling model is as follows:

[0029]

[0030] In the formula, N c is the scheduling time domain; Δt is the scheduling time step; is the electricity purchase cost of ICES at time t; P e (t) is the electricity purchase power of ICES at time t; c e (t) is the electricity price at time t; is the gas purchase cost of ICES at time t; P g (t) is the gas purchase power of ICES at time t; c g (t) is the gas price at time t.

[0031] As an optimal solution of the ICES model predictive control method based on GRU-driven dynamic energy hub, the constraint conditions of the ICES optimal scheduling model include: operation constraints of energy conversion equipment, operation constraints of energy storage equipment, energy balance constraints, and interaction constraints with the superior network;

[0032] The operation constraints of the energy conversion equipment are as follows:

[0033] 0 ≤ P out,i,t ≤ P cap,i

[0034] In the formula, P out,i,t is the output power of the energy conversion equipment i in ICES at time t; P cap,i represents the rated output power of the energy conversion equipment i;

[0035] The operation constraints of the energy storage equipment are as follows:

[0036] W j,t+1 = W j,t (1 - σ j ) + (P j,C,t η j,C - P j,D,t / η j,D )Δt

[0037] In the formula, W j,t and W j,t+1 are the stored energy of the energy storage device j (j = 1, 2 corresponding to BAT and TST respectively) at time t and t + 1; σ j is the self-loss coefficient of device j; P j,C,t and P j,D,t are the charging and discharging powers of device j respectively; η j,C and η j,D are the charging and discharging efficiencies of device j respectively; Δt is the unit scheduling step size;

[0038] 0 ≤ P j,C,t U j,C,t ≤ P j,C,max

[0039] 0 ≤ P j,D,t U j,D,t ≤ P j,D,max

[0040] U j,C,t + U j,D,t ≤ 1

[0041] In the formula, P j,C,max and P j,D,max represent the maximum charging and discharging powers of the energy storage device j respectively; U j,C,t and U j,D,t are binary 0-1 variables representing the charging and discharging states of the energy storage device j, restricting the device from charging and discharging simultaneously, U j,C,t = 1 represents that the energy storage device j is in the charging state; U j,D,t = 1 represents that the device j is in the discharging state;

[0042] W j,min ≤ W j,t ≤ W j,max

[0043] In the formula, W j,min and W j,max are the upper and lower limit values of the allowable stored energy of the energy storage device j respectively;

[0044] W j,start = W j,end

[0045] In the formula, W j,start and W j,end are the stored energies of the energy storage device j at the initial and termination moments of the scheduling period respectively;

[0046] The energy balance constraint is:

[0047]

[0048] The upper-level network interaction constraint is as follows:

[0049] P e (t) ≤ P e,cap

[0050] P g (t) ≤ P g,cap

[0051] In the formula, P e,cap and P g,cap are respectively the upper limits of the interactive power between ICES and the power network and the natural gas network.

[0052] As an optimal solution of the ICES model predictive control method based on GRU-driven dynamic energy hubs, by calculating the absolute difference between the supply value and the true value of L e and L h within the scheduling day, the accuracy of the optimal scheduling strategy is evaluated; the calculation formula is:

[0053]

[0054] In the formula, is the absolute difference between the supply value and the true value of L e in the optimal scheduling strategy within the scheduling day; and are respectively the supply error components of L e caused by the prediction error of the equipment efficiency parameter at time t and the supply error component of L e caused by the prediction error of e ; both being positive indicates a shortage in the supply of L e under the optimal scheduling strategy, and vice versa, indicating an excess in the supply of L e ; and are respectively the predicted values of the CHP power generation efficiency, CHP thermoelectric ratio, GB heating efficiency, and HP heating efficiency in the optimal scheduling strategy at time t; and are respectively the true values of the CHP power generation efficiency, CHP thermoelectric ratio, GB heating efficiency, and HP heating efficiency at time t; is the true value of L e at time t; is the predicted value of L e at time t; is the absolute difference between the supply value and the true value of L h in the optimal scheduling strategy within the scheduling day; and are respectively the supply error components of L h caused by the prediction error of the equipment efficiency parameter at time t and the supply error component of L hL caused by prediction error h supply error component. If both are positive, it indicates that under the optimized scheduling strategy, there is a shortage in the L h supply. Conversely, it indicates that there is h an oversupply in the L The true value of L at time t h ; The predicted value of L at time t h .

[0055] The present invention also provides an ICES model predictive control device based on a GRU-driven dynamic energy hub. Based on the above ICES model predictive control method based on a GRU-driven dynamic energy hub, it includes:

[0056] A DEH model construction module for modeling the coupling relationship between the ICES input and output based on the coupling matrix of a multivariable nonlinear system to obtain a DEH model;

[0057] A device efficiency parameter correction model construction and training module for constructing a device efficiency parameter correction model based on a GRU network; training the device efficiency parameter correction model through a training strategy to obtain the trained device efficiency parameter correction model;

[0058] A device efficiency parameter correction model processing module for inputting the input matrix of a set time period into the trained device efficiency parameter correction model for correction processing and outputting the device efficiency parameter prediction matrix of the set time period;

[0059] A G-DEH model construction module for constructing a G-DEH model of ICES based on the trained device efficiency parameter correction model and the DEH model;

[0060] An ICES optimized scheduling model construction module for constructing an ICES optimized scheduling model based on the G-DEH model;

[0061] A G-DEH model parameter update module for updating the efficiency parameters of the G-DEH model in the previous time period according to the device efficiency parameter prediction matrix of the set time period to obtain the G-DEH model and the ICES optimized scheduling model of the set time period;

[0062] An ICES optimized scheduling model solving module for solving the ICES optimized scheduling model of the set time period to obtain the scheduling strategy of the set time period;

[0063] A scheduling strategy execution module for performing scheduling execution on ICES according to the scheduling strategy of the set time period; after the scheduling execution is completed, ICES updates the set parameters to optimize the scheduling strategy for the next time period.

[0064] As an optimal solution of the ICES model predictive control device based on the GRU-driven dynamic energy hub, in the DEH model construction module, the expression of the DEH model is:

[0065]

[0066] In the formula, L e,t and L h,t are the electricity and heat load demands of the ICES at time t, respectively; P e,t and P g,t are the electricity purchase and gas purchase powers of the ICES at time t, respectively; is the correction value of the heating efficiency of the HP at time t; is the correction value of the heating efficiency of the GB at time t; is the correction value of the power generation efficiency of the CHP at time t; is the correction value of the thermoelectric ratio of the CHP at time t; ν ee,t is the distribution factor of the electricity purchase power, representing the proportion of the electricity purchase power directly supplied to the electricity load at time t; (1 - ν ee,t ) is the proportion of the electricity purchase power converted into heat by the HP at time t; ν gh,t is the distribution factor of the gas purchase power at time t, representing the proportion of natural gas supplied to the CHP; (1 - ν gh,t ) is the proportion of the gas purchase power converted into heat by the GB at time t; P BAT,t is the charge and discharge power of the BAT at time t, taking a positive value for charging and a negative value for discharging; P TST,t is the charge and heat release power of the TST at time t, taking a positive value for heat charging and a negative value for heat release.

[0067] As an optimal solution of the ICES model predictive control device based on the GRU-driven dynamic energy hub, in the device efficiency parameter correction model construction and training module, the device efficiency parameter correction model includes four layers of networks: an input layer, a hidden layer, a fully connected layer, and an output layer; the input matrix enters the hidden layer through the input layer, is processed by the hidden layer, and outputs a hidden state matrix; the hidden state matrix is transformed and processed by the fully connected layer to obtain a device efficiency parameter prediction matrix; the device efficiency parameter prediction matrix is output through the output layer;

[0068] The expression of the input matrix is:

[0069]

[0070] In the formula, N GRU is the length of the prediction time domain of the device efficiency parameter correction model; k is the k-th GRU of the hidden layer (1 ≤ k ≤ N GRU ); is the input matrix for the k-th GRU;

[0071] The expression of the hidden state matrix is:

[0072]

[0073] In the formula, is the hidden state value at time k + t; N GRU is the number of GRUs;

[0074] The expression of the device efficiency parameter prediction matrix is:

[0075]

[0076] In the formula, is t 0 + k moment prediction vector composed of device efficiency parameters.

[0077] As an optimal solution of the ICES model predictive control device based on GRU-driven dynamic energy hub, in the ICES optimal scheduling model construction module, the ICES optimal scheduling model aims at the optimal operation economy of ICES within the scheduling time domain; the expression of the ICES optimal scheduling model is:

[0078]

[0079] In the formula, N c is the scheduling time domain; Δt is the scheduling time step; is the electricity purchase cost of ICES in the t period; P e (t) is the electricity purchase power of ICES in the t period; c e (t) is the electricity price in the t period; is the gas purchase cost of ICES in the t period; P g (t) is the gas purchase power of ICES in the t period; c g (t) is the gas price in the t period.

[0080] As an optimal solution of the ICES model predictive control device based on GRU-driven dynamic energy hub, in the ICES optimal scheduling model construction module, the constraint conditions of the ICES optimal scheduling model include: energy conversion equipment operation constraint, energy storage equipment operation constraint, energy balance constraint and superior network interaction constraint;

[0081] The energy conversion equipment operation constraint is:

[0082] 0 ≤ P out,i,t ≤ P cap,i

[0083] In the formula, P out,i,tis the output power of the energy conversion device i in the ICES during the t period; P cap,i represents the rated output power of the energy conversion device i;

[0084] The operating constraints of the energy storage device are:

[0085] W j,t+1 = W j,t (1 - σ j ) + (P j,C,t η j,C - P j,D,t / η j,D )Δt

[0086] In the formula, W j,t and W j,t+1 are the stored energies of the energy storage device j (j = 1, 2 corresponding to BAT and TST respectively) at time t and t + 1; σ j is the self-loss coefficient of device j; P j,C,t and P j,D,t are the charging and discharging powers of device j respectively; η j,C and η j,D are the charging and discharging efficiencies of device j respectively; Δt is the unit scheduling step;

[0087] 0 ≤ P j,C,t U j,C,t ≤ P j,C,max

[0088] 0 ≤ P j,D,t U j,D,t ≤ P j,D,max

[0089] U j,C,t + U j,D,t ≤ 1

[0090] In the formula, P j,C,max and P j,D,max represent the maximum charging and discharging powers of the energy storage device j respectively; U j,C,t and U j,D,t are binary 0-1 variables indicating the charging and discharging states of the energy storage device j, restricting the device from charging and discharging simultaneously, U j,C,t = 1 represents that the energy storage device j is in the charging state; U j,D,t = 1 represents that device j is in the discharging state;

[0091] W j,min ≤ W j,t ≤ W j,max

[0092] In the formula, W j,min and W j,maxThey are the upper and lower limit values of the allowable energy storage capacity of the energy storage device j, respectively;

[0093] W j,start = W j,end

[0094] In the formula, W j,start and W j,end are the energy storage levels of the energy storage device j at the initial and termination moments of the scheduling period, respectively;

[0095] The energy balance constraint is:

[0096]

[0097] The upper-level network interaction constraint is:

[0098] P e (t) ≤ P e,cap

[0099] P g (t) ≤ P g,cap

[0100] In the formula, P e,cap and P g,cap are the upper limits of the interaction powers between the ICES and the power network and the natural gas network, respectively.

[0101] As an optimal solution of the ICES model predictive control device based on the GRU-driven dynamic energy hub, in the ICES optimal scheduling model solving module, by calculating the absolute difference between the supply value and the true value of L e and L h within the scheduling day, the accuracy of the optimal scheduling strategy is evaluated; the calculation formula is:

[0102]

[0103]

[0104] In the formula, is the absolute difference between the supply value and the true value of L e in the optimal scheduling strategy within the scheduling day; and are the supply error components of L e caused by the prediction error of the device efficiency parameter and the supply error component of L e caused by the prediction error at time t, respectively. A positive value for both indicates a shortage in the supply of L e under the optimal scheduling strategy, and vice versa, indicating an excess supply of L e ; e Supply surplus; and They are the predicted values of the CHP power generation efficiency, CHP heat-electricity ratio, GB heating efficiency, and HP heating efficiency in the optimized scheduling strategy for period t, respectively; and They are the true values of the CHP power generation efficiency, CHP heat-electricity ratio, GB heating efficiency, and HP heating efficiency in period t, respectively; is the true value of L in period t e true value; is the predicted value of L in period t e predicted value; is the absolute difference between the supply value and the true value of L h in the optimized scheduling strategy within the scheduling day; and They are the supply error components of L h caused by the prediction error of the equipment efficiency parameter and the supply error components of L h caused by the prediction error in period t, respectively. A positive value for both indicates a shortage in the supply of L h under the optimized scheduling strategy. Conversely, it indicates an excess supply of L h ; h supply surplus; is the true value of L in period t h true value; is the predicted value of L in period t h predicted value.

[0105] The present invention has the following advantages: Based on the coupling matrix of a multivariable nonlinear system, the present invention models the coupling relationship between the input and output of ICES to obtain a DEH model; based on the GRU network, a device efficiency parameter correction model is constructed; the device efficiency parameter correction model is trained through a training strategy to obtain the trained device efficiency parameter correction model; the input matrix of a set time period is input into the trained device efficiency parameter correction model for correction processing, and the device efficiency parameter prediction matrix of the set time period is output; based on the trained device efficiency parameter correction model and the DEH model, a G-DEH model of ICES is constructed; based on the G-DEH model, an ICES optimal scheduling model is constructed; according to the device efficiency parameter prediction matrix of the set time period, the efficiency parameters of the previous time period of the G-DEH model are updated to obtain the G-DEH model and the ICES optimal scheduling model of the set time period; the ICES optimal scheduling model of the set time period is solved to obtain the scheduling strategy of the set time period; according to the scheduling strategy of the set time period, ICES performs scheduling execution; after the scheduling execution is completed, ICES updates the set parameters to optimize the scheduling strategy for the next time period. The present invention constructs a G-DEH model of ICES and proposes an MPC strategy with the goal of optimal economy based on the G-DEH model. The G-DEH model can consider the correction effect of the operating conditions of the previous time section on the device efficiency of the current time section, more accurately describe the time series relationship between the device load rate, temperature, and pressure sequence and the device efficiency parameters, and further improve the matching degree between the ICES optimal scheduling model and the actual operating conditions. The MPC strategy based on G-DEH can realize the rolling iterative update of the device load rate and device efficiency parameters of the energy conversion device within the scheduling period. On the premise of maintaining the supply-demand balance of ICES, this strategy further improves the accuracy and economy of the ICES optimal scheduling scheme. BRIEF DESCRIPTION OF THE DRAWINGS

[0106] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only exemplary, and for those of ordinary skill in the art, without creative efforts, other implementation drawings can be obtained according to the provided drawings.

[0107] The structures, ratios, sizes, etc. illustrated in this specification are only used to match the content disclosed in the specification for those familiar with this technology to understand and read, and are not used to limit the implementation conditions of the present invention. Therefore, they do not have substantial technical significance. Any modification of the structure, change in the ratio relationship, or adjustment of the size, without affecting the effects that the present invention can produce and the purposes that can be achieved, should still fall within the scope covered by the technical content disclosed in the present invention.

[0108] Figure 1 It is a schematic flow chart of the ICES model predictive control method based on GRU-driven dynamic energy hub provided in Embodiment 1 of the present invention;

[0109] Figure 2 It is a schematic specific implementation flow chart of the ICES model predictive control method based on GRU-driven dynamic energy hub provided in Embodiment 1 of the present invention;

[0110] Figure 3 It is a schematic diagram of the G-DEH model of ICES in the ICES model predictive control method based on GRU-driven dynamic energy hub provided in Embodiment 1 of the present invention;

[0111] Figure 4 It is a schematic diagram of the device efficiency parameter correction model in the ICES model predictive control method based on GRU-driven dynamic energy hub provided in Embodiment 1 of the present invention;

[0112] Figure 5 It is a schematic diagram of the loss function curves of the device efficiency parameter correction model and the BPNN model in the ICES model predictive control method based on GRU-driven dynamic energy hub provided in Embodiment 1 of the present invention;

[0113] Figure 6 It is the L of the typical day ICES in a possible embodiment provided in Embodiment 1 of the present invention e 、L h 、T and F curve schematic diagrams;

[0114] Figure 7 It is a schematic diagram of the device efficiency parameter prediction curves in Scenario II and Scenario III (E = 200) in a possible embodiment provided in Embodiment 1 of the present invention;

[0115] Figure 8 It is a schematic diagram of the electrical and thermal power balance and error component results in Scenario I in a possible embodiment provided in Embodiment 1 of the present invention;

[0116] Figure 9 It is a schematic diagram of the electrical and thermal power balance and error component results in Scenario II in a possible embodiment provided in Embodiment 1 of the present invention;

[0117] Figure 10 Schematic diagram of the electrical and thermal power balance and error component results in Scenario III (E = 200) in a possible embodiment provided in Embodiment 1 of the present invention;

[0118] Figure 11 Schematic diagram of the architecture of the ICES model predictive control device based on the GRU-driven dynamic energy hub provided in Embodiment 2 of the present invention. Detailed implementation manners

[0119] The following specific embodiments illustrate the implementation manners of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0120] Embodiment 1

[0121] Refer to Figure 1 and Figure 2 Embodiment 1 of the present invention provides an ICES model predictive control method based on a GRU-driven dynamic energy hub, including the following steps:

[0122] S1. Based on the coupling matrix of the multivariable nonlinear system, model the coupling relationship between the input and output of ICES to obtain the DEH model;

[0123] S2. Based on the GRU network, construct a device efficiency parameter correction model; train the device efficiency parameter correction model through a training strategy to obtain the trained device efficiency parameter correction model;

[0124] S3. Input the input matrix of the set time period into the trained device efficiency parameter correction model for correction processing, and output the device efficiency parameter prediction matrix of the set time period;

[0125] S4. Based on the trained device efficiency parameter correction model and the DEH model, construct the G-DEH model of ICES;

[0126] S5. Based on the G-DEH model, construct an ICES optimal scheduling model;

[0127] S6. According to the device efficiency parameter prediction matrix of the set time period, update the efficiency parameters of the G-DEH model in the previous time period to obtain the G-DEH model and the ICES optimal scheduling model of the set time period;

[0128] S7. Solve the ICES optimal scheduling model for the set time period to obtain the scheduling strategy for the set time period;

[0129] S8. According to the scheduling strategy for the set time period, ICES performs scheduling execution; after the scheduling execution is completed, ICES updates the set parameters and optimizes the scheduling strategy for the next time period.

[0130] In this embodiment, the framework of the ICES model predictive control method driven by G-DEH (GRU-driven dynamic energy hub) is as Figure 2 shown. To simplify the complex continuous-time process, the scheduling day is divided into N discrete time periods, and each time period is a scheduling time step (Δt). N p is the prediction horizon (consistent with the value of N GRU ), N c is the scheduling horizon, and it satisfies: N ≥ N p ≥ N c . First, train the device efficiency parameter correction model. After the training is completed, construct the G-DEH model. Within N p , use the device efficiency parameter correction model to predict the device efficiency parameter prediction matrix within N p based on the actual data of the operating conditions to correct the efficiency parameters of the G-DEH model; within N c , solve the optimal scheduling model of ICES with the lowest operating cost as the goal and update for outputting the next device efficiency parameter prediction matrix. This model predictive control method is a sequential solution process, so the device efficiency parameter correction model and the optimal scheduling model within the scheduling day will both be called N times.

[0131] In this embodiment, in step S1, based on the coupling matrix of the multivariable nonlinear system, model the coupling relationship between the input and output of ICES to obtain the DEH model;

[0132] Specifically, based on the coupling matrix of the multivariable nonlinear system, model the coupling relationship between the input and output of ICES to obtain the DEH model. DEH (driven dynamic energy hub) is a dynamic energy hub; through the DEH model, the true operating state of ICES under the dynamic change of device efficiency parameters can be accurately reflected. The expression of the DEH model is:

[0133]

[0134] In the formula, L e,t and L h,tThe electrical and heat load demands of the ICES in period t are respectively; P e,t and P g,t The electricity purchase and gas purchase powers of the ICES in period t are respectively; is the correction value of the heating efficiency of the HP in period t; is the correction value of the heating efficiency of the GB in period t; is the correction value of the power generation efficiency of the CHP in period t; is the correction value of the thermoelectric ratio of the CHP in period t; ν ee,t is the distribution factor of the electricity purchase power, representing the proportion of the electricity purchase power directly supplied to the electrical load in period t; (1 - ν ee,t ) is the proportion of the electricity purchase power converted into heat by the HP in period t; ν gh,t is the distribution factor of the gas purchase power in period t, representing the proportion of natural gas supplied to the CHP; (1 - ν gh,t ) is the proportion of the gas purchase power converted into heat by the GB in period t; P BAT,t is the charge and discharge power of the BAT in period t, taking a positive value for charging and a negative value for discharging; P TST,t is the charge and heat release power of the TST in period t, taking a positive value for heat charging and a negative value for heat release.

[0135] In this embodiment, in step S2, based on the GRU network, a device efficiency parameter correction model is constructed; the device efficiency parameter correction model is trained through a training strategy to obtain the trained device efficiency parameter correction model;

[0136] Specifically, as Figure 4 shown, based on the GRU (Gated Recurrent Unit) network, a device efficiency parameter correction model is constructed; the device efficiency parameter correction model includes four layers of networks: an input layer, a hidden layer, a fully connected layer, and an output layer; the input matrix enters the hidden layer through the input layer, and through the processing of the hidden layer, a hidden state matrix is output; the hidden state matrix is transformed through the fully connected processing layer to obtain a device efficiency parameter prediction matrix; the device efficiency parameter prediction matrix is output through the output layer;

[0137] At the initial moment t 0 in period t, the expression of the input matrix X t0 of the input layer is:

[0138]

[0139] In the formula, N GRU is the length of the prediction time domain of the device efficiency parameter correction model; k is the k-th GRU in the hidden layer (1 ≤ k ≤ N GRU ); is the input matrix of the k-th GRU;

[0140] Load factor The temperature T and the air pressure F are the three operating condition parameters that have the greatest impact on the equipment efficiency parameters. Therefore, is t 0 -N GRU +k moment by Figure 3 The three operating condition parameters of the three energy conversion devices HP, GB, and CHP in form a matrix with a length of 3 and a width of 3, and the expression is:

[0141]

[0142] In the formula, is t 0 -N GRU +k moment of the load factor of HP, and the meanings of the other parameters can be understood by analogy.

[0143] Such as Figure 4 shown, the hidden layer of the efficiency parameter correction model consists of N GRU GRUs connected in sequence according to the time series, where the kth GRU is responsible for predicting the prediction vector 0 +k moment composed of four equipment efficiency parameters As shown in formula (4):

[0144]

[0145] The operation process inside the kth GRU is as follows:

[0146] GRU first calculates the candidate hidden state value As shown in formula (5):

[0147]

[0148] Not only depends on the current input It is also affected by the hidden state of the previous time section . The value of the reset gate is between 0 and 1, and the closer its value is to 1, it means that the equipment efficiency correction model will retain more operation condition data information of the equipment under the previous time section, avoiding prediction deviation of equipment efficiency parameters caused by excessive forgetting of operation condition historical data. After calculating , GRU calculates the hidden state value of the current k+t moment through the values of the update gate and As shown in formula (6):

[0149]

[0150] Determines​ and The weight ratio of the equipment efficiency parameter prediction in the current time section is Ability to balance the current time period with the previous N GRU -1 Time period surface equipment operating condition information, so as to more accurately capture the dynamic characteristics of equipment efficiency parameters. and The calculation formulas are shown in formula (7) and formula (8):

[0151]

[0152] In the above calculation process inside the kth GRU, in the calculation formula, is the Hadamard product of the matrix; “×” is the matrix multiplication operation; “+” is the matrix addition operation; σ(·) and tanh(·) are the sigmoid function and the hyperbolic tangent function, respectively, as the activation function of GRU; W r , W u , W c , U r , U u and U c is the weight matrix; b r 、b u and b c is the bias vector.

[0153] The expression of the hidden state matrix output by the hidden layer is:

[0154]

[0155] In the formula, is the hidden state value at time k+t; N GRU is the number of GRU;

[0156] A fully connected layer is set after the hidden layer, and the hidden state matrix of the hidden layer output is Transform the learned multi-dimensional information features into a low-dimensional sample space to obtain the equipment efficiency parameter prediction matrix Equipment efficiency parameter prediction matrix For correction Figure 4 Efficiency parameters of various equipment of G-DEH in NGRU.

[0157] The expression of the equipment efficiency parameter prediction matrix is:

[0158]

[0159] In the formula, t 0 +The prediction vector consisting of equipment efficiency parameters at time k.

[0160] In this embodiment, the device efficiency parameter correction model is trained through a training strategy to obtain the trained device efficiency parameter correction model;

[0161] Specifically, the training process of the device efficiency parameter correction model is as follows:

[0162] Preprocess the data: The input data set is {X τ , Y τ} S , with a total of S sample data, and τ is the τ-th sample data. X τ is a feature matrix with a length of 3×N and a width of 3, which is composed of the historical data of T and F of HP, GB, and CHP, as shown in Equation (2); Y GRU is a device efficiency label value matrix with a length of N τ and a width of 4, as shown in Equation (10). First, {X GRU , Y τ} τ is cleaned to fill in missing values and remove outliers. Then, the'min-max' method is used to standardize the data to accelerate the training process. S Train the device efficiency parameter correction model: The training time step t

[0163] , learning rate α, maximum number of iterations n, number of neurons in the hidden layer N s , and number of neurons in the fully connected layer N h are set according to existing literature. Before training, the weight matrices and bias vectors in Equations (5)-(8) are initialized using a random initialization function. During training, B samples {X f , Y τ} τ are input in each iteration process, and the device efficiency parameter correction model performs forward propagation on the input B samples to finally obtain a sequence of device efficiency parameter prediction matrices B . The predicted values are compared with the true values {Y } τ and the loss function L is calculated for HP, GB, and CHP respectively, as shown in Equation (11): B In the formula, || ||

[0164]

[0165] is the Frobenius norm; F ; and and They are the predicted values of the equipment efficiency parameters required for HP, GB, and CHP in the τ-th sample; Y HP,τ , Y GB,τ , and Y CHP,τ are the actual values of the equipment efficiency parameters required for HP, GB, and CHP in the τ-th sample respectively.

[0166] The Adaptive Moment Estimation Algorithm (Adam) is used to adjust the weight matrix and bias vector of the equipment efficiency parameter correction model, ensuring a relatively fast convergence rate. After multiple iterations, the value of L continuously decreases until the change in the value of L between two consecutive iterations is very small and the change trend tends to be stable, then it can be considered that the training result has converged.

[0167] In this embodiment, Figure 3 the operating condition data collected by the ICES field sensors shown is used as the data set {X τ , Y τ} S to compare and verify the advantages of the equipment efficiency parameter correction model and the BPNN model in this paper in processing time series data. The hyperparameter settings of the equipment efficiency parameter correction model are shown in Table 1:

[0168] <![CDATA[t s > α n <![CDATA[N h > <![CDATA[N f > 8 0.02 400 32 32

[0169] Table 1 Hyperparameter settings of the equipment efficiency parameter correction model

[0170] As Figure 5 shown, they are the loss function curves of the equipment efficiency parameter correction model and the BPNN model. By comparing Figure 5 (a) and Figure 5 (b), it can be obtained that in terms of the convergence speed, the L of the equipment efficiency correction model drops rapidly in the initial stage of training and approaches 0 when the number of iterations E = 150, indicating that it can quickly capture and learn the time series laws and characteristics of the equipment historical data; while the BPNN model needs to be iterated until E is about 500 to reach convergence, and the convergence speed is significantly slower. In terms of the convergence accuracy, when E = 150, the L values of the equipment efficiency correction models for HP, GB, and CHP are 0.0163, 0.0231, and 0.0117 respectively, and the L values of the BPNN model are 0.0686, 0.0416, and 0.0580 respectively, which are significantly higher than those of the equipment efficiency correction model. This shows that the equipment efficiency correction model has higher accuracy in predicting the equipment efficiency parameters considering the time series correlation of the operating conditions of energy conversion equipment compared with the BPNN.

[0171] In this embodiment, in step S3, the input matrix of the set time period is input into the trained equipment efficiency parameter correction model for correction processing, and the equipment efficiency parameter prediction matrix of the set time period is output;

[0172] In this embodiment, in step S4, based on the trained device efficiency parameter correction model and the DEH model, a G-DEH model of ICES is constructed.

[0173] Specifically, as Figure 3 shown, the G-DEH model of ICES includes three parts: the ICES architecture, the DEH model, and the device efficiency correction model based on the GRU network. A typical electrical-thermal-gas coupled ICES includes five energy devices: a combined heat and power (CHP) unit, a gas boiler (GB), a heat pump (HP), a battery (BAT), and a thermal storage tank (TST). The G-DEH model consists of the DEH model and the device efficiency correction model to accurately depict the impact of the dynamic change of device efficiency parameters on the actual operating state of ICES. The DEH model is based on the coupling matrix of a multivariable nonlinear system to model the complex coupling relationship between the input and output of ICES, aiming to accurately reflect the true operating state of ICES under the dynamic change of device efficiency parameters. The device efficiency correction model is composed of GRUs, and the arrow direction connecting each GRU is the positive time direction. The device efficiency correction model can fully capture the temporal correlation of device efficiency parameters with the change of operating conditions, predict and dynamically correct the efficiency parameters of each device in Equation (1) at the initial moment of each scheduling period, so as to accurately reflect the real-time operating state of ICES.

[0174] In this embodiment, in step S5, based on the G-DEH model, an ICES optimal scheduling model is constructed.

[0175] Among them, assuming that the variable values involved in the ICES optimal scheduling model only change at the start or end of Δt, the ICES optimal scheduling model can be written in discrete form.

[0176] Specifically, the ICES optimal scheduling model aims to maximize the operating economy of ICES within N c , and the expression of the ICES optimal scheduling model is shown in Equation (12):

[0177]

[0178] In the formula, N c is the scheduling time domain; Δt is the scheduling time step; is the power purchase cost of ICES at time t; P e (t) is the power purchase power of ICES at time t; c e (t) is the electricity price at time t; The gas purchase cost of ICES during period t; P g (t) is the gas purchase power of ICES during period t; c g (t) is the gas price during period t.

[0179] The constraint conditions of the ICES optimal scheduling model include: the operation constraints of energy conversion equipment, the operation constraints of energy storage equipment, the energy balance constraints, and the interaction constraints with the superior network;

[0180] The operation constraints of energy conversion equipment mean that the output power of energy conversion equipment i (i = 1, 2, 3 corresponding to HP, GB, and CHP respectively) is restricted by its installed capacity; the expression is as shown in Equation (15):

[0181] 0 ≤ P out,i,t ≤ P cap,i (15)

[0182] In the formula, P out,i,t is the output power of energy conversion equipment i in ICES during period t; P cap,i represents the rated output power of energy conversion equipment i;

[0183] The operation constraints of energy storage equipment:

[0184] The charging and discharging mechanisms of different energy storage equipment have certain similarities. The change in stored energy is related to parameters such as the initial stored energy, charging and discharging power, charging and discharging efficiency, and self-loss rate. The expression is as shown in Equation (16):

[0185] W j,t+1 = W j,t (1 - σ j ) + (P j,C,t η j,C - P j,D,t / η j,D )Δt (16)

[0186] In the formula, W j,t and W j,t+1 are the stored energies of energy storage equipment j (j = 1, 2 corresponding to BAT and TST respectively) at times t and t + 1; σ j is the self-loss coefficient of equipment j; P j,C,t and P j,D,t are the charging and discharging powers of equipment j respectively; η j,C and η j,D are the charging and discharging efficiencies of equipment j respectively; Δt is the unit scheduling step length;

[0187] The charging and discharging powers of energy storage equipment need to be restricted within a certain range, as shown in Equations (17)-(19):

[0188] 0 ≤ P j,C,tU j,C,t ≤P j,C,max (17)

[0189] 0≤P j,D,t U j,D,t ≤P j,D,max (18)

[0190] U j,C,t +U j,D,t ≤1(19)

[0191] In the formula, P j,C,max and P j,D,max respectively represent the maximum charging and discharging powers of the energy storage device j; U j,C,t and U j,D,t are binary 0-1 variables representing the charging and discharging states of the energy storage device j, restricting the device from charging and discharging simultaneously. U j,C,t =1 represents that the energy storage device j is in the charging state; U j,D,t =1 represents that the device j is in the discharging state;

[0192] The stored energy of the energy storage device j should satisfy the following constraints to prevent overcharging and discharging from harming the battery life:

[0193] W j,min ≤W j,t ≤W j,max (20)

[0194] In the formula, W j,min and W j,max are respectively the upper and lower limit values of the allowable stored energy of the energy storage device j;

[0195] In addition, the stored energy of the energy storage device j should be equal at the initial and termination moments of the scheduling period to maintain the charging and discharging balance of a complete scheduling period:

[0196] W j,start =W j,end (21)

[0197] In the formula, W j,start and W j,end are respectively the stored energies of the energy storage device j at the initial and termination moments of the scheduling period.

[0198] Energy balance constraint: The ICES optimal scheduling model should satisfy the energy balance constraint based on the G-DEH model, as shown in Equation (1).

[0199] Superior network interaction constraint:

[0200] The interaction powers between the ICES and the superior power network and natural gas network are restricted by Equations (22)-(23):

[0201] P e (t) ≤ P e,cap (22)

[0202] P g (t) ≤ P g,cap (23)

[0203] Wherein, P e,cap and P g,cap are respectively the upper limits of the interactive power between the ICES and the power grid and the natural gas grid.

[0204] In this embodiment, there must be a certain error between the device output in the optimized scheduling scheme and the actual values of L e and L h There are two factors causing this error: 1) There is an error in the predicted values of L e and L h ; 2) There is an error in the prediction of the device efficiency parameters. According to the training results of the device efficiency parameter correction model, the present invention assumes that there is no error in the device efficiency parameters predicted by the device efficiency parameter correction model when E = 200, and calculates the absolute difference between the supply value and the true value of L e and L h within the scheduling day, so as to reflect the accuracy of the optimized scheduling strategy, as shown in Equations (24) and (25):

[0205]

[0206] Wherein, is the absolute difference between the supply value and the true value of L e in the optimized scheduling strategy within the scheduling day; and are respectively the supply error components of L e caused by the prediction error of the device efficiency parameters at time t and the supply error component of L e caused by the prediction error of L e . If both are positive, it means that there is a shortage in the supply of L e under the optimized scheduling strategy. On the contrary, it means that the supply of L e is excessive; and are respectively the predicted values of the CHP power generation efficiency, CHP thermoelectric ratio, GB heating efficiency and HP heating efficiency in the optimized scheduling strategy at time t; and are respectively the true values of the CHP power generation efficiency, CHP thermoelectric ratio, GB heating efficiency and HP heating efficiency at time t; is the true value of L e at time t; is the predicted value of L e at time t; For the L within the scheduling day h The absolute difference between the supply value and the true value in the optimized scheduling strategy; and Are respectively the L supply error components caused by the prediction error of the equipment efficiency parameter in period t h And the L supply error component caused by the prediction error. If both are positive, it indicates that there is a shortage in the L supply under the optimized scheduling strategy. Conversely, it indicates that h The L supply error component caused by the prediction error. If both are positive, it indicates that there is a shortage in the L supply under the optimized scheduling strategy. Conversely, it indicates that h There is an L supply shortage under the optimized scheduling strategy. Conversely, it indicates that h There is an L supply shortage under the optimized scheduling strategy. Conversely, it indicates that h There is an oversupply; Is the true value of L in period t h ; Is the predicted value of L in period t h ;

[0207] In this embodiment, in step S6, according to the equipment efficiency parameter prediction matrix of the set time period, update the efficiency parameters of the G-DEH model in the previous time period to obtain the G-DEH model and the ICES optimized scheduling model of the set time period;

[0208] In this embodiment, in step S7, solve the ICES optimized scheduling model of the set time period to obtain the scheduling strategy of the set time period;

[0209] Specifically, call the Gurobi solver through Python to solve the ICES optimized scheduling model of the set time period to obtain the scheduling strategy of the set time period;

[0210] In this embodiment, in step S8, according to the scheduling strategy of the set time period, ICES performs scheduling execution; after the scheduling execution is completed, ICES updates the set parameters and optimizes the scheduling strategy for the next time period.

[0211] In this embodiment, an MPC strategy is provided to implement model predictive control of the G-DEH-driven ICES model. The specific steps are as follows:

[0212] T1. Train the equipment efficiency parameter correction model. After the training is completed, initialize the G-DEH model;

[0213] T2. t = 1, 2,..., N;

[0214] T3. Collect the L e,t and L h,t ; in the time period [t, t + 1];

[0215] T4. Take the p in the time period [t - N Input device efficiency parameter correction model, outputting [[t, t + N p -1].

[0216] T5. According to update the efficiency parameters of the G-DEH model for the [[t, t + N p -1] period;

[0217] T6. Solve the ICES optimal scheduling model to obtain the scheduling plan of ICES for the [[t, t + N c -1] period;

[0218] T7. ICES only executes the instructions in the first period of the scheduling plan, that is, the scheduling plan for the [[t + 1, t + 2]] period;

[0219] T8. Update the [[t + 1, t + 2]] period and

[0220] T9. Let t = t + 1, and determine whether t is equal to N; if t ≠ N, return to step T3; if t = N, jump out of the loop.

[0221] In a possible embodiment, a specific optimal scheduling example is provided as follows:

[0222] I. Parameter setting:

[0223] Select the Figure 3 shown typical ICES for analysis. The rated parameters of the devices in ICES are shown in Table 2, and the electricity price and natural gas purchase price are shown in Table 3.

[0224]

[0225] Table 2 Device rated parameters

[0226]

[0227]

[0228] Table 3 Energy purchase price

[0229] This invention focuses on the efficiency change of energy conversion devices, so the prediction methods of L e , L h , T, and F are not discussed. Without loss of generality, this invention assumes that the actual data of L e , L h , T, and F are simulated by adding random noise of normal distribution to the predicted data. Among them, L e and L hThe expected error is taken as 0, and the mean square error is taken as 2%; the expected errors of T and F are taken as 0, and the mean square error is taken as 5%. The L of the typical dispatching day e and L h , T and F curves are as Figure 6 shown. N is 48, Δt is 0.5 h, N p is 4 h, N c is 2 h.

[0230] To verify the effectiveness of the method proposed in the present invention, three scenarios are set up for comparative analysis. All three scenarios solve the ICES optimal dispatching model based on the MPC strategy.

[0231] Scenario I: The EH model with constant equipment efficiency parameters is adopted;

[0232] Scenario II: The equipment efficiency parameters of the EH model are corrected by the BPNN;

[0233] Scenario III: The G-DEH-driven ICES model predictive control method proposed in this paper. At the same time, the influence of E on the optimal dispatching result is analyzed.

[0234] The dispatching strategies of the three scenarios are applied to Figure 6 the L shown in e and L h , T and F predicted values, and without considering the power and heat load deficits in the actual operation of ICES, the predicted operation cost of ICES is obtained. However, in actual operation, due to the existence of error components in Equations (24) and (25), the output of each device in Scenarios I to III needs to be corrected to meet the actual power and heat load demands. The correction strategy adopted in the present invention is: when the energy supply is short, the power load deficit is made up by purchasing electricity, and the heat load deficit is made up by GB, and the corresponding compensation cost is calculated; when the energy supply is excessive, the excessive energy is stored by BAT and TST. The relative error of the optimal dispatching strategy for each scenario is the error of the predicted operation cost of each scenario relative to the actual operation cost.

[0235] II. Analysis of example results:

[0236] The predicted operation cost, actual operation cost, relative error, absolute difference in L e supply, and absolute difference in L h supply of ICES in Scenarios I to III are shown in Table 4:

[0237]

[0238]

[0239] Table 4 Comparison of optimization results for three scenarios

[0240] In Scenario I, the predicted operating cost of ICES is 53,457.71 yuan, and the L e absolute difference in supply is 1,838.82 kWh, and the L h absolute difference in supply is 4,541.96 kWh. The actual operating cost of ICES after correction is 58,650.48 yuan, and the relative error is 8.85%. The results show that ignoring the variable operating conditions and time-series related characteristics of energy conversion equipment will have an adverse impact on the accuracy of the optimal scheduling scheme.

[0241] In Scenario II, the BPNN method is used to correct the efficiency parameters of energy conversion equipment. The predicted operating cost of ICES is 56,956.79 yuan, and the L e absolute difference in supply is 1,346.06 kWh, and the L h absolute difference in supply is 2,453.59 kWh. The actual operating cost after correction is 58,248.26 yuan, which is 0.69% lower than that in Scenario I. The relative error of the scheduling scheme is 6.63% lower than that in Scenario I, indicating that the economy and accuracy of the optimal scheduling strategy have been improved to a certain extent.

[0242] To analyze the influence of E on the optimal scheduling results, four optimal scheduling strategies are generated in Scenario III when E = 50, 100, 150, and 200 respectively. Among them, the actual operating cost and relative error of Scenario III (E = 50) are higher than those of Scenario II, indicating that the selection of the number of training iterations of the equipment efficiency parameter correction model has a significant impact on the economy and accuracy of the optimal scheduling results. It can be seen from Table 3 that as E increases, the economy and accuracy of the optimal scheduling strategy are also continuously improved. Among them, the predicted operating cost of ICES in Scenario III (E = 200) is 56,583.01 yuan, and the L e absolute difference in supply is 1,217.27 kWh, and the L h absolute difference in supply is 1,988.39 kWh. The actual operating cost of ICES in Scenario III (E = 200) is 57,403.24 yuan, which is 2.13% lower than that in Scenario I and 1.45% lower than that in Scenario II. The relative error of Scenario III (E = 200) is only 1.43%, which is 7.42% lower than that in Scenario I and 0.82% lower than that in Scenario II. The above results show that the optimal scheduling strategy based on the G-DEH model proposed in the present invention can fully consider the influence of the operating conditions of the previous time section on the equipment efficiency parameters of the current time section, and its combined application with the MPC strategy can effectively enhance the economy and accuracy of the ICES optimal scheduling strategy.

[0243] To further verify the effectiveness of the method proposed in the present invention, the optimized scheduling strategies of Scenario I, Scenario II, and Scenario III (E = 200) are compared and analyzed, and the predicted curves of the efficiency parameters of CHP, GB, and HP in Scenario II and Scenario III (E = 200) are obtained as Figure 7 shown. The results of the electricity and heat power balance and error components under the three scenarios are as Figures 8 - 10 shown. Among them, the red shadow represents the part of the load supply deficit, and the blue shadow represents the part of the load supply surplus; the positive error component corresponds to the load supply deficit, and vice versa corresponds to the load supply surplus.

[0244] In Scenario I, the equipment efficiency parameters are regarded as rated values. Under this condition, the CHP operates in the heat-to-power mode to supply part of the electrical load. It has high energy utilization efficiency and good economy, and most of the heat load is satisfied by the CHP. However, Scenario I ignores that the actual operating efficiency of the CHP at low load levels is lower than the rated efficiency. As Figure 8 can be seen, this results in the equipment output being on the small side during most periods in the optimized scheduling strategy of Scenario I. There are a large number of positive supply error components for the electrical load and heat load, and the load deficit is large. It is necessary to purchase electricity and gas from the power grid and gas network additionally, which instead increases the actual operating cost of the ICES. Therefore, the scheduling scheme under constant equipment efficiency parameters may be somewhat unreasonable.

[0245] In Scenario III (E = 200), as Figure 10 can be seen, the electrical load demand is mainly supplied by the CHP and the superior power grid together, and the heat load demand is mainly satisfied by the output of the CHP and the GB. During the period from 00:00 to 03:00, the heat load level is low. At this time, the CHP will be at a low load level and have a low operating efficiency. Only relying on the GB and HP with higher efficiency as the main heating equipment can meet the heat load demand of the ICES, and the electrical load demand is completely supplied by the superior power grid. During the period from 03:00 to 06:00, the heat load level exceeds the sum of the rated capacities of the GB and HP. During this period, the GB and HP are still the main heating equipment, the CHP supplies a small amount of heat load, and at the same time supplies a small amount of electrical load through the heat-to-power mode. During the period from 06:00 to 22:00, the CHP serves as the main heating equipment to maintain the efficient operation of the ICES. The reason is that the electricity price is high during this period, and the heat-to-power mode of the CHP can supply part of the electrical load, and the comprehensive economy is good. Combining Figure 7 can be seen, compared with the period from 00:00 to 03:00, the CHP has a higher efficiency and better comprehensive economy during the period from 22:00 to 24:00. Therefore, different from the period from 00:00 to 03:00, the CHP is still the main heating equipment during this period.

[0246] As Figure 9As shown in Figure 2, the scheduling results of scenario II have a certain error compared with scenario III (E=200). For example, during the period of 02:00-04:00, Figure 7 It can be seen that the predicted value of CHP efficiency parameter in scenario II is much greater than that in scenario III (E=200), which results in the heat load demand of ICES being met by CHP instead of GB during this period. At the same time, due to the overestimation of CHP efficiency parameter, there is a positive supply error component in the electric load and heat load during this period caused by the prediction error of equipment efficiency parameter, such as Figure 9 As shown in the orange column in (b). The load supply error component caused by the equipment efficiency parameter prediction error and the load supply error component caused by the load prediction error follow the principle of superposition of errors in the same direction and cancellation of errors in the opposite direction, and the total load supply error in this period is obtained. Obviously, there is a certain shortage of both the electric load and the thermal load in the period of 02:00-04:00, and additional electricity and gas need to be purchased from the power grid and gas grid. This error causes the economy and accuracy of the optimization scheduling strategy of scenario II to be worse than that of scenario III (E=200).

[0247] The above results show that considering the time series correlation of equipment operating status can reduce the impact of prediction error on ICES optimization scheduling to a certain extent, and the G-DEH model with time series data processing capability can improve the economy and accuracy of ICES optimization scheduling scheme.

[0248] In summary, the present invention models the coupling relationship between the input and output of ICES based on the coupling matrix of a multivariable nonlinear system to obtain a DEH model; constructs an equipment efficiency parameter correction model based on a GRU network; trains the equipment efficiency parameter correction model through a training strategy to obtain the trained equipment efficiency parameter correction model; inputs the input matrix of a set time period into the trained equipment efficiency parameter correction model for correction processing to output the equipment efficiency parameter prediction matrix of the set time period; constructs a G-DEH model of ICES based on the trained equipment efficiency parameter correction model and the DEH model; constructs an ICES optimal scheduling model based on the G-DEH model; updates the efficiency parameters of the previous time period of the G-DEH model according to the equipment efficiency parameter prediction matrix of the set time period to obtain the G-DEH model and the ICES optimal scheduling model of the set time period; solves the ICES optimal scheduling model of the set time period to obtain the scheduling strategy of the set time period; ICES performs scheduling execution according to the scheduling strategy of the set time period; after the scheduling execution is completed, ICES updates the set parameters to optimize the scheduling strategy for the next time period. The present invention constructs a G-DEH model of ICES and proposes an MPC strategy with the goal of optimal economy based on the G-DEH model. The G-DEH model can consider the correction effect of the operating conditions of the previous time section on the equipment efficiency of the current time section, more accurately describe the time series relationship between the equipment load rate, temperature and pressure sequence and the equipment efficiency parameters, and further improve the matching degree between the ICES optimal scheduling model and the actual operating conditions. The MPC strategy based on G-DEH can realize the rolling iterative update of the equipment load rate and equipment efficiency parameters of the energy conversion equipment within the scheduling period. This strategy further improves the accuracy and economy of the ICES optimal scheduling scheme on the premise of maintaining the supply-demand balance of ICES.

[0249] It should be noted that the method of the embodiments of the present disclosure can be executed by a single device, such as a computer or a server. The method of this embodiment can also be applied to a distributed scenario and completed by multiple devices cooperating with each other. In this case of a distributed scenario, one of the multiple devices can only execute one or more steps of the method of the embodiments of the present disclosure, and these multiple devices will interact with each other to complete the described method.

[0250] It should be noted that some embodiments of the present disclosure have been described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the above embodiments and still achieve the desired results. Additionally, the processes depicted in the drawings do not necessarily require the particular order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0251] Embodiment 2

[0252] See Figure 11 , Embodiment 2 of the present invention further provides an ICES model predictive control device based on a GRU-driven dynamic energy hub, including:

[0253] A DEH model construction module 001, configured to model the coupling relationship between the ICES input and output based on the coupling matrix of the multivariable nonlinear system to obtain a DEH model;

[0254] A device efficiency parameter correction model construction and training module 002, configured to construct a device efficiency parameter correction model based on a GRU network; train the device efficiency parameter correction model through a training strategy to obtain the trained device efficiency parameter correction model;

[0255] A device efficiency parameter correction model processing module 003, configured to input the input matrix of a set time period into the trained device efficiency parameter correction model for correction processing, and output the device efficiency parameter prediction matrix of the set time period;

[0256] A G-DEH model construction module 004, configured to construct a G-DEH model of the ICES based on the trained device efficiency parameter correction model and the DEH model;

[0257] An ICES optimal scheduling model construction module 005, configured to construct an ICES optimal scheduling model based on the G-DEH model;

[0258] A G-DEH model parameter update module 006, configured to update the efficiency parameters of the G-DEH model in the previous time period according to the device efficiency parameter prediction matrix of the set time period to obtain the G-DEH model and the ICES optimal scheduling model of the set time period;

[0259] An ICES optimal scheduling model solving module 007, configured to solve the ICES optimal scheduling model of the set time period to obtain the scheduling strategy of the set time period;

[0260] The scheduling policy execution module 008 is used to perform scheduling execution according to the scheduling policy in the set time period; after the scheduling execution is completed, ICES updates the set parameters and optimizes the scheduling policy for the next time period.

[0261] In this embodiment, in the DEH model construction module 001, the expression of the DEH model is:

[0262]

[0263] In the formula, L e,t and L h,t are the electricity and heat load demands of ICES at time t respectively; P e,t and P g,t are the electricity purchase and gas purchase powers of ICES at time t respectively; is the correction value of the heating efficiency of HP at time t; is the correction value of the heating efficiency of GB at time t; is the correction value of the power generation efficiency of CHP at time t; is the correction value of the thermoelectric ratio of CHP at time t; ν ee,t is the distribution factor of the electricity purchase power, indicating the proportion of the electricity purchase power directly supplied to the electric load at time t; (1 - ν ee,t ) is the proportion of the electricity purchase power converted into heat by HP at time t; ν gh,t is the distribution factor of the gas purchase power at time t, representing the proportion of natural gas supplied to CHP; (1 - ν gh,t ) is the proportion of the gas purchase power converted into heat by GB at time t; P BAT,t is the charge and discharge power of BAT at time t, taking a positive value for charging and a negative value for discharging; P TST,t is the charge and heat release power of TST at time t, taking a positive value for heat charging and a negative value for heat release.

[0264] In this embodiment, in the equipment efficiency parameter correction model construction and training module 002, the equipment efficiency parameter correction model includes a four-layer network of an input layer, a hidden layer, a fully connected layer, and an output layer; the input matrix enters the hidden layer through the input layer, is processed by the hidden layer, and outputs a hidden state matrix; the hidden state matrix is transformed and processed through the fully connected layer to obtain an equipment efficiency parameter prediction matrix; the equipment efficiency parameter prediction matrix is output through the output layer;

[0265] The expression of the input matrix is:

[0266]

[0267] In the formula, N GRU is the length of the prediction time domain of the equipment efficiency parameter correction model; k is the kth GRU of the hidden layer (1 ≤ k ≤ NGRU ) is the input matrix for the k-th GRU;

[0268] The expression of the hidden state matrix is:

[0269]

[0270] wherein, is the hidden state value at time k + t; N GRU is the number of GRUs;

[0271] The expression of the device efficiency parameter prediction matrix is:

[0272]

[0273] wherein, is t 0 + k moment prediction vector composed of device efficiency parameters.

[0274] In this embodiment, in the ICES optimal scheduling model construction module 005, the ICES optimal scheduling model aims at the optimal operation economy of ICES within the scheduling time domain; the expression of the ICES optimal scheduling model is:

[0275]

[0276] wherein, N c is the scheduling time domain; Δt is the scheduling time step; is the electricity purchase cost of ICES in time period t; P e (t) is the electricity purchase power of ICES in time period t; c e (t) is the electricity price in time period t; is the gas purchase cost of ICES in time period t; P g (t) is the gas purchase power of ICES in time period t; c g (t) is the gas price in time period t.

[0277] In this embodiment, in the ICES optimal scheduling model construction module 005, the constraint conditions of the ICES optimal scheduling model include: energy conversion device operation constraint, energy storage device operation constraint, energy balance constraint and superior network interaction constraint;

[0278] The energy conversion device operation constraint is:

[0279] 0 ≤ P out,i,t ≤ P cap,i

[0280] wherein, P out,i,t is the output power of the energy conversion device i in the ICES at time period t; Pcap,i Represents the rated output power of the energy conversion device i;

[0281] The operating constraints of the energy storage device are as follows:

[0282] W j,t+1 = W j,t (1 - σ j ) + (P j,C,t η j,C - P j,D,t / η j,D )Δt

[0283] In the formula, W j,t and W j,t+1 are the stored energy of the energy storage device j (j = 1, 2 corresponding to BAT and TST respectively) at time t and t + 1; σ j is the self-loss coefficient of device j; P j,C,t and P j,D,t are the charging and discharging powers of device j respectively; η j,C and η j,D are the charging and discharging efficiencies of device j respectively; Δt is the unit scheduling step size;

[0284] 0 ≤ P j,C,t U j,C,t ≤ P j,C,max

[0285] 0 ≤ P j,D,t U j,D,t ≤ P j,D,max

[0286] U j,C,t + U j,D,t ≤ 1

[0287] In the formula, P j,C,max and P j,D,max represent the maximum charging and discharging powers of the energy storage device j respectively; U j,C,t and U j,D,t are binary 0-1 variables representing the charging and discharging states of the energy storage device j, restricting the device from charging and discharging simultaneously, U j,C,t = 1 represents that the energy storage device j is in the charging state; U j,D,t = 1 represents that the device j is in the discharging state;

[0288] W j,min ≤ W j,t ≤ W j,max

[0289] In the formula, W j,min and W j,max are the upper and lower limit values of the allowable stored energy of the energy storage device j respectively;

[0290] W j,start = W j,end

[0291] Wherein, W j,start and W j,end are respectively the stored energy of the energy storage device j at the initial and termination moments of the scheduling period;

[0292] The energy balance constraint is:

[0293]

[0294] The upper-level network interaction constraint is:

[0295] P e (t) ≤ P e,cap

[0296] P g (t) ≤ P g,cap

[0297] Wherein, P e,cap and P g,cap are respectively the upper limits of the interaction power between the ICES and the power network and the natural gas network.

[0298] In this embodiment, in the ICES optimal scheduling model solving module 007, by calculating the absolute difference between the supply value and the true value of L e and L h during the scheduling day, the accuracy of the optimal scheduling strategy is evaluated; the calculation formula is:

[0299]

[0300]

[0301] Wherein, is the absolute difference between the supply value and the true value of L e in the optimal scheduling strategy during the scheduling day; and are respectively the supply error components of L e caused by the prediction error of the device efficiency parameter at time t and the supply error component of L e caused by the prediction error. The positive values of the two indicate that there is a shortage in the supply of L e under the optimal scheduling strategy, and vice versa, indicating that the supply of L e is excessive; e ; and are respectively the predicted values of the CHP power generation efficiency, CHP thermoelectric ratio, GB heating efficiency, and HP heating efficiency in the optimal scheduling strategy at time t; and They are the true values of the CHP power generation efficiency, CHP heat-electricity ratio, GB heating efficiency, and HP heating efficiency in period t, respectively; is the L in period t e true value; is the L in period t e predicted value; is the absolute difference between the supply value and the true value of L h in the optimized scheduling strategy within the scheduling day; and are the supply error components of L caused by the prediction error of the equipment efficiency parameters in period t h and the supply error components of L h caused by the prediction error, respectively. A positive value for both indicates a shortage in the supply of L under the optimized scheduling strategy, and vice versa, indicating h an excess supply; h is the L in period t h true value; is the L in period t h true value; is the L in period t h predicted value.

[0302] It should be noted that the information interaction, execution process, etc. among the above system modules are based on the same concept as the method embodiment in Embodiment 1 of the present application, and the technical effects brought by them are the same as those of the method embodiment of the present application. For specific content, reference can be made to the description in the method embodiment shown above in the present application, and details are not repeated here.

[0303] Embodiment 3

[0304] Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium, in which program codes for the ICES model predictive control method based on GRU-driven dynamic energy hub are stored, and the program codes include instructions for executing the ICES model predictive control method based on GRU-driven dynamic energy hub in Embodiment 1 or any possible implementation manner thereof.

[0305] The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center integrating one or more available media. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid-state disk (SSD)), etc.

[0306] Embodiment 4

[0307] Embodiment 4 of the present invention provides an electronic device, including: a memory and a processor;

[0308] The processor and the memory complete communication with each other through a bus; the memory stores program instructions executable by the processor, and the processor can execute the ICES model predictive control method of the GRU-driven dynamic energy hub according to Embodiment 1 or any possible implementation thereof by invoking the program instructions.

[0309] Specifically, the processor can be implemented by hardware or software. When implemented by hardware, the processor can be a logic circuit, an integrated circuit, etc.; when implemented by software, the processor can be a general-purpose processor, which is implemented by reading software code stored in the memory. The memory can be integrated in the processor or can exist independently outside the processor.

[0310] In the above embodiments, they can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, they can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a dedicated computer, a computer network, or other programmable systems. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wirelessly (such as infrared, wireless, microwave, etc.).

[0311] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing system. They can be concentrated on a single computing system or distributed on a network composed of multiple computing systems. Optionally, they can be implemented by program code executable by the computing system, so that they can be stored in a storage system and executed by the computing system. And in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules respectively, or multiple modules or steps of them can be made into a single integrated circuit module to implement. Thus, the present invention is not limited to any specific combination of hardware and software.

[0312] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made thereto based on the present invention, which will be obvious to those skilled in the art. Therefore, these modifications or improvements made without departing from the spirit of the present invention all fall within the scope of the present invention claimed.

Claims

1. ICES model predictive control method based on GRU driven dynamic energy hub, characterized in that: include: Based on the coupling matrix of the multivariable nonlinear system, the coupling relationship between ICES input and output is modeled to obtain the DEH model; Based on the GRU network, an equipment efficiency parameter correction model is constructed; the equipment efficiency parameter correction model is trained through a training strategy to obtain a trained equipment efficiency parameter correction model; Inputting the input matrix of the set time period into the trained equipment efficiency parameter correction model for correction processing, and outputting the equipment efficiency parameter prediction matrix of the set time period; Based on the trained equipment efficiency parameter correction model and the DEH model, construct the G-DEH model of ICES; Based on the G-DEH model, an ICES optimization scheduling model is constructed; According to the equipment efficiency parameter prediction matrix of the set time period, the efficiency parameters of the G-DEH model in the previous time period are updated to obtain the G-DEH model and the ICES optimization scheduling model of the set time period; Solving the ICES optimization scheduling model for the set time period to obtain a scheduling strategy for the set time period; According to the scheduling strategy of the set time period, ICES performs scheduling execution; after the scheduling execution is completed, ICES updates the set parameters and optimizes the scheduling strategy for the next time period.

2. The ICES model predictive control method based on GRU driven dynamic energy hub according to claim 1 is characterized in that: The expression of the DEH model is: Where, L e,t and L h,t are the electricity and heat load requirements of ICES during period t; P e,t and P g,t are the electricity and gas purchase power of ICES during period t respectively; is the correction value of HP heating efficiency during period t; is the correction value of GB heating efficiency during period t; is the corrected value of CHP power generation efficiency during period t; is the corrected value of the CHP thermal-electric ratio during period t; ν ee,t is the allocation factor of purchased power, indicating the proportion of purchased power directly supplied to the electric load during period t; (1-ν ee,t ) is the proportion of purchased electricity converted from HP to produce heat during period t; ν gh,t is the allocation factor of gas purchase power in period t, representing the proportion of natural gas supplied to CHP; (1-ν gh,t ) is the ratio of the purchased gas power converted into heat by GB during period t; P BAT,t is the charging and discharging power of BAT during period t, with positive value for charging and negative value for discharging; P TST,t It is the charging and discharging power of TST in period t. Charging takes positive value and discharging takes negative value.

3. The ICES model predictive control method based on GRU driven dynamic energy hub according to claim 2 is characterized in that: The equipment efficiency parameter correction model includes a four-layer network of an input layer, a hidden layer, a fully connected layer and an output layer; the input matrix enters the hidden layer through the input layer, is processed by the hidden layer, and outputs a hidden state matrix; the hidden state matrix is ​​transformed by the fully connected layer to obtain an equipment efficiency parameter prediction matrix; the equipment efficiency parameter prediction matrix is ​​output through the output layer; The expression of the input matrix is: Where N GRU is the length of the prediction time domain of the equipment efficiency parameter correction model; k is the kth GRU (1≤k≤N GRU ); is the input matrix of the kth GRU; The expression of the hidden state matrix is: In the formula, is the hidden state value at time k+t; N GRU is the number of GRU; The expression of the equipment efficiency parameter prediction matrix is: In the formula, It is the prediction vector composed of equipment efficiency parameters at time t0+k.

4. The ICES model predictive control method based on GRU driven dynamic energy hub according to claim 3 is characterized in that: The ICES optimization scheduling model aims to optimize the economic efficiency of ICES operation within the scheduling time domain; the expression of the ICES optimization scheduling model is: Where N c is the scheduling time domain; Δt is the scheduling time step; P is the electricity purchase cost of ICES during period t; e (t) is the power purchased by ICES during period t; c e (t) is the electricity price during period t; P is the gas purchase cost of ICES during period t; g (t) is the gas purchasing power of ICES during period t; c g (t) is the gas price during period t.

5. The ICES model predictive control method based on GRU driven dynamic energy hub according to claim 4 is characterized in that: The constraints of the ICES optimization scheduling model include: energy conversion equipment operation constraints, energy storage equipment operation constraints, energy balance constraints and upper network interaction constraints; The energy conversion equipment operation constraints are: 0≤P out,i,t ≤P cap,i Where P out,i,t is the output power of energy conversion device i in ICES during period t; P cap,i represents the rated output power of energy conversion device i; The energy storage device operation constraints are: W j,t+1 =W j,t (1-p j )+(P j,C,t or j,C -P j,D,t / or j,D )Δt Where W j,t and W j,t+1 are the energy storage capacity of energy storage device j (j=1, 2 correspond to BAT and TST respectively) at time t and t+1 respectively; σ j is the self-loss coefficient of device j; P j,C,t and P j,D,t are the charging and discharging power of device j respectively; η j,C and η j,D are the charging and discharging efficiencies of device j respectively; Δt is the unit scheduling step; 0≤P j,C,t IN j,C,t ≤P j,C,max 0≤P j,D,t IN j,D,t ≤P j,D,max IN j,C,t +U j,D,t ≤1 Where P j,C,max and P j,D,max Respectively represent the maximum charging and discharging power of energy storage device j; U j,C,t and U j,D,t It is a binary 0-1 variable representing the charging and discharging status of energy storage device j, which restricts the device from charging and discharging at the same time. j,C,t =1 means that the energy storage device j is in the charging state; U j,D,t =1 means that device j is in the energy-discharging state; IN j,min ≤In j,t ≤In j,max Where W j,min and W j,max are the upper and lower limits of the energy storage capacity allowed by energy storage device j, respectively; IN j,start =In j,end Where W j,start and W j,end are the energy storage capacity of energy storage device j at the initial and final moments of the scheduling period, respectively; The energy balance constraint is: The upper network interaction constraints are: P e (t)≤P e,cap P g (t)≤P g,cap Where P e,cap and P g,cap They are the upper limits of the interaction power between ICES and the power grid and the natural gas grid, respectively.

6. The ICES model predictive control method based on GRU driven dynamic energy hub according to claim 5 is characterized in that: By scheduling the day L e and L h The absolute difference between the supply value and the true value is calculated to evaluate the accuracy of the optimized scheduling strategy; the calculation formula is: In the formula, L for scheduling day e The absolute difference between the supplied value and the true value in the optimal scheduling policy; and are L caused by the prediction error of equipment efficiency parameters in period t. e Supply error component and L e L caused by prediction error e Supply error component, both are positive, indicating that L e There is a shortage in supply, otherwise, it means L e Oversupply; and are the predicted values ​​of CHP power generation efficiency, CHP heat-to-electricity ratio, GB heating efficiency and HP heating efficiency in the optimal scheduling strategy for period t; and are the true values ​​of CHP power generation efficiency, CHP heat-to-electricity ratio, GB heating efficiency and HP heating efficiency during period t, respectively; is L in period t e True value; is L in period t e Predicted value; L for scheduling day h The absolute difference between the supplied value and the true value in the optimal scheduling policy; and are L caused by the prediction error of equipment efficiency parameters in period t. h Supply error component and L h L caused by prediction error h Supply error component, both are positive, indicating that L h There is a shortage in supply, otherwise, it means L h Oversupply; is L in period t h True value; is L in period t h Predicted value.

7. An ICES model predictive control device based on a GRU-driven dynamic energy hub, using an ICES model predictive control method based on a GRU-driven dynamic energy hub according to any one of claims 1 to 6, characterized in that: include: The DEH model building module is used to model the coupling relationship between ICES input and output based on the coupling matrix of the multivariable nonlinear system to obtain the DEH model; The equipment efficiency parameter correction model construction and training module is used to construct an equipment efficiency parameter correction model based on the GRU network; the equipment efficiency parameter correction model is trained through a training strategy to obtain the trained equipment efficiency parameter correction model; An equipment efficiency parameter correction model processing module is used to input the input matrix of a set time period into the trained equipment efficiency parameter correction model for correction processing, and output the equipment efficiency parameter prediction matrix of the set time period; A G-DEH model building module, used to build a G-DEH model of ICES based on the trained equipment efficiency parameter correction model and the DEH model; An ICES optimization scheduling model construction module is used to construct an ICES optimization scheduling model based on the G-DEH model; A G-DEH model parameter updating module, used to update the efficiency parameters of the G-DEH model in the previous time period according to the equipment efficiency parameter prediction matrix of the set time period, and obtain the G-DEH model and the ICES optimization scheduling model of the set time period; An ICES optimization scheduling model solving module, used to solve the ICES optimization scheduling model for the set time period to obtain a scheduling strategy for the set time period; The scheduling strategy execution module is used for ICES to perform scheduling execution according to the scheduling strategy of the set time period; after the scheduling execution is completed, ICES updates the set parameters and optimizes the scheduling strategy for the next time period.

8. The ICES model predictive control device based on GRU driven dynamic energy hub according to claim 7, characterized in that: In the DEH model construction module, the expression of the DEH model is: Where, L e,t and L h,t are the electricity and heat load requirements of ICES during period t; P e,t and P g,t are the electricity and gas purchase power of ICES during period t respectively; is the correction value of HP heating efficiency during period t; is the correction value of GB heating efficiency during period t; is the corrected value of CHP power generation efficiency during period t; is the corrected value of the CHP thermal-electric ratio during period t; ν ee,t is the allocation factor of purchased power, indicating the proportion of purchased power directly supplied to the electric load during period t; (1-ν ee,t ) is the proportion of purchased electricity converted from HP to produce heat during period t; ν gh,t is the allocation factor of gas purchase power in period t, representing the proportion of natural gas supplied to CHP; (1-ν gh,t ) is the ratio of the purchased gas power converted into heat by GB during period t; P BAT,t is the charging and discharging power of BAT during period t, with positive value for charging and negative value for discharging; P TST,t It is the charging and discharging power of TST in period t. Charging takes positive value and discharging takes negative value.

9. The ICES model predictive control device based on GRU driven dynamic energy hub according to claim 8, characterized in that: In the equipment efficiency parameter correction model construction and training module, the equipment efficiency parameter correction model includes a four-layer network of an input layer, a hidden layer, a fully connected layer and an output layer; the input matrix enters the hidden layer through the input layer, is processed by the hidden layer, and outputs a hidden state matrix; the hidden state matrix is ​​transformed by the fully connected layer to obtain an equipment efficiency parameter prediction matrix; the equipment efficiency parameter prediction matrix is ​​output through the output layer; The expression of the input matrix is: Where N GRU is the length of the prediction time domain of the equipment efficiency parameter correction model; k is the kth GRU (1≤k≤N GRU ); is the input matrix of the kth GRU; The expression of the hidden state matrix is: In the formula, is the hidden state value at time k+t; N GRU is the number of GRU; The expression of the equipment efficiency parameter prediction matrix is: In the formula, It is the prediction vector composed of equipment efficiency parameters at time t0+k.

10. The ICES model predictive control device based on GRU driven dynamic energy hub according to claim 9, characterized in that: In the ICES optimization scheduling model construction module, the ICES optimization scheduling model aims to optimize the economic efficiency of ICES operation within the scheduling time domain; the expression of the ICES optimization scheduling model is: Where N c is the scheduling time domain; Δt is the scheduling time step; P is the electricity purchase cost of ICES during period t; e (t) is the power purchased by ICES during period t; c e (t) is the electricity price during period t; P is the gas purchase cost of ICES during period t; g (t) is the gas purchasing power of ICES during period t; c g (t) is the gas price during period t.

Images