A method for constructing a flexible operation domain of an electricity-gas-hydrogen hybrid integrated energy system

Abstract

The present invention proposes a method for constructing a flexible operation domain of an electric-gas-hydrogen hybrid integrated energy system. By using a graph convolutional neural network, the feasibility of the electric-hydrogen power point is explored, and the intrinsic connection between the system topology, operation constraints, and scenario data is realized. On the basis of being able to accurately identify whether the electric-hydrogen power point is a feasible point, the flexible operation domain of the electric-gas-hydrogen hybrid integrated energy system under different scenario conditions is quickly constructed. With the goal of minimizing the load shedding of the electric-gas-hydrogen hybrid integrated energy system, a basic physical model for identifying whether the electric-hydrogen power point is a feasible point is established on the basis of the electric-gas-hydrogen hybrid integrated energy system model; a graph convolutional neural network is introduced, and the changes in the gas grid operation topology are considered, and a model for constructing the flexible operation domain of the electric-gas-hydrogen hybrid integrated energy system based on the graph convolutional neural network is established.

Description

Technical Field

[0001] The present invention belongs to the technical field of coordinated optimization of electricity-natural gas networks, and specifically relates to a method for constructing a flexible operation domain of an electricity-gas-hydrogen hybrid integrated energy system. Background Art

[0002] The global energy system is undergoing a transition from fossil fuels to renewable energy to cope with climate change and achieve low-carbon development goals. Renewable energy sources such as wind and solar energy are intermittent and volatile, which leads to challenges in the balance of supply and demand in the power system. Hydrogen energy, as a clean energy, is regarded as one of the key technologies for energy transformation. Electric hydrogen production technology can convert excess electricity into hydrogen. As an energy carrier, hydrogen can be transported or stored through pipelines, thereby balancing the load of the power grid and the gas grid and improving the utilization rate of renewable energy. In addition, ensuring that the load is met is of great significance for ensuring the stable operation of the power system, promoting economic development, maintaining social stability, promoting technological progress, and achieving environmental sustainability. Therefore, as an important carrier of supply and demand balance, it is very important that the power of the electric hydrogen production device can timely meet all the operating constraints in the electric-gas-hydrogen hybrid integrated energy system without shedding the load. That is, the timely construction of the flexible operation domain of the integrated energy system becomes crucial.

[0003] However, traditional construction methods often take a long time to identify feasible power points for hydrogen production and to construct a flexible operating domain for the system, which is very unfavorable for absorbing intermittent and fluctuating renewable energy. And there are certain limitations in using general machine learning methods to identify whether a power point is feasible. As a powerful deep learning model, graph convolutional neural network can process graph structured data and is very suitable for describing and analyzing complex relationships and dynamic changes in the system. It can be used to predict and evaluate the dynamic stability of the system, help operators formulate preventive control strategies, optimize energy distribution, and improve the overall efficiency and reliability of the system. Based on the graph convolutional neural network model, the complex interactions between the hydrogen production device and other parts of the system can be learned to provide support for decision-making. Therefore, it is urgent to develop a method for constructing the flexible operating domain of the electric-gas-hydrogen hybrid integrated energy system based on graph convolutional neural network, so as to greatly reduce the construction time of the flexible operating domain of the electric-gas-hydrogen integrated energy system while maintaining the high accuracy of feasibility identification of hydrogen production power points and construction of the system flexible operating domain. Summary of the invention

[0004] Purpose of the invention: The technical problem to be solved by the present invention is to provide a method for constructing a flexible operation domain of an electric-gas-hydrogen hybrid integrated energy system in response to the shortcomings of traditional technologies. The present invention takes into account the coordinated coupling of the power system and the natural gas system, and constructs the flexible operation domain of the system by introducing a graph convolutional neural network to mine the characteristics of system topology, operation constraints, scenario data and other conditions. This method can further be used to obtain the flexible operation domain of the system under different scenario data. The present invention can not only realize the high-accuracy identification of whether the power point of hydrogen production is a feasible point and the high-accuracy construction of the flexible operation domain of the electric-gas-hydrogen hybrid integrated energy system, but also greatly reduces the time consumed in identifying whether the power point of hydrogen production is a feasible point and constructing the flexible operation domain, thereby ensuring the coordinated coupling operation of the power point of hydrogen production.

[0005] Technical solution: In order to solve the above technical problems, the present invention provides a method for constructing a flexible operation domain of an electricity-gas-hydrogen hybrid integrated energy system, comprising the following steps:

[0006] Step 1, obtaining network coefficients and operation coefficients of the power grid, gas grid, and electric hydrogen production model, wherein the network coefficients include power grid topology, gas grid topology, line resistance and impedance, natural gas, and hydrogen calorific value, and the operation coefficients include generator set power generation coefficients, wind power photovoltaic inverter coefficients, gas source parameters, electric hydrogen production device parameters, and booster station parameters;

[0007] Step 2: Obtain the scenario data of power grid load demand, gas grid load demand, wind power and photovoltaic output, and power operation data set of hydrogen production device at different time sections;

[0008] Step 3: Taking the minimum load shedding of the electricity-gas-hydrogen hybrid integrated energy system as the objective function, a basic physical model for identifying whether the power point of hydrogen production is a feasible point is established. Based on the obtained network coefficients, operating coefficients and scenario data, a nonlinear optimization solver is used to solve the basic physical model to identify feasible points and infeasible points under corresponding operating conditions, and a neural network training sample data set is generated under different gas network topologies with scenario data as the feature and corresponding feasible point or infeasible point recognition results as the label;

[0009] Step 4: For the neural network training sample data set generated in step 3, train and establish a model for constructing the flexible operation domain of the electric-gas-hydrogen hybrid integrated energy system based on a graph convolutional neural network, which takes the scenario data at other different time sections as input and the feasible points or infeasible points identified in batches as output;

[0010] Step 5. For any gas grid topology involved in the training, the electric hydrogen production power point is uniformly given with a certain accuracy within the power constraint of the electric hydrogen production device. Based on the graph convolutional neural network established in step 4, the model of the flexible operation domain of the electric-gas-hydrogen hybrid integrated energy system is constructed. The load demand data, wind power and photovoltaic output data and all uniformly given electric hydrogen production power point data at any time section are batch input, all feasible points are identified and a convex envelope is constructed to form the system's flexible operation domain under the gas grid topology, time section and a certain accuracy.

[0011] Furthermore, in step 3, the basic physical model for identifying whether the power point of hydrogen production by electricity is a feasible point with the minimum load shedding of the electricity-gas-hydrogen hybrid integrated energy system as the objective function is as follows:

[0012] (1) Objective function:

[0013]

[0014] In the formula, It represents the required power of the electric load d; It represents the actual absorbed power of the electric load d; It represents the energy demand of gas load e; It represents the actual energy consumed by the gas load e;

[0015] The objective function of formula (A-1) represents the minimum sum of the power cut load and the gas cut load;

[0016] (2) Power system operation model:

[0017]

[0018] θ REF =0(A-8)

[0019] In the formula, subscripts i and j represent power buses; subscript v represents generators; subscript r represents new energy units; subscript d represents power loads; subscript h represents hydrogen production by electricity; subscript REF represents reference buses; U v (i) represents the set of generator sets connected to bus i; U r (i) represents the set of new energy units connected to bus i; U d (i) represents the set of power loads connected to bus i; U h (i) represents the set of electric hydrogen production connected to bus i; U(i) represents the set of lines connected to bus i; represents the actual power generated by the generator v; P r W Indicates the grid-connected power of the new energy unit r; represents the power of hydrogen production from electricity; θ i represents the voltage phase angle of bus i; x ij and They represent the susceptance and transmission capacity of line ij respectively; and Respectively represent the minimum and maximum power generation of unit v; Indicates the predicted power of new energy units; They represent the upper and lower limits of the electric power consumed by electric hydrogen production h; θ REF Represents the phase angle of the reference bus.

[0020] Formula (A-2) is the node active power balance equation; Formula (A-3) is the active load demand constraint; Formula (A-4) is the line transmission capacity constraint; Formula (A-5) is the upper and lower limits of the generator active output; Formula (A-6) is the grid-connected new energy output constraint; Formula (A-7) is the upper and lower limits of the power of hydrogen production; Formula (A-8) stipulates that the phase angle of the reference bus is 0;

[0021] (3) Electricity-generated hydrogen model:

[0022] The electric hydrogen production device is an important component of the coupling between the power grid and the gas grid. In the power system, the electric hydrogen production device is equivalent to a controllable load, and in the natural gas system, the electric hydrogen production device is equivalent to a gas source. Its model is as follows:

[0023]

[0024] In the formula, It represents the hydrogen energy produced by electric hydrogen production h; It represents the hydrogen flow rate produced by electric hydrogen production h; is the maximum hydrogen energy that can be produced by hydrogen production from electricity; It represents the calorific value of hydrogen, which is 12.75MJ / m 3 ; η h It represents the energy conversion efficiency of electric hydrogen production;

[0025] Formula (A-9) represents the relationship between the electric energy consumed by electric hydrogen production and the energy of hydrogen produced; Formula (A-10) represents the relationship between the energy of hydrogen produced and the flow rate; Formula (A-11) describes the upper and lower limit constraints of the hydrogen energy produced by electric hydrogen production;

[0026] (4) Steady-state natural gas system operation model taking into account pipeline hydrogen blending:

[0027]

[0028]

[0029] In the formula, subscripts m and n represent natural gas nodes; subscript e represents gas load; subscript k represents natural gas compression station; subscript w represents natural gas source; subscript mn represents pipeline mn between natural gas nodes m and n; G w (m) represents the gas source set connected to node m; G h (m) represents the electric hydrogen production set connected to node m; G e (m) represents the gas load set connected to node m; G g (m) represents the set of gas generators connected to node m; G k (m) represents the set of booster stations connected to node m; G(m) represents the set of pipelines connected to node m; It represents the energy provided by the gas source w; It represents the natural gas energy consumed by gas generator set v; E mn With E nm They represent the energy of the natural gas at the beginning and end of the pipeline mn respectively; represents the energy of the natural gas flowing through the compression station k; θ k represents the percentage of natural gas consumption at the booster station k to the transmission flow; F mn With F nm They represent the natural gas flow at the beginning and end of the pipeline mn respectively; W mn represents the Weymouth constant of the pipeline mn; π m and π n It represents the pressure of nodes m and n; Represents the calorific value of the pipeline mn; H gas It represents the calorific value of natural gas, which is 38.29MJ / m 3 ;H m It represents the heat value of node m; It represents the natural gas flow provided by the gas source w; represents the natural gas flow through the booster station k; η v It represents the power generation efficiency of the gas unit; and They represent the maximum and minimum pressure values ​​of node m respectively; It represents the upper limit of the flow rate at the head end of the pipeline; and They represent the maximum and minimum supply energies of the gas source w respectively; and They represent the inlet and outlet pressures at the booster station k respectively; and They represent the maximum and minimum pressurization ratios of the pressurization station k respectively; represents the maximum transmission energy of the booster station k; α represents the hydrogen mixing ratio;

[0030] Formula (A-12) is the energy balance constraint equation; Formula (A-13) is an equation that describes the nonlinear relationship between pipeline flow and pressure drop at the beginning and end of the pipeline; Formula (A-14) is a necessary constraint condition for a steady-state system, that is, without considering the pipe storage, the flow at the beginning of the pipeline is considered to be equal to the flow at the end; Formula (A-15) is a formula for calculating the calorific value: the energy flowing into the node is equal to the energy flowing out of the node plus the energy consumption of the booster station; Formula (A-16) is an energy calculation formula; Formula (A-17) is the gas load demand energy constraint; Formula (A-18) is the node pressure constraint; Formula (A-19) is the constraint on the flow at the beginning and end of the pipeline; Formula (A-20) is the gas source supply constraint; Formula (A-21) is the maximum and minimum boost ratio constraint of the booster station; Formula (A-22) is the booster station transmission energy constraint; Formula (A-23) is the pipeline calorific value constraint; Formula (A-24) is the node calorific value constraint;

[0031] Based on the above basic physical model for identifying whether the hydrogen production power point is a feasible point, the hydrogen production power point under the condition that all constraints of the basic physical model can be satisfied without load shedding is called a feasible point under the corresponding conditions; the hydrogen production power point under the condition that all constraints of the basic physical model can be satisfied only by load shedding is called an infeasible point under the corresponding conditions. The power set space of all feasible points under the same operating conditions constitutes the flexible operating domain of the system under this condition.

[0032] Furthermore, in step 4, the model of the flexible operation domain of the electricity-gas-hydrogen hybrid integrated energy system is constructed based on the graph convolutional neural network as follows:

[0033] (1) Network structure model:

[0034] The constructed graph convolutional neural network structure mainly consists of three graph convolutional layers plus one linear layer, and its mathematical expression is as follows:

[0035]

[0036] In the formula, x represents the input initial feature matrix composed of all node features; is the symmetric normalized adjacency matrix obtained by normalizing the adjacency matrix A; ReLU(·) is the rectified linear unit activation function; the superscripts (1), (2), (3), and (4) denote the 1st, 2nd, 3rd, and 4th layers; z (1) ,z (2) ,z (3) They are the output feature matrix representations of the 1st, 2nd, and 3rd layer graph convolution operations respectively; The feature representation of node k that represents the output of the 3rd layer graph convolution operation; represents the global feature; k represents the node number; N is the number of nodes; W (1) ,W (2) ,W (3) are the parameter weight matrices of the 1st, 2nd, and 3rd graph convolutional layers respectively; W (4) represents the parameter weight matrix of the 4th linear layer; b (1) ,b (2) ,b (3) are the bias vectors of the 1st, 2nd, and 3rd graph convolutional layers respectively; b (4) represents the bias vector of the 4th linear layer; y represents the final output recognition result, that is, the classification result of the graph;

[0037] In the forward propagation process of the network structure, the node features pass through three layers of graph convolutional layers. Each layer uses the ReLU activation function to map the features to a higher dimension. Then, the multi-dimensional feature vectors after propagation through the three layers of convolutional layers are summarized by global average pooling to integrate the node features in the graph into an overall feature. Finally, they are mapped to the output space through linear transformation to obtain the final recognition result. The first layer is the input layer, and the input dimension corresponds to the number of features of each node; the second layer input dimension is equal to the first layer output dimension; the third layer input dimension is equal to the second layer output dimension; the fourth layer input dimension is equal to the third layer output dimension, and the output dimension is 1, and the output is the probability of being identified as a positive category;

[0038] (2) Graph convolution operation model:

[0039] The general mathematical model of the graph convolution operation of each node in the network structure is as follows:

[0040]

[0041] In the formula, l represents the node number; x l represents the feature representation of node l; z k represents the new feature representation of node k; neighbors(k) represents the set of neighboring nodes of node k; deg(k) represents the degree of node k; deg(l) represents the degree of node l; W represents the weight matrix; b represents the bias vector;

[0042] (3) Loss function:

[0043] Select the function BCEWithLogitsLoss(·) as the loss function:

[0044]

[0045] In the formula, logits represents the output of the model without the activation function, which is the score belonging to the positive category; labels represents the true binary classification label, which is 0 or 1.

[0046] Beneficial effects: Compared with the traditional technology, the technical solution of the present invention has the following beneficial effects:

[0047] Compared with the traditional basic scheme for constructing a flexible operation domain of an electricity-gas-hydrogen integrated energy system and the method for constructing a flexible operation domain of an electricity-gas-hydrogen integrated energy system based on a general deep neural network without topological features, the technical scheme of the present invention constructs a flexible operation domain of an electricity-gas-hydrogen integrated energy system by introducing a graph convolutional neural network to mine the characteristics of conditions such as system topology, operation constraints, and scenario data. The results of the example test show that compared with the traditional scheme for constructing a flexible operation domain of an electricity-gas-hydrogen integrated energy system, the present invention greatly reduces the time consumed for operating domain construction on the basis of high accuracy. Compared with the method for constructing a flexible operation domain of an electricity-gas-hydrogen integrated energy system based on a general deep neural network without topological features, the construction accuracy of the method of the present invention is higher, and the time for neural network training in the method of the present invention is also relatively reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 is a flow chart of the method of the present invention;

[0049] Figure 2 This is an example diagram of a comprehensive energy system with three hydrogen production devices connected;

[0050] Figure 3 It is a comparison chart of the results of the method of constructing a flexible operation domain based on graph convolutional neural network and the basic solution of constructing a flexible operation domain based on the traditional method under the scenario data of a certain time section;

[0051] Figure 4 is with Figure 3 A comparison chart of the results of the method of constructing a flexible operating domain based on a general deep neural network without topological features and the basic solution of constructing a flexible operating domain based on traditional methods under the same time section scene data. DETAILED DESCRIPTION

[0052] The present invention is further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent modifications of the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0053] like Figure 1 As shown, the invention provides a method for constructing a flexible operation domain of an electricity-gas-hydrogen hybrid integrated energy system, comprising the following steps:

[0054] Step 1, obtaining network coefficients and operation coefficients of the power grid, gas grid, and electric hydrogen production model, wherein the network coefficients include power grid topology, gas grid topology, line resistance and impedance, natural gas, and hydrogen calorific value, and the operation coefficients include generator set power generation coefficients, wind power photovoltaic inverter coefficients, gas source parameters, electric hydrogen production device parameters, and booster station parameters;

[0055] Step 2: Obtain the scenario data of power grid load demand, gas grid load demand, wind power and photovoltaic output, and power operation data set of hydrogen production device at different time sections;

[0056] Step 3: Taking the minimum load shedding of the electricity-gas-hydrogen hybrid integrated energy system as the objective function, a basic physical model for identifying whether the power point of hydrogen production is a feasible point is established. Based on the obtained network coefficients, operating coefficients and scenario data, a nonlinear optimization solver is used to solve the basic physical model to identify feasible points and infeasible points under corresponding operating conditions, and a neural network training sample data set is generated under different gas network topologies with scenario data as the feature and corresponding feasible point or infeasible point recognition results as the label;

[0057] Step 4: For the neural network training sample data set generated in step 3, train and establish a model for constructing the flexible operation domain of the electric-gas-hydrogen hybrid integrated energy system based on a graph convolutional neural network, which takes the scenario data at other different time sections as input and the feasible points or infeasible points identified in batches as output;

[0058] Step 5. For any gas grid topology involved in the training, the electric hydrogen production power point is uniformly given with a certain accuracy within the power constraint of the electric hydrogen production device. Based on the graph convolutional neural network established in step 4, the model of the flexible operation domain of the electric-gas-hydrogen hybrid integrated energy system is constructed. The load demand data, wind power and photovoltaic output data and all uniformly given electric hydrogen production power point data at any time section are batch input, all feasible points are identified and a convex envelope is constructed to form the system's flexible operation domain under the gas grid topology, time section and a certain accuracy.

[0059] Furthermore, in step 3, the basic physical model for identifying whether the power point of hydrogen production by electricity is a feasible point with the minimum load shedding of the electricity-gas-hydrogen hybrid integrated energy system as the objective function is as follows:

[0060] (1) Objective function:

[0061]

[0062] In the formula, It represents the power demand of the electric load d; It represents the actual absorbed power of the electric load d; It represents the energy demand of gas load e; It represents the actual energy consumed by the gas load e;

[0063] The objective function of formula (A-1) represents the minimum sum of the power cut load and the gas cut load;

[0064] (2) Power system operation model:

[0065]

[0066] θ REF =0(A-8)

[0067] In the formula, subscripts i and j represent power buses; subscript v represents generators; subscript r represents new energy units; subscript d represents power loads; subscript h represents hydrogen production by electricity; subscript REF represents reference buses; U v (i) represents the set of generator sets connected to bus i; U r (i) represents the set of new energy units connected to bus i; U d (i) represents the set of power loads connected to bus i; U h (i) represents the set of electric hydrogen production connected to bus i; U(i) represents the set of lines connected to bus i; represents the actual power generated by the generator v; P r W Indicates the grid-connected power of the new energy unit r; represents the power of hydrogen production from electricity; θ i represents the voltage phase angle of bus i; x ij and They represent the susceptance and transmission capacity of line ij respectively; and Respectively represent the minimum and maximum power generation of unit v; Indicates the predicted power of new energy units; They represent the upper and lower limits of the electric power consumed by electric hydrogen production h; θ REF Represents the phase angle of the reference bus.

[0068] Formula (A-2) is the node active power balance equation; Formula (A-3) is the active load demand constraint; Formula (A-4) is the line transmission capacity constraint; Formula (A-5) is the upper and lower limits of the generator active output; Formula (A-6) is the grid-connected new energy output constraint; Formula (A-7) is the upper and lower limits of the power of hydrogen production; Formula (A-8) stipulates that the phase angle of the reference bus is 0;

[0069] (3) Electricity-generated hydrogen model:

[0070] The electric hydrogen production device is an important component of the coupling between the power grid and the gas grid. In the power system, the electric hydrogen production device is equivalent to a controllable load, and in the natural gas system, the electric hydrogen production device is equivalent to a gas source. Its model is as follows:

[0071]

[0072] In the formula, It represents the hydrogen energy produced by electric hydrogen production h; It represents the hydrogen flow rate produced by electric hydrogen production h; is the maximum hydrogen energy that can be produced by hydrogen production from electricity; It represents the calorific value of hydrogen, which is 12.75MJ / m 3 ; η h It represents the energy conversion efficiency of electric hydrogen production;

[0073] Formula (A-9) represents the relationship between the electric energy consumed by electric hydrogen production and the energy of hydrogen produced; Formula (A-10) represents the relationship between the energy of hydrogen produced and the flow rate; Formula (A-11) describes the upper and lower limit constraints of the hydrogen energy produced by electric hydrogen production;

[0074] (4) Steady-state natural gas system operation model taking into account pipeline hydrogen blending:

[0075]

[0076]

[0077] In the formula, subscripts m and n represent natural gas nodes; subscript e represents gas load; subscript k represents natural gas compression station; subscript w represents natural gas source; subscript mn represents pipeline mn between natural gas nodes m and n; G w (m) represents the gas source set connected to node m; G h (m) represents the electric hydrogen production set connected to node m; G e (m) represents the gas load set connected to node m; G g (m) represents the set of gas generators connected to node m; G k (m) represents the set of booster stations connected to node m; G(m) represents the set of pipelines connected to node m; It represents the energy provided by the gas source w; It represents the natural gas energy consumed by gas generator set v; E mn With E nm They represent the energy of the natural gas at the beginning and end of the pipeline mn respectively; represents the energy of the natural gas flowing through the compression station k; θ krepresents the percentage of natural gas consumption at the booster station k to the transmission flow; F mn With F nm They represent the natural gas flow at the beginning and end of the pipeline mn respectively; W mn represents the Weymouth constant of the pipeline mn; π m and π n It represents the pressure of nodes m and n; Represents the calorific value of the pipeline mn; H gas It represents the calorific value of natural gas, which is 38.29MJ / m 3 ;H m It represents the heat value of node m; It represents the natural gas flow provided by the gas source w; represents the natural gas flow through the booster station k; η v It represents the power generation efficiency of the gas unit; and They represent the maximum and minimum pressure values ​​of node m respectively; It represents the upper limit of the flow rate at the head end of the pipeline; and They represent the maximum and minimum supply energies of the gas source w respectively; and They represent the inlet and outlet pressures at the booster station k respectively; and They represent the maximum and minimum pressurization ratios of the pressurization station k respectively; represents the maximum transmission energy of the booster station k; α represents the hydrogen mixing ratio;

[0078] Formula (A-12) is the energy balance constraint equation; Formula (A-13) is an equation that describes the nonlinear relationship between pipeline flow and pressure drop at the beginning and end of the pipeline; Formula (A-14) is a necessary constraint condition for a steady-state system, that is, without considering the pipe storage, the flow at the beginning of the pipeline is considered to be equal to the flow at the end; Formula (A-15) is a formula for calculating the calorific value: the energy flowing into the node is equal to the energy flowing out of the node plus the energy consumption of the booster station; Formula (A-16) is an energy calculation formula; Formula (A-17) is the gas load demand energy constraint; Formula (A-18) is the node pressure constraint; Formula (A-19) is the constraint on the flow at the beginning and end of the pipeline; Formula (A-20) is the gas source supply constraint; Formula (A-21) is the maximum and minimum boost ratio constraint of the booster station; Formula (A-22) is the booster station transmission energy constraint; Formula (A-23) is the pipeline calorific value constraint; Formula (A-24) is the node calorific value constraint;

[0079] Based on the above basic physical model for identifying whether the hydrogen production power point is a feasible point, the hydrogen production power point under the condition that all constraints of the basic physical model can be satisfied without load shedding is called a feasible point under the corresponding conditions; the hydrogen production power point under the condition that all constraints of the basic physical model can be satisfied only by load shedding is called an infeasible point under the corresponding conditions. The power set space of all feasible points under the same operating conditions constitutes the flexible operating domain of the system under this condition.

[0080] Furthermore, in step 4, the model of the flexible operation domain of the electricity-gas-hydrogen hybrid integrated energy system is constructed based on the graph convolutional neural network as follows:

[0081] (1) Network structure model:

[0082] The constructed graph convolutional neural network structure mainly consists of three graph convolutional layers plus one linear layer, and its mathematical expression is as follows:

[0083]

[0084]

[0085] In the formula, x represents the input initial feature matrix composed of all node features; is the symmetric normalized adjacency matrix obtained by normalizing the adjacency matrix A; ReLU(·) is the rectified linear unit activation function; the superscripts (1), (2), (3), and (4) denote the 1st, 2nd, 3rd, and 4th layers; z (1) ,z (2) ,z (3) They are the output feature matrix representations of the 1st, 2nd, and 3rd layer graph convolution operations respectively; The feature representation of node k that represents the output of the 3rd layer graph convolution operation; represents the global feature; k represents the node number; N is the number of nodes; W (1) ,W (2) ,W (3) are the parameter weight matrices of the 1st, 2nd, and 3rd graph convolutional layers respectively; W (4) represents the parameter weight matrix of the 4th linear layer; b (1) ,b (2) ,b (3) are the bias vectors of the 1st, 2nd, and 3rd graph convolutional layers respectively; b (4) represents the bias vector of the 4th linear layer; y represents the final output recognition result, that is, the classification result of the graph;

[0086] In the forward propagation process of the network structure, the node features pass through three layers of graph convolutional layers. Each layer uses the ReLU activation function to map the features to a higher dimension. Then, the multi-dimensional feature vectors after propagation through the three layers of convolutional layers are summarized by global average pooling to integrate the node features in the graph into an overall feature. Finally, they are mapped to the output space through linear transformation to obtain the final recognition result. The first layer is the input layer, and the input dimension corresponds to the number of features of each node; the second layer input dimension is equal to the first layer output dimension; the third layer input dimension is equal to the second layer output dimension; the fourth layer input dimension is equal to the third layer output dimension, and the output dimension is 1, and the output is the probability of being identified as a positive category;

[0087] (2) Graph convolution operation model:

[0088] The general mathematical model of the graph convolution operation of each node in the network structure is as follows:

[0089]

[0090] In the formula, l represents the node number; x l represents the feature representation of node l; z k represents the new feature representation of node k; neighbors(k) represents the set of neighboring nodes of node k; deg(k) represents the degree of node k; deg(l) represents the degree of node l; W represents the weight matrix; b represents the bias vector;

[0091] (3) Loss function:

[0092] Select the function BCEWithLogitsLoss(·) as the loss function:

[0093]

[0094] In the formula, logits represents the output of the model without the activation function, which is the score belonging to the positive category; labels represents the true binary classification label, which is 0 or 1.

[0095] Case Analysis

[0096] The following example illustrates the superiority of the method for constructing the flexible operation domain of the electric-gas-hydrogen hybrid integrated energy system based on graph convolutional neural network described in the present invention. Figure 2The 118-node power system and the 25-node natural gas system shown. In order to compare the superiority of the method proposed in the present invention, the traditional electricity-gas-hydrogen integrated energy system flexibility operation domain construction scheme and the method for constructing the electricity-gas-hydrogen integrated energy system flexibility operation domain based on deep neural network and the electricity-gas-hydrogen hybrid integrated energy system flexibility operation domain construction method based on graph convolutional neural network proposed in the present invention are used to compare the performance and construction results. The present invention solves the nonlinear programming problem of the basic scheme through the GAMS optimization platform with the BARON solver, and uses Python 3.11 and Pytorch to process deep neural network and graph convolutional neural network problems.

[0097] Based on this example, the recognition performance of the traditional basic scheme for constructing the flexible operation domain of the electricity-gas-hydrogen integrated energy system, the method for constructing the flexible operation domain of the electricity-gas-hydrogen integrated energy system based on a deep neural network, and the method for constructing the flexible operation domain of the electricity-gas-hydrogen hybrid integrated energy system based on a graph convolutional neural network proposed in the present invention are compared (the results are shown in Table 1), and the flexible operation domain results constructed by the method for constructing the flexible operation domain of the electricity-gas-hydrogen integrated energy system based on a deep neural network and the method for constructing the flexible operation domain of the electricity-gas-hydrogen hybrid integrated energy system based on a graph convolutional neural network proposed in the present invention are compared with the actual results (the results are shown in Table 2). Figure 3 , Figure 4 ). This shows that the graph convolutional neural network has better performance and results in identifying the feasibility of the power point of hydrogen production and building the flexible operation domain than the deep neural network in the construction of the system flexible operation domain by mining the characteristics of system topology, operation constraints, scenario data and other conditions. Compared with the construction of the flexible operation domain based on the deep neural network, the construction of the flexible operation domain by the graph convolutional neural network improves the recognition accuracy by adding the characteristics of the network topology (Table 1 and Figure 3 , Figure 4), and the training time of the method based on graph convolutional neural network for constructing the system flexibility operation domain is lower than that of the method based on deep neural network, and the recognition time of the recognition method based on graph convolutional neural network is only 1 / 3630 of the recognition time of the basic scheme for constructing the flexibility operation domain of the traditional electricity-gas-hydrogen integrated energy system (the flexibility operation domain in this example is constructed with an accuracy of 10MW, and the power upper limits of the three hydrogen production are 100MW, 70MW, and 100MW respectively. Therefore, the feasibility of 968 hydrogen production power points needs to be identified in space. The basic scheme can only construct the system's flexibility operation domain by identifying whether each hydrogen production power point is feasible in turn. The basic scheme takes about 30 seconds to identify each hydrogen production power point). It shows that the method for constructing the flexibility operation domain of the electricity-gas-hydrogen hybrid integrated energy system based on graph convolutional neural network has greatly reduced the time consumed for construction on the basis of having a very high accuracy.

[0098] Table 1 Comparison of recognition performance and time consumption of different methods

[0099]

[0100] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention.

Claims

1. A method for constructing a flexible operation domain of an electricity-gas-hydrogen hybrid integrated energy system, characterized in that: The following steps are involved: Step 1, obtaining network coefficients and operation coefficients of the power grid, gas grid, and electric hydrogen production model, wherein the network coefficients include power grid topology, gas grid topology, line resistance and impedance, natural gas, and hydrogen calorific value, and the operation coefficients include generator set power generation coefficients, wind power photovoltaic inverter coefficients, gas source parameters, electric hydrogen production device parameters, and booster station parameters; Step 2: Obtain the scenario data of power grid load demand, gas grid load demand, wind power and photovoltaic output, and power operation data set of hydrogen production device at different time sections; Step 3: Taking the minimum load shedding of the electricity-gas-hydrogen hybrid integrated energy system as the objective function, a basic physical model for identifying whether the power point of hydrogen production is a feasible point is established. Based on the obtained network coefficients, operating coefficients and scenario data, a nonlinear optimization solver is used to solve the basic physical model to identify feasible points and infeasible points under corresponding operating conditions, and a neural network training sample data set is generated under different gas network topologies with scenario data as the feature and corresponding feasible point or infeasible point recognition results as the label; Step 4: For the neural network training sample data set generated in step 3, train and establish a model for constructing the flexible operation domain of the electric-gas-hydrogen hybrid integrated energy system based on a graph convolutional neural network, which takes the scene data at different time sections as input and the feasible points or infeasible points identified in batches as output; Step 5: For any gas grid topology involved in the training, the electric hydrogen production power point is uniformly given with a certain accuracy within the power constraint range of the electric hydrogen production device. Based on the graph convolutional neural network established in step 4, the model of the flexible operation domain of the electric-gas-hydrogen hybrid integrated energy system is constructed. The load demand data, wind power and photovoltaic output data and all uniformly given electric hydrogen production power point data at any time section are batch input, all feasible points are identified, and a convex envelope is constructed to form the system's flexible operation domain under the gas grid topology, time section and a certain accuracy; In step 3, the basic physical model for identifying whether the power point of hydrogen production is a feasible point with the minimum load shedding of the electricity-gas-hydrogen hybrid integrated energy system as the objective function is as follows: (1) Objective function In the formula, It represents the required power of the electric load d; It represents the actual absorbed power of the electric load d; It represents the energy demand of gas load e; It represents the actual energy consumed by the gas load e; The objective function of formula (A-1) represents the minimum sum of the power cut load and the gas cut load; (2) Power system operation model i REF =0 (A-8) In the formula, subscripts i and j represent power buses; subscript v represents generators; subscript r represents new energy units; subscript d represents power loads; subscript h represents hydrogen production by electricity; subscript REF represents reference buses; U v (i) represents the set of generator sets connected to bus i; U r (i) represents the set of new energy units connected to bus i; U d (i) represents the set of power loads connected to bus i; U h (i) represents the set of electric hydrogen production connected to bus i; U(i) represents the set of lines connected to bus i; P v G represents the actual power generated by the generator v; P r W Indicates the grid-connected power of the new energy unit r; represents the power of hydrogen production from electricity; θ i represents the voltage phase angle of bus i; x ij and They represent the susceptance and transmission capacity of line ij respectively; and Respectively represent the minimum and maximum power generation of unit v; Indicates the predicted power of new energy units; They represent the upper and lower limits of the electric power consumed by electric hydrogen production h; θ REF represents the phase angle of the reference bus; Formula (A-2) is the node active power balance equation; Formula (A-3) is the active load demand constraint; Formula (A-4) is the line transmission capacity constraint; Formula (A-5) is the upper and lower limits of the generator active output; Formula (A-6) is the grid-connected new energy output constraint; Formula (A-7) is the upper and lower limits of the power of hydrogen production; Formula (A-8) stipulates that the phase angle of the reference bus is 0; (3) Hydrogen production model The electric hydrogen production device is an important component of the coupling between the power grid and the gas grid. In the power system, the electric hydrogen production device is equivalent to a controllable load, and in the natural gas system, the electric hydrogen production device is equivalent to a gas source. Its model is as follows: In the formula, It represents the hydrogen energy produced by electric hydrogen production h; It represents the hydrogen flow rate produced by electric hydrogen production h; is the maximum hydrogen energy that can be produced by hydrogen production from electricity; It represents the calorific value of hydrogen, which is 12.75MJ / m 3 ; η h It represents the energy conversion efficiency of electric hydrogen production; Formula (A-9) represents the relationship between the electric energy consumed by electric hydrogen production and the energy of hydrogen produced; Formula (A-10) represents the relationship between the energy of hydrogen produced and the flow rate; Formula (A-11) describes the upper and lower limit constraints of the hydrogen energy produced by electric hydrogen production; (4) Steady-state natural gas system operation model considering pipeline hydrogen blending In the formula, subscripts m and n represent natural gas nodes; subscript e represents gas load; subscript k represents natural gas compression station; subscript w represents natural gas source; subscript mn represents pipeline mn between natural gas nodes m and n; G w (m) represents the gas source set connected to node m; G h (m) represents the electric hydrogen production set connected to node m; G e (m) represents the gas load set connected to node m; G g (m) represents the set of gas generators connected to node m; G k (m) represents the set of booster stations connected to node m; G(m) represents the set of pipelines connected to node m; It represents the energy provided by the gas source w; It represents the natural gas energy consumed by gas generator set v; E mn With E nm They represent the energy of the natural gas at the beginning and end of the pipeline mn respectively; It represents the energy possessed by the natural gas flowing through the compression station k; represents the percentage of natural gas consumption at the booster station k to the transmission flow; F mn With F nm They represent the natural gas flow at the beginning and end of the pipeline mn respectively; W mn represents the Weymouth constant of the pipeline mn; π m and π n It represents the pressure of nodes m and n; Represents the calorific value of the pipeline mn; H gas It represents the calorific value of natural gas, which is 38.29MJ / m 3 ; H m It represents the heat value of node m; It represents the natural gas flow provided by the gas source w; It represents the natural gas flow rate flowing through the compression station k; η v It represents the power generation efficiency of the gas unit; and They represent the maximum and minimum pressure values ​​of node m respectively; It represents the upper limit of the flow rate at the head end of the pipeline; and They represent the maximum and minimum supply energies of the gas source w respectively; and They represent the inlet and outlet pressures at the booster station k respectively; and They represent the maximum and minimum pressurization ratios of the pressurization station k respectively; represents the maximum transmission energy of the booster station k; α represents the hydrogen mixing ratio; Formula (A-12) is the energy balance constraint equation; Formula (A-13) is an equation that describes the nonlinear relationship between the pipeline flow rate and the pressure drop at the beginning and end of the pipeline; Formula (A-14) is a necessary constraint condition for the steady-state system, that is, without considering the pipe storage, it is assumed that the flow rate at the beginning of the pipeline is equal to the flow rate at the end; Formula (A-15) is the formula for calculating the calorific value: the energy flowing into the node is equal to the energy flowing out of the node plus the energy consumption of the booster station; Formula (A-16) is the energy calculation formula; Formula (A-17) is the gas load demand energy constraint; Formula (A-18) is the node pressure constraint; Formula (A-19) is the constraint on the flow rate at the beginning and end of the pipeline; Formula (A-20) is the gas source supply constraint; Formula (A-21) is the maximum and minimum boost ratio constraint of the booster station; Formula (A-22) is the booster station transmission energy constraint; Formula (A-23) is the pipeline calorific value constraint; Formula (A-24) is the node calorific value constraint; Based on the above basic physical model for identifying whether the hydrogen production power point is a feasible point, the hydrogen production power point under the condition that all constraints of the basic physical model can be satisfied without load shedding is called a feasible point under the corresponding conditions; The power point of hydrogen production under the condition that all constraints of the basic physical model can only be satisfied by shedding load is called an infeasible point under the corresponding condition, and the power set space of all feasible points under the same operating conditions constitutes the flexible operating domain of the system under this condition; In step 4, the model of the flexible operation domain of the electricity-gas-hydrogen hybrid integrated energy system is constructed based on the graph convolutional neural network: (1) Network structure model The constructed graph convolutional neural network structure consists of three graph convolutional layers plus one linear layer, and its mathematical expression is as follows: In the formula, x represents the input initial feature matrix composed of all node features; is the symmetric normalized adjacency matrix obtained by normalizing the adjacency matrix A; ReLU(·) is the rectified linear unit activation function; the superscripts (1), (2), (3), and (4) denote the 1st, 2nd, 3rd, and 4th layers; z (1) ,z (2) ,z (3) They are the output feature matrix representations of the 1st, 2nd, and 3rd layer graph convolution operations respectively; The feature representation of node k that represents the output of the 3rd layer graph convolution operation; represents the global feature; k represents the node number; N is the number of nodes; W (1) ,W (2) ,W (3) are the parameter weight matrices of the 1st, 2nd, and 3rd graph convolutional layers respectively; W (4) represents the parameter weight matrix of the 4th linear layer; b (1) ,b (2) ,b (3) are the bias vectors of the 1st, 2nd, and 3rd graph convolutional layers respectively; b (4) represents the bias vector of the 4th linear layer; y represents the final output recognition result, that is, the classification result of the graph; In the forward propagation process of the network structure, the node features pass through three layers of graph convolutional layers. Each layer uses the ReLU activation function to map the features to a higher dimension. The multi-dimensional feature vectors after propagation through the three layers of convolutional layers are summarized by global average pooling to integrate the node features in the graph into an overall feature, which is mapped to the output space through linear transformation to obtain the final recognition result. The first layer is the input layer, and the input dimension corresponds to the number of features of each node; the second layer input dimension is equal to the first layer output dimension; the third layer input dimension is equal to the second layer output dimension; the fourth layer input dimension is equal to the third layer output dimension, and the output dimension is 1, and the output is the probability of being recognized as a positive category; (2) Graph convolution operation model The general mathematical model of the graph convolution operation of each node in the network structure is as follows: In the formula, l represents the node number; x l represents the feature representation of node l; z k Represents the new feature representation of node k; neighbors(k) represents the set of neighbor nodes of node k; deg(k) represents the degree of node k; deg(l) represents the degree of node l; W represents the weight matrix; b represents the bias vector; (3) Loss function: The function BCEWithLogitsLoss(·) is selected as the loss function as follows: In the formula, logits represents the output of the model without the activation function, which is the score belonging to the positive category; labels represents the true binary classification label, which is 0 or 1.

Images