Electric heating integrated energy system optimization planning method considering seasonal heat storage, equipment and medium

Abstract

The invention relates to an electric heating integrated energy system optimization planning method and device considering seasonal heat storage and a medium, and the method comprises the following steps: constructing a seasonal heat storage model based on a heat charging and discharging state and an internal thermal power state, the internal thermal power state comprising a weekly state and an intra-period state, and being obtained through the prediction of a deep neural network; based on the seasonal heat storage model, an electric heating integrated energy system coordinated planning model is built, and the electric heating integrated energy system coordinated planning model takes the sum of minimum annual investment cost and operation cost as a target function and builds constraint conditions based on investment constraints and operation constraints; and carrying out hybrid linear transformation on the electric heating integrated energy system coordinated planning model by adopting a large m method, and solving to obtain an optimal planning scheme of the electric heating integrated energy system. Compared with the prior art, the method has the advantages that the electric heating comprehensive energy system can be accurately modeled, the nonlinear optimization problem of a traditional physical modeling method is solved, and scientific and efficient energy planning is achieved.

Description

Technical Field

[0001] The present invention belongs to the field of integrated energy system planning, and in particular relates to an optimization planning method, device and medium for an electric-thermal integrated energy system taking seasonal heat storage into consideration. Background Art

[0002] With the large-scale application of renewable energy, the integrated electric and thermal energy system (IEHS) has become an important development direction. The power system and the thermal system are connected together through components such as cogeneration units and heat pumps to realize the conversion of electric energy and thermal energy, which can reasonably optimize the production and conversion process of electric energy and thermal energy. The seasonal thermal energy storage system (STES) is installed on the node together with the cogeneration unit and heat pump. It can store excess heat energy in summer and release it in winter to meet heat demand, thereby improving energy utilization efficiency. Traditional seasonal heat storage models mostly use physical modeling methods, which have problems such as complex models, large amount of calculations, and difficulty in handling nonlinear dynamic changes. This makes the capacity planning of STES very challenging and difficult to accurately describe its complex behavior.

[0003] The paper "Optimal Planning of Integrated Electricity and Heat System Considering Seasonal and Short-Term Thermal Energy Storage" (J. Tan, Q. Wu and X. Zhang, IEEE Transactions on Smart Grid, vol. 14, no. 4, pp. 2697-2708) proposed a three-stage planning model for an integrated electric and heat energy system (IEHS) considering seasonal thermal energy storage (STES) and short-term energy storage. The above method takes into account the different energy cycle characteristics of STES and short-term energy storage, and coordinates to solve multi-scale uncertainties, but there are problems of insufficient accuracy and low computational efficiency. Therefore, it is necessary to redesign an optimal planning method for an integrated electric and heat energy system considering seasonal heat storage, which can accurately model the integrated electric and heat energy system, solve the nonlinear optimization problems of traditional physical modeling methods, and achieve scientific and efficient energy planning. Summary of the invention

[0004] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide an optimization planning method, equipment and medium for an electric-thermal integrated energy system taking into account seasonal heat storage, to accurately model the electric-thermal integrated energy system, and to solve the nonlinear optimization problems of traditional physical modeling methods, so as to achieve scientific and efficient energy planning.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] The present invention provides an optimization planning method for an electric-thermal integrated energy system considering seasonal heat storage, comprising the following steps:

[0007] A seasonal heat storage model is constructed based on the heat charging and discharging state and the internal thermodynamic state. The heat charging and discharging state includes the heat charging state and the heat discharging state, both of which are binary variables. The internal thermodynamic state includes the inter-cycle state and the intra-cycle state, which respectively represent the initial temperature of the first hour of each day and the temperature change of each hour of each day, both of which are nonlinear variables. The inter-cycle state and the intra-cycle state are respectively obtained by deep neural network prediction;

[0008] Based on the seasonal heat storage model, a coordinated planning model for an electric-thermal integrated energy system is constructed. The coordinated planning model for the electric-thermal integrated energy system takes minimizing the sum of annual investment cost and operating cost as an objective function, and constructs constraint conditions based on investment constraints and operating constraints. The operating cost includes maintenance cost, fuel and load reduction cost, electricity cost and carbon emission cost. The operating constraints include cogeneration units, heat pumps and thermal load operating constraints, distributed heating network operating constraints and power system operating constraints.

[0009] The big-m method is used to perform mixed linear transformation on the coordinated planning model of the electric-thermal integrated energy system to obtain the optimal planning scheme of the electric-thermal integrated energy system.

[0010] Furthermore, the expression of the seasonal heat storage model is as follows:

[0011]

[0012] Among them, c w and ρ w are the specific heat capacity and density of water, R and h are the radius and height of the seasonal heat storage model, is the internal thermodynamic state of the seasonal heat storage model at time t, is the mass flow rate of the seasonal heat storage model at time t, and are the supply end temperature and return end temperature of the seasonal heat storage model at time t, and are the heat losses at the top, bottom and sides of the seasonal heat storage model at time t, and d represents the day.

[0013] Furthermore, the internal thermodynamic state of the seasonal heat storage model at time t is The specific expression is as follows:

[0014]

[0015] in, For the periodic state, is the state within the cycle, η is the heat loss coefficient;

[0016]

[0017] Among them, Φ RP To represent the day collection, and They are respectively the charging state and the releasing state. is the soil ambient temperature at time t.

[0018] Furthermore, the expression of the heat loss coefficient η is as follows:

[0019]

[0020] Among them, s t 、s b and s are the thickness of the top, bottom and sides, respectively, and λ t,wall , b,wall and λ s,wall are the thermal conductivities of the top, bottom and side surfaces, respectively, is the basic mass flow rate.

[0021] Furthermore, the deep neural network is a fully connected neural network with a ReLU activation function, and its network parameters are obtained by iterative training by minimizing a loss function, and the loss function includes a mean square error and a Huber loss function.

[0022] Furthermore, the expression of the objective function of the electric-thermal integrated energy system coordination planning model is as follows:

[0023]

[0024]

[0025] C m =αC inv

[0026]

[0027] Among them, C inv is the investment cost, C m is the maintenance cost, C F&LS is the fuel and load shedding cost, C grid For electricity cost, is the carbon emission cost, r is the interest rate, β is the component life, c CHP 、c HP 、c STES and c WTare the unit investment costs of the CHP unit, heat pump, seasonal heat storage model and fan, respectively. i CHP,cap , P i HP,cap and P i WT,cap are the capacities of the cogeneration unit, heat pump and fan, V i STES is the physical volume of the seasonal heat storage model, α is the maintenance cost coefficient, W d is the length of the day, c fuel and c LS are the unit fuel cost and load reduction penalty cost of the CHP unit, and are the power generation and load reduction of the cogeneration unit, and The power purchase price and power sales price of the power grid are respectively and are the electricity purchased and sold from the power grid, c carbon is the unit carbon emission cost, a CHP and a grid are the carbon emission coefficients of the cogeneration unit and grid power generation, Φ CHP , Φ HP , Φ STES , Φ WT and Φ EL They are combined heat and power unit kit, heat pump kit, seasonal thermal storage model kit, fan kit and electrical load kit.

[0028] Furthermore, the constraint conditions of the electric-thermal integrated energy system coordination planning model are specifically expressed as follows:

[0029] Investment constraints:

[0030]

[0031] Where R and h are the radius and height of the seasonal heat storage model, respectively;

[0032] Cogeneration units, heat pumps and heat load operating constraints:

[0033]

[0034] in, and are the calorific value of the cogeneration unit and the heat pump, and are the power heat transfer coefficients of the cogeneration unit and the heat pump, and are the power generation of the cogeneration unit and the power consumption of the heat pump, respectively; c is the specific heat capacity of water, and are the mass flow rates of the CHP unit, heat pump and heat load, respectively, and are the temperatures of the cogeneration unit, heat pump and heat load supply end, and are the temperatures of the cogeneration unit, heat pump and heat load return end, respectively. and They are heat charging state and heat releasing state respectively;

[0035] Distributed heating network operation constraints:

[0036]

[0037] in, is the mass flow rate in the pipeline, and are the mass flow rate temperature of the heat supply network outlet and the heat recovery network, and are the mass flow rate temperature at the outlet of the heating network and the inlet of the heat recovery network, respectively. and are the temperatures of the nodes ND of the heating network and the heat recovery network, is the basic mass flow rate;

[0038] Power system operation constraints:

[0039]

[0040]

[0041] in, For large-scale wind power generation, and are the total input flow and total output flow of the corresponding date, time and node respectively. is the power input corresponding to the date, time and node, is the power load corresponding to the date, time and node, V j,t,d is the voltage of node j at the corresponding date, time, V i,t,d is the voltage of node i at the corresponding date, time, and r ij and x ij are the resistance parameters and reactance parameters of nodes i to j, respectively, ij,t,d and Q ij,t,d are the real-time power of nodes i to j at date d, time t, and V ref is the reference voltage, ΔV max is the maximum voltage deviation.

[0042] Furthermore, the big-m method is used to linearize the bilinear terms in the constraints of the coordinated planning model of the electric-thermal integrated energy system, and the problem of solving the coordinated planning model of the electric-thermal integrated energy system is converted into a mixed integer linear programming MILP problem and solved.

[0043] The present invention also provides an electronic device, comprising a memory, a processor, and a program stored in the memory, wherein the processor implements the above method when executing the program.

[0044] The present invention also provides a computer-readable storage medium on which a computer program is stored, and the program implements the above method when executed by a processor.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] The present invention provides an optimization planning method for an electric-thermal integrated energy system considering seasonal heat storage, and constructs a seasonal heat storage model based on a heat charging and discharging state and an internal thermodynamic state, wherein the heat charging and discharging state includes a heat charging state and a heat discharging state, both of which are binary variables, so that the model is more direct and accurate in describing the energy flow of an energy storage system, simplifies the complex heat charging and discharging process, and is convenient for subsequent calculations and analyses; the internal thermodynamic state includes an inter-cycle state and an intra-cycle state, which respectively represent the initial temperature of the first hour of each day and the temperature change every hour of each day, so that the model can more accurately capture the dynamic change of the internal temperature of the energy storage system, thereby improving the simulation accuracy of the model for actual physical processes, and can also capture temperature trends on a longer time scale, providing strong support for long-term planning and management; the inter-cycle state and the intra-cycle state are both nonlinear variables, which are respectively obtained through deep neural network prediction, can efficiently process high-dimensional and large-volume data, and improve the accuracy and flexibility of the seasonal heat storage model. The present invention constructs a coordinated planning model for an electric-thermal integrated energy system based on a seasonal heat storage model, takes minimizing the sum of annual investment costs and operating costs as the objective function, and constructs constraint conditions based on investment constraints and operating constraints, wherein operating costs include maintenance costs, fuel and load reduction costs, electricity costs, and carbon emission costs; operating constraints include cogeneration units, heat pumps, and thermal load operating constraints, distributed heating network operating constraints, and power system operating constraints; the large-m method is used to perform a mixed linear transformation on the coordinated planning model for the electric-thermal integrated energy system, and the optimal planning scheme for the electric-thermal integrated energy system is obtained by solving the method. In the above process, multiple factors are fully considered to make the model closer to the actual application scenario, solve the nonlinear optimization problem of traditional physical modeling methods, simplify the model solving process, and achieve efficient and scientific energy planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is a schematic diagram of the internal thermodynamic state.

[0048] Among them, annual hours is the number of hours in a year (a total of 8760 hours in a year), annual input data is the data for the whole year, which can be simplified to the representative day D, Ψ is the typical day, that is, the day with similar STES output in the whole year, consecutive days are consecutive days, inter-period and intra-period are the inter-period stage and intra-period stage respectively, Internal Temp is the internal temperature of STES, is the internal thermodynamic state, For the periodic state, It is the state within the cycle;

[0049] Figure 2 This is a schematic diagram of the structure of the integrated electric and thermal energy system (IEHS) including the seasonal thermal energy storage model (STES).

[0050] Among them, CHP is combined heat and power unit, HP is heat pump, WT is fan, GB is gas boiler, Upper power is upstream power grid, gas system is gas system, and End-users is end users;

[0051] Figure 3 Schematic diagram of the district heating network (DHN).

[0052] Among them, Supply pipelines represents water supply pipelines, which are represented by red, Return pipelines represent return pipelines, which are represented by blue, HL (Heat load) represents heat load, Heat station represents heat station, and nd represents node;

[0053] Figure 4 This is a schematic diagram of neural network input and output.

[0054] Among them, (4a) represents The corresponding neural network, (4b) represents The corresponding neural network, (4c) represents The corresponding neural network, (4d) represents The corresponding neural network;

[0055] Figure 5 It is the IEHS topology with STES test;

[0056] Figure 6 The loss function results on the training data set of four neural networks

[0057] Among them, (6a) represents The corresponding neural network, (6b) represents The corresponding neural network, (6c) represents The corresponding neural network, (6d) represents The corresponding neural network;

[0058] Figure 7 The temperature comparison results of neural network-assisted STES and physical STES are shown in Figure 2.

[0059] Among them, (7a) represents the variable (7b) represents a variable (7c) represents a variable (7d) represents a variable

[0060] Figure 8 is the optimal power and heat dispatch for Case 1 and Case 4,

[0061] Among them, (8a) represents the power operation result under case 1, (8b) represents the thermal operation result under case 1, (8c) represents the power operation result under case 4, and (8d) represents the thermal operation result under case 4. DETAILED DESCRIPTION

[0062] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0063] Example:

[0064] like Figure 2 As shown in the figure, in the integrated electric and thermal energy system (IEHS), the seasonal thermal storage model (STES) is installed at the node together with the combined heat and power unit and the heat pump. The combined heat and power unit is responsible for balancing the variable mass flow rate according to the charging and discharging state of the STES to ensure the constant mass flow rate of the heat source node. The heat pump converts electrical energy into thermal energy, stores it in the STES, and then releases it during the heating season. The IEHS can sell the surplus electricity to the grid when the wind power output is large, and purchase electricity from the grid during peak hours.

[0065] This embodiment provides an optimization planning method for an electric-thermal integrated energy system considering seasonal heat storage. By accurately modeling and optimizing STES, the operating efficiency and economy of the system are improved to meet the thermal energy needs of different seasons. The linearization processing of neural networks is performed to solve the nonlinear optimization problems of traditional physical modeling methods, improve the operating efficiency and flexibility of the system, and realize low-carbon and efficient energy optimization planning. Specifically, the following steps are included:

[0066] S1. A seasonal heat storage model STES is constructed based on the charging and discharging state and the internal thermodynamic state. The charging and discharging state includes the charging state and the discharging state, both of which are binary variables. The internal thermodynamic state includes the inter-cycle state and the intra-cycle state, which respectively represent the initial temperature in the first hour of each day and the temperature change every hour of each day, both of which are nonlinear variables.

[0067] Constructing seasonal heat storage model STES:

[0068]

[0069] Among them, Q STES is the thermal energy of STES, c w and ρ w are the specific heat capacity and density of water, R and h are the radius and height of the seasonal heat storage model, τ in is the internal temperature of STES, is the sum of the inlet and outlet enthalpy flows of the STES, Q loss is the total heat loss of STES from top, bottom and sides to its surroundings.

[0070] By introducing a binary variable representing the STES charging and discharging state and The nonlinear differential thermodynamics in equation (1) can be expressed as a time step of 1 hour, as follows:

[0071]

[0072]

[0073] in, is the internal temperature of STES at time t (i.e., the internal thermodynamic state), is the mass flow rate of the seasonal heat storage model at time t, and are the supply end temperature and return end temperature of the seasonal heat storage model at time t, and are the heat losses at the top, bottom and sides of the seasonal heat storage model at time t, and d represents the day.

[0074] Assuming that the mass flow rate flowing into the STES in the charging state is the same as the mass flow rate flowing out of the STES in the heat releasing state, the mass flow rate of the STES is expressed as formula (3):

[0075]

[0076] Since the charging and discharging cycles of STES are less than those of short-term heat storage, STES keeps running all day long, and equation (4) is used to ensure that the charging and discharging states do not occur at the same time:

[0077]

[0078] Equations (5) and (6) represent the heat loss from the top, bottom and sides, which mainly depend on the surface area and the temperature difference between the internal storage medium and the surrounding environment:

[0079]

[0080] Among them, λ t / b,wall represents the thermal conductivity of the top and bottom of the STES at time t, λ s,wall represents the thermal conductivity of the side surface of the STES at time t, represents the soil ambient temperature at time t, s t / b and s Represents the thickness of the top, bottom and sides respectively.

[0081] The hourly internal thermal dynamic model of STES is as follows:

[0082]

[0083] in, represents the basic mass flow rate of STES, represents the return end temperature of STES at time t.

[0084] The relevant constraint expressions of the STES hourly internal thermodynamic model are:

[0085]

[0086]

[0087] in, and are the maximum and minimum temperatures of STES, and are the minimum return end temperature and maximum return end temperature of STES, respectively. In the state of heat charging and heat discharging, the return and supply end temperatures of STES are equal to the internal temperature of STES, respectively, as expressed by equations (8) and (9); equation (10) is used to ensure that the internal temperature of STES is not lower than its return end temperature and not higher than its supply end temperature; equation (11) is used to ensure that the initial heat energy of STES is consistent with the heat stored at the end of the cycle; equation (12) is used to limit the maximum heat energy stored in STES.

[0088] like Figure 1 As shown in the figure, the data of 8760 hours throughout the year are simplified into 365 representative days D, each of which belongs to a different typical day Ψ. In this way, the STES temperature (i.e., the state between cycles) can be ) is decomposed into two sub-states for representation:

[0089]

[0090] Among them, Φ RP To represent the day collection, is the periodic state, indicating the initial temperature of the first hour of each day, expressed by formula (14); is the intra-cycle state, indicating the temperature change every hour of every day, which is zero in the first hour of every day, as expressed by formula (15).

[0091] The heat loss coefficient η is:

[0092]

[0093] Through the heat loss coefficient η, the simplified time-sequential continuous STES model can be obtained as follows:

[0094]

[0095]

[0096] in, is the periodic state of the representative day, expressed by formula (17), The status of the cycle is the last hour of a typical day; is the state of the STES cycle for two consecutive hours, expressed by equation (18). According to equations (17) and (18), equation (7) is re-expressed as a simplified STES model that is continuous in time order, expressed by equation (19). Equation (7) is the calculation method of STES itself. If it is not simplified by the typical day method, then it is necessary to calculate the temperature change for 8760 hours throughout the year, which makes the calculation complicated and difficult. The principle of the typical day method is: calculate the temperature change for 24 hours every day for each typical day, calculate the temperature at the first moment of each day for 365 days throughout the year, and then express the two as the final STES temperature.

[0097] S2. Based on the seasonal heat storage model STES, the coordinated planning model IEHS of the electric-thermal integrated energy system is constructed. The coordinated planning model IEHS of the electric-thermal integrated energy system is used to minimize the annual investment cost C. inv The sum of the cost of maintenance and the cost of operation is the objective function, and the constraint conditions are constructed based on the investment constraint and the operation constraint. The operation cost includes the maintenance cost C m , fuel and load shedding costs C F &LS , electricity cost C grid and the cost of carbon emissions Operation constraints include cogeneration units, heat pumps and heat load operation constraints, distributed heating network operation constraints, and power system operation constraints.

[0098] The expression of the objective function is as follows:

[0099]

[0100] C m =αC inv (twenty two)

[0101]

[0102] Where r is the interest rate, β is the component life, c CHP 、c HP 、c STES and c WT are the unit investment costs of the CHP unit, heat pump, seasonal heat storage model and fan, respectively. i CHP,cap , P i HP,cap and P i WT,cap are the capacities of the cogeneration unit, heat pump and fan, V i STES is the physical volume of the seasonal heat storage model, α is the maintenance cost coefficient, W d is the length of the day, c fuel and c LS are the unit fuel cost and load reduction penalty cost of the CHP unit, and are the power generation and load reduction of the cogeneration unit, and The power purchase price and power sales price of the power grid are respectively and are the electricity purchased and sold from the power grid, c carbon is the unit carbon emission cost, a CHP and a grid are the carbon emission coefficients of the cogeneration unit and grid power generation, Φ CHP , Φ HP , Φ STES , Φ WT and Φ EL They are combined heat and power unit kit, heat pump kit, seasonal thermal storage model kit, fan kit and electrical load kit.

[0103] The expression of the constraint conditions of the coordinated planning model of the electric and thermal integrated energy system is as follows:

[0104] (a) Investment constraints:

[0105]

[0106] Among them, formula (26) indicates that the capacity of the cogeneration unit, heat pump and fan unit is limited by the maximum installed capacity, and formula (27) limits the maximum installation volume of STES, which corresponds to the optimized variable R.

[0107] (b) Cogeneration units, heat pumps and heat load operating constraints:

[0108]

[0109] in, and are the calorific value of the cogeneration unit and the heat pump, and are the power heat transfer coefficients of the cogeneration unit and the heat pump, and are the power generation of the cogeneration unit and the power consumption of the heat pump, respectively; c is the specific heat capacity of water, and are the mass flow rates of the CHP unit, heat pump and heat load, respectively, and are the temperatures of the cogeneration unit, heat pump and heat load supply end, and They are the temperatures of the cogeneration unit, heat pump and heat load return end respectively.

[0110] The heat output of the CHP unit and the heat pump is linearly related to the power output and consumption, as shown in equations (28) and (29), respectively. Equations (30) and (31) implement the power generation and power consumption limits of the CHP unit and the heat pump. The heat output of the CHP unit and the heat pump is calculated by equations (32) and (33), and the heat consumption of the heat load is calculated by equation (34). In order to balance the mass flow rate changes from the STES, the CHP unit adjusts its mass flow rate according to the charging and discharging state of the STES to ensure that the total mass flow rate of the heat source node is constant, as expressed by equation (35). Equations (36), (37), and (38) limit the supply and return temperatures of the CHP unit, the heat pump, and the heat load, respectively. The mass flow rates flowing into and out of the STES are considered to be constant.

[0111] (c) Distributed heating network (DHN) operation constraints:

[0112]

[0113] in, is the mass flow rate in the pipeline, and are the mass flow rate temperature of the heat supply network outlet and the heat recovery network, and are the mass flow rate temperature at the outlet of the heating network and the inlet of the heat recovery network, respectively. and are the temperatures of the nodes ND of the heating network and the heat recovery network respectively. Figure 3 As shown in Figure 2, in the distributed heating network (DHN), the temperature of each heat node is calculated as the mixed temperature of the mass flow rate flowing into the node. Equations (39) and (40) give the energy conservation law at each heat node in the supply and return pipes. The STES acts as a heat source when releasing heat and as a heat load when charging heat. Equations (41) and (42) ensure that the supply and return temperature of any cogeneration unit, heat pump, heat load or pipe flowing out of a node is equal to the mixed temperature of the node. The heat node supply and return temperature is constrained by equation (43).

[0114] (d) Power system (EPS) operation constraints:

[0115]

[0116] in, For large-scale wind power generation, and are the total input flow and total output flow of the corresponding date, time and node respectively. is the power input corresponding to the date, time and node, is the power load corresponding to the date, time point, and node, V j,t,d is the voltage of node j at the corresponding date, time, V i,t,d is the voltage of node i at the corresponding date, time, and r ij and x ij are the resistance parameters and reactance parameters of nodes i to j, respectively, ij,t,d and Q ij,t,d are the real-time power of nodes i to j at date d, time t, and V ref is the reference voltage, ΔV max is the maximum voltage deviation.

[0117] The power generation of the cogeneration unit and the consumption of the heat pump are constrained by equation (44), equation (45) is the limit of available wind power generation, equation (46) constrains the maximum amount of electricity traded with the grid, and equation (47) indicates that the load reduction is less than the power load. The active and reactive power flow constraints do not consider the losses on each bus, and are expressed by equations (48) and (49) respectively. Equation (50) calculates the voltage loss of each branch, and equation (51) limits the voltage of each bus to Vref = 1p.u.

[0118] S3. The inter-cycle state and the intra-cycle state are obtained through deep neural network prediction, and the large-m method is used to perform mixed linear transformation on the coordinated planning model IEHS of the electric-thermal integrated energy system to obtain the optimal planning scheme of the electric-thermal integrated energy system.

[0119] The deep neural network is a fully connected neural network with ReLU activation function, including an input layer, M hidden layers and an output layer. The input layer has N input features and the hidden layer contains N M neurons, and the output layer consists of one neuron, represented by Y:

[0120]

[0121] Each hidden layer uses ReLU as the activation function, and the output layer uses a linear activation function. The output vectors of each hidden layer and output layer are shown in equations (53) and (54):

[0122]

[0123] The nonlinear ReLU can be transformed into a set of MILP constraints through the big m method, introducing a binary variable and And H is large enough, then the MILP formula of ReLU is expressed as:

[0124]

[0125] The DHN coupled STES model with simplified time representation established by equations (16) and (19) is nonlinear and computationally complex. The nonlinear thermal dynamic state is encoded into and Four neural networks, respectively Figure 4 As shown in (4a), (4b), (4c) and (4d), each neural network data set consists of a set of input variables X and a set of output variables τ. As can be seen from formula (18), the state within the cycle You can select and As the input variable, since the η of the state in the initial period is 0, Depend on and In addition, for the periodic state in equation (17) and is selected as the input variable. Since in formula (19), yes and The superposition of variables and As an input variable, since the optimization problem involves the capacity configuration of STES, all four DNNs take R as an input variable.

[0126] Four neural networks are trained using a data set consisting of input variables and output variables to determine the parameters of the neural network. The training process includes forward propagation and backward propagation, calculating the loss function and updating the values ​​of weights w and bias b respectively. The neural network uses the mean square error (MSE) as the loss function, as shown in formula (58), and for and The neural network uses the Huber loss function, as shown in formula (59). MSE is the mean of the sum of squares of the differences between the predicted value and the actual value. It gives a greater penalty to large errors and is therefore very sensitive to outliers. Huber loss is a combination of MSE and absolute error. It introduces a threshold to determine when to reduce sensitivity to large error samples. The specific selection criteria are determined by the quality of the effect.

[0127]

[0128]

[0129] Using the big m method, we can solve the bilinear term in equation (33) By linearizing the temperature mixing constraints in equations (39) and (40), we can use the complementary relationship between the two binary variables to solve the problem of The bilinear terms of are linearized.

[0130] Based on the exact mixed integer linear reconstruction of the neural network trained by equations (52)-(54) and (55)-(57), the optimal planning solution of the MILP problem of IEHS planning with network-connected STES can be obtained:

[0131]

[0132] XT h +Da≤h (59)

[0133] Ex h +Fy e ≤g (59)

[0134] 0≤x h ,0≤y e ,0≤q n ,a∈{0,1} (59)

[0135] Among them, x hrepresents all continuous variables in the DHS, including coupled components (such as CHP units and heat pumps), y e represents all continuous variables in EPS, q n represents all continuous variables in the neural network, a represents the binary variable of the STES charging and discharging state and the integer variable introduced in the linearization, ABC (transpose) are the coefficient matrices of the heat network, power grid and neural network respectively, CDEF are the variable coefficient matrices in the constraints respectively, and hg is the constant coefficient matrix.

[0136] To verify the effectiveness of the above method, this embodiment Figure 5 The improved IEEE-33 bus system shown was tested on the IEHS and the DHS of the CHP unit and STES.

[0137] A. Basic data

[0138] The maintenance cost factor is 2%, the life of the CHP unit, high-voltage unit, STES unit, and wind turbine are all assumed to be 20 years, and the interest rate is 5%. The penalty cost of load shedding is set to 5$ / kWh, the maximum power exchanged with the upper grid is 1000kW, the carbon emission cost is stipulated to be 0.1$ / kg, the power emission factor of the upper grid is 0.537kg / kWh, the thermal conductivity (λt,wall, λs,wall) and thickness (st and ss), the top wall and side wall of STES are 0.104 and 0.5W / (m*K) and 0.24 and 0.2m respectively, and the mass flow rate of STES is 15kg / s. The supply and return pipe temperatures are limited to [50℃, 70℃] and [30℃, 50℃] respectively. In order to construct the STES model with reduced time representation, 18 representative days are selected by clustering method. The MILP for the IEHS optimal planning was constructed via YALMIP in MATLAB R2018a and solved using the CPLEX solver on a laptop with an Intel Core i5 and 8GB RAM.

[0139] The economic and technical parameters of the candidate components are shown in Table 1:

[0140] Table 1 Economic and technical parameters of candidate components

[0141]

[0142] B. Case Analysis

[0143] Ⅰ. Effectiveness of Neural Network-Assisted STES Model in Optimizing IEHS Planning

[0144] In order to demonstrate the effectiveness of the optimal IEHS planning of the proposed neural network-assisted STES model, three cases are designed and tested in this embodiment for comparison. In Case 1, the IEHS planning is optimized using the proposed neural network-assisted STES model, where R is considered to be the optimal variable. In Case 2, the IEHS planning is optimized using the physics-based STES model, and R is considered to be the parameter obtained from Case 1. Therefore, the IEHS planning can be solved by a commercial solver. In addition, Case 3 is added to show the computational effectiveness of the proposed deep learning-assisted STES model, in which the IEHS planning is directly optimized using the linearized STES model.

[0145] (1) Effectiveness of neural network encoding of STES nonlinear thermal dynamics

[0146] Since the four proposed neural networks are designed to approximate the nonlinear thermal dynamic model in STES, the training and testing datasets It is created by simulating with discrete values ​​of the input variables. In order to verify the accuracy of the proposed neural network-assisted STES model, about 80% of the dataset is used as training data, while the rest is used for testing. The structures of the four trained neural networks are shown in Table 2:

[0147] Table 2 Structures of four trained neural networks

[0148]

[0149] Table 3 shows the training and testing results of four neural networks under simulated thermal dynamic states, obtained from the proposed neural network-assisted STES model and the physics-based STES model, respectively.

[0150] Table 3 MSE of four neural networks using training and test datasets

[0151]

[0152] The comparison between the predicted output and the labeled output is shown in Figure 6 As shown, from Figure 6 As can be seen from (6a), the neural network is The prediction error is the lowest, about 0.4%, from Figure 6 It can be seen from (6b), (6c) and (6d) that The test MSE of are all relatively low, and the corresponding e is high. This shows that the proposed four neural networks can achieve high accuracy and satisfactory performance.

[0153] also, Figure 7The results of STES calculations using neural network-assisted and physics-based STES models are presented. The comparison results are as follows: Figure 7 As shown in (7a), (7b), (7c), and (7d), the STES model assisted by the neural network and the STES model based on physics have good The daily profiles of are very similar, with the maximum MSE less than 0.01, which corresponds to the results in Table 3. and The prediction error is slightly higher than However, the proposed neural network-assisted STES model has a good prediction error. The overall performance in this regard is satisfactory, which confirms that the neural network model can effectively capture the thermodynamic behavior inside STES and reflect the physical properties of STES.

[0154] (2) Effectiveness of the neural network-assisted STES model in IEHS planning

[0155] The comparison results of the annual investment and operation costs under the three cases and the optimal capacity of the candidate components are shown in Table 4, where the operation and maintenance costs include maintenance costs, fuel and load reduction costs, grid interaction costs, and carbon emission costs. It can be seen that the annual total costs of Case 1 and Case 2 are 205.29×10 4 USD and 275.58×10 4 The investment and operating costs of Case 1 are similar to those of Case 2, so it can be concluded that incorporating the proposed neural network-assisted STES model in IEHS planning can successfully solve the problem with satisfactory accuracy using commercial solvers.

[0156] Table 4 Annual cost and optimal production capacity under three working conditions

[0157]

[0158]

[0159] II. Comparison of Optimization Planning and Operational Effects between IEHS and STES

[0160] In order to illustrate the economic and environmental advantages of incorporating STES into IEHS planning, Case 4, in which STES is not incorporated into IEHS planning, is added.

[0161] (1) Table 5 shows the detailed investment results of all candidate components in Case 1 and Case 4. Compared with Case 1, the capacity of the CHP unit in Case 4 is increased by 25.15%. This is because the addition of STES to the IEHS planning of Case 1 can transfer the surplus heat energy in the non-heating season to the heating season, partially cutting the peak heat load, thereby reducing the heat source capacity. It proves the effectiveness of STES in compensating for the imbalance of heat energy, which leads to a reduction in the required installed capacity of the heat source.

[0162] Table 5 Optimal capacity of candidate components under Case 1 and Case 4

[0163]

[0164] The annual investment and operation costs of IEHS in Case 1 and Case 4 are shown in Table 6. The annual total costs of Case 1 and Case 4 are $205.29×104 and $275.58×104, respectively. In both cases, investment and fuel costs are the two largest parts of the annual total cost. By considering STES in IEHS planning, it can be seen from Table 6 that the installed capacity of the cogeneration unit in Case 1 is 323.68 kW less than that in Case 4. Therefore, the fuel cost and emission cost of Case 1 are reduced by 20.77% and 11.61% compared with Case 4, respectively. Since STES can effectively reduce load reduction through thermal compensation, the load reduction in Case 1 is also lower than that in Case 4. It demonstrates the economic benefits of STES in improving system investment and operation.

[0165] Table 6 Annual investment and operating costs of IEHS in Case 1 and Case 4

[0166]

[0167]

[0168] (2) Comparison of IEHS and STES optimization results

[0169] Figure 8 The results of electricity and heat operation at 20-day intervals under Case 1 and Case 4 show that wind power generation (Wind) constitutes the majority of power generation within the IEHS in Case 1 and Case 4. The heat pump (HP) is one of the main heat suppliers throughout the year due to its high energy conversion efficiency. Figure 8 As shown in (8a) and (8c) of , the introduction of STES in Case 1 leads to a slight reduction in the total power generated by the CHP units during the high heat demand winter period compared to the optimal power and heat dispatch in Case 4. In addition, there is a large amount of surplus power sold to the grid. Figure 8(8b) shows that STES effectively achieves seasonal energy transfer by charging heat in the non-heating season and releasing heat in the heating season, thereby replacing the heat production of the CHP unit. Figure 8 In (8d), the heating demand is mainly met by the heat pump, and the supplementary heat is provided by the CHP unit, which requires an increase in the CHP unit capacity, as shown in Table 5. Therefore, the fuel cost and carbon emissions are greatly reduced as shown in Table 6. The optimal scheduling results show the effectiveness of STES in compensating for daily and seasonal heat imbalances.

[0170] If the above method is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0171] The above description of the embodiments is to facilitate the understanding and use of the invention by those skilled in the art. It is obvious that those skilled in the art can easily make various modifications to these embodiments and apply the general principles described herein to other embodiments without creative work. Therefore, the present invention is not limited to the above embodiments, and improvements and modifications made by those skilled in the art based on the disclosure of the present invention without departing from the scope of the present invention should be within the scope of protection of the present invention.

Claims

1. An optimization planning method for an electric-thermal integrated energy system considering seasonal heat storage, characterized in that: The following steps are involved: A seasonal heat storage model is constructed based on the heat charging and discharging state and the internal thermodynamic state. The heat charging and discharging state includes the heat charging state and the heat discharging state, both of which are binary variables. The internal thermodynamic state includes the inter-cycle state and the intra-cycle state, which respectively represent the initial temperature of the first hour of each day and the temperature change of each hour of each day, both of which are nonlinear variables. The inter-cycle state and the intra-cycle state are respectively obtained by deep neural network prediction; Based on the seasonal heat storage model, a coordinated planning model for an electric-thermal integrated energy system is constructed. The coordinated planning model for the electric-thermal integrated energy system takes minimizing the sum of annual investment cost and operating cost as an objective function, and constructs constraint conditions based on investment constraints and operating constraints. The operating cost includes maintenance cost, fuel and load reduction cost, electricity cost and carbon emission cost. The operating constraints include cogeneration units, heat pumps and thermal load operating constraints, distributed heating network operating constraints and power system operating constraints. The big-m method is used to perform mixed linear transformation on the coordinated planning model of the electric-thermal integrated energy system to obtain the optimal planning scheme of the electric-thermal integrated energy system.

2. The optimization planning method of an electric-thermal integrated energy system considering seasonal heat storage according to claim 1 is characterized in that: The expression of the seasonal heat storage model is as follows: Among them, c w and ρ w are the specific heat capacity and density of water, R and h are the radius and height of the seasonal heat storage model, is the internal thermodynamic state of the seasonal heat storage model at time t, is the mass flow rate of the seasonal heat storage model at time t, and are the supply end temperature and return end temperature of the seasonal heat storage model at time t, and are the heat losses at the top, bottom and sides of the seasonal heat storage model at time t, and d represents the day.

3. The optimization planning method of an electric-thermal integrated energy system considering seasonal heat storage according to claim 2 is characterized in that: Internal thermodynamic state of the seasonal heat storage model at time t The specific expression is as follows: in, For the periodic state, is the state within the cycle, η is the heat loss coefficient; Among them, Φ RP To represent the day collection, and are respectively the charging state and the releasing state. is the soil ambient temperature at time t.

4. The optimization planning method of an electric-thermal integrated energy system considering seasonal heat storage according to claim 3 is characterized in that: The expression of heat loss coefficient η is as follows: Among them, s t 、s b and s are the thickness of the top, bottom and sides, respectively, and λ t,wall , b,wall and λ s,wall are the thermal conductivities of the top, bottom and side surfaces, respectively, is the basic mass flow rate.

5. The optimization planning method of an electric-thermal integrated energy system considering seasonal heat storage according to claim 1 is characterized in that: The deep neural network is a fully connected neural network with a ReLU activation function, and its network parameters are obtained by iterative training by minimizing a loss function, wherein the loss function includes a mean square error and a Huber loss function.

6. The optimization planning method of an electric-thermal integrated energy system considering seasonal heat storage according to claim 1 is characterized in that: The expression of the objective function of the coordinated planning model of the electric-thermal integrated energy system is as follows: C m =αC inv Among them, C inv is the investment cost, C m is the maintenance cost, C F&LS is the fuel and load shedding cost, C grid For electricity cost, is the carbon emission cost, r is the interest rate, β is the component life, c CHP 、c HP 、c STES and c WT are the unit investment costs of the CHP unit, heat pump, seasonal heat storage model and fan, respectively. i CHP,cap , P i HP,cap and are the capacities of the cogeneration unit, heat pump and fan, respectively. is the physical volume of the seasonal heat storage model, α is the maintenance cost coefficient, W d is the length of the day, c fuel and c LS are the unit fuel cost and load reduction penalty cost of the CHP unit, and are the power generation and load reduction of the cogeneration unit, and The power purchase price and power sales price of the power grid are respectively and are the electricity purchased and sold from the power grid, c carbon is the unit carbon emission cost, a CHP and a grid are the carbon emission coefficients of the cogeneration unit and grid power generation, Φ CHP , Φ HP , Φ STES , Φ WT and Φ EL They are the combined heat and power unit kit, heat pump kit, seasonal thermal storage model kit, fan kit and electrical load kit.

7. The optimization planning method of electric-thermal integrated energy system considering seasonal heat storage according to claim 6 is characterized in that: The expression of the constraint conditions of the coordinated planning model of the electric-thermal integrated energy system is as follows: Investment constraints: Where R and h are the radius and height of the seasonal heat storage model, respectively; Cogeneration units, heat pumps and heat load operating constraints: in, and are the calorific value of the cogeneration unit and the heat pump, respectively. and are the power heat transfer coefficients of the cogeneration unit and the heat pump, and are the power generation of the cogeneration unit and the power consumption of the heat pump, respectively; c is the specific heat capacity of water, and are the mass flow rates of the CHP unit, heat pump and heat load, respectively, and are the temperatures of the cogeneration unit, heat pump and heat load supply end, and are the temperatures of the cogeneration unit, heat pump and heat load return end, respectively. and They are heat charging state and heat releasing state respectively; Distributed heating network operation constraints: in, is the mass flow rate in the pipeline, and are the mass flow rate temperature of the heat supply network outlet and the heat recovery network, and are the mass flow rate temperature at the outlet of the heating network and the inlet of the heat recovery network, respectively. and are the temperatures of the nodes ND of the heating network and the heat recovery network, is the basic mass flow rate; Power system operation constraints: in, For large-scale wind power generation, and are the total input flow and total output flow of the corresponding date, time and node respectively. is the power input corresponding to the date, time and node, is the power load corresponding to the date, time point, and node, V j,t,d is the voltage of node j at the corresponding date, time, V i,t,d is the voltage of node i at the corresponding date, time, and r ij and x ij are the resistance parameters and reactance parameters of nodes i to j, respectively, ij,t,d and Q ij,t,d are the real-time power of nodes i to j at date d, time t, and V ref is the reference voltage, ΔV max is the maximum voltage deviation.

8. The optimization planning method of electric-thermal integrated energy system considering seasonal heat storage according to claim 1 is characterized in that: The big-m method is used to linearize the bilinear terms in the constraint conditions of the coordinated planning model of the electric-thermal integrated energy system, and the solution problem of the coordinated planning model of the electric-thermal integrated energy system is converted into a mixed integer linear programming MILP problem and solved.

9. An electronic device comprising a memory, a processor, and a program stored in the memory, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images