Hydrogen-containing multi-energy microgrid scheduling method

Abstract

A hydrogen-containing multi-energy micro-grid scheduling method, comprising: based on the obtained historical operation data of the hydrogen-containing multi-energy micro-grid, constructing a multi-level polyhedral uncertainty set of uncertainty parameters; the uncertainty parameters include renewable energy output power and multi-type load demand of electricity, heat, cold, gas and hydrogen; based on the multi-level polyhedral uncertainty set, a random-robust optimization day-ahead scheduling model of the hydrogen-containing multi-energy micro-grid is established; the model aims to minimize the weighted expectation of the worst-case comprehensive cost under each level of uncertainty set, and meets the unit combination constraint, energy conversion constraint and multi-energy flow balance constraint; the random-robust optimization day-ahead scheduling model is solved by using an improved column and constraint generation algorithm to obtain a day-ahead optimal scheduling plan; based on the day-ahead optimal scheduling plan, a model predictive control method is used to solve a real-time economic scheduling model in each time period to determine the real-time scheduling decision within a day. The method can reduce the robust scheduling cost of the hydrogen-containing multi-energy micro-grid.

Description

Technical Field

[0001] This invention relates to the field of microgrid integrated optimization technology, and in particular to a hydrogen-containing multi-energy microgrid scheduling method. Background Technology

[0002] To reduce fossil fuel pollution, the proportion of renewable energy installed capacity has been increasing year by year. As a system integrating renewable energy power generation and energy storage, multi-energy microgrids (MEMGs) can comprehensively supply various heterogeneous energy flow loads, fully utilizing various energy types through energy conversion equipment to improve overall utilization efficiency. Hydrogen energy, with its high calorific value and pollution-free characteristics, is gradually becoming an important component of multi-energy microgrids. Therefore, research on hydrogen-based MEMGs (H-MEMGs) coupled with hydrogen energy is essential. Due to the uncertainty of renewable energy output and load, higher requirements are placed on H-MEMG scheduling. However, traditional stochastic optimization methods have insufficient anti-interference capabilities and cannot guarantee feasibility across all scenarios. Traditional robust optimization models, on the other hand, suffer from overly conservative results and poor economic efficiency. Therefore, reducing the robust scheduling cost of H-MEMGs is an urgent technical problem to be solved. Summary of the Invention

[0003] To address the problems existing in the prior art, embodiments of this application provide a method, apparatus, computing device, computer storage medium, and product containing a computer program for dispatching hydrogen-containing multi-energy microgrids, which can reduce the cost of robust dispatching of H-MEMGs.

[0004] In a first aspect, embodiments of this application provide a scheduling method for a hydrogen-containing multi-energy microgrid, comprising: constructing a multi-level polyhedral uncertainty set of uncertain parameters in the hydrogen-containing multi-energy microgrid based on acquired historical operating data; the uncertain parameters include renewable energy output power and demand for multiple types of loads such as electricity, heat, cooling, gas, and hydrogen; establishing a stochastic-robust optimization day-ahead scheduling model for the hydrogen-containing multi-energy microgrid based on the multi-level polyhedral uncertainty set; the model aims to minimize the weighted expectation of the worst-case comprehensive cost under each level of uncertainty set, satisfying unit combination constraints, energy conversion constraints, and multi-energy flow balance constraints; solving the stochastic-robust optimization day-ahead scheduling model using an improved column and constraint generation algorithm to obtain the day-ahead optimal scheduling plan; and using a model predictive control method to solve the real-time economic scheduling model on a rolling basis for each time period to determine the intraday real-time scheduling decision.

[0005] In some possible implementations, a multi-level polyhedral uncertainty set of uncertainty parameters in a hydrogen-containing multi-energy microgrid is constructed, including: dividing the original uncertainty set into B-level nested sub-uncertain sets, satisfying... Each layer of uncertainty corresponds to a preset worst-case probability, and the sum of the worst-case probabilities of all layers is 1. For each scheduling period and each uncertainty parameter, the upper and lower deviation boundaries of the b-th layer of uncertainty are determined based on the historical data points of that period and the shortest interval containing a preset number of data points. The preset number is determined by the product of the worst-case probability and the total number of historical data points.

[0006] In some possible implementations, the multi-level polyhedral uncertainty set of uncertainty parameters in a hydrogen-containing multi-energy microgrid is expressed by the following formula:

[0007]

[0008] In the formula, This represents the total number of floors. Characterizing Uncertainty Parameters The uncertain set of layer b, Characterizing the uncertainty parameter at time t The actual value of layer b. Characterizing the uncertainty parameter at time t The predicted value, The direction representing the actual deviation is positive. Characterizing the parameters at time t The largest prediction up-bias in layer b The direction representing the actual deviation is negative. Characterizing the parameters at time t The maximum prediction down bias in layer b Characterizes the scheduling period, Characterization parameters Uncertain budget parameters.

[0009] In some possible implementations, the stochastic-robust optimized day-ahead scheduling model is represented as:

[0010]

[0011] In the formula, x represents the 0-1 integer decision variable of the current day stage. Characterize the feasible region of x. Characterizing the start-up and shutdown costs associated with the current-day decision variables, Characterizing the worst-case probability of the uncertainty set at level b, Characterizing the uncertainty parameter values ​​within the b-th level uncertainty set. Characterizing continuous decision variables during intraday phases, Characterizing 0-1 decision variables during the intraday phase, Characterizes the overall cost of intraday operations. The feasible region characterizes the intraday decision variables under given daily plans and uncertainties.

[0012] Among some possible implementations, the feasible region Constrained by the following formula:

[0013]

[0014] In the formula, C, D, E, F are constant coefficient matrices, d is a constant vector, and the constraints include energy conversion constraints, multi-energy flow balance constraints, and energy storage device operation constraints.

[0015] In some possible implementations, the improved column and constraint generation algorithm includes: decomposing the original problem into a relaxed master problem and subproblems, setting an iteration counter, lower bound, upper bound, and convergence gap; solving the relaxed master problem to obtain the current-day optimal plan for the current iteration, and updating the lower bound; using the current-day optimal plan as known parameters, solving each layer of subproblems in parallel; for each layer of subproblems, relaxing the 0-1 variables of the energy storage charging and discharging state into continuous variables, using duality theory to transform the relaxed subproblems into single-layer maximization problems, and solving to obtain the initial worst-case scenario; using the initial worst-case scenario as known quantities, solving the subproblems under the fixed scenario to obtain the optimal charging and discharging state of the energy storage device; fixing the charging and discharging state, resolving the subproblems to obtain the final determined worst-case scenario and corresponding cost, and updating the upper bound; determining whether the difference between the upper and lower bounds is less than the convergence gap, if so, terminating the iteration and outputting the current-day optimal plan; otherwise, adding the constraints corresponding to the current worst-case scenario to the master problem and continuing the next iteration.

[0016] In some possible implementations, the main problem formula is expressed as:

[0017]

[0018] In the formula, n represents the iteration count of the current loop, and N is the total number of iterations. This represents the worst-case scenario set at level b, determined by the solutions to subsequent subproblems. and These represent the intraday continuous decision variables and the 0-1 decision variables of the energy storage charging and discharging state, respectively, introduced under the worst-case scenario of the b-th layer reference in the nth iteration. As an auxiliary variable, it is used to characterize the worst-case intraday operating cost under the uncertainty set of level b. The worst-case scenario for the b-th layer, determined by the subproblems, is given in the nth iteration.

[0019] In some possible implementations, unit combination constraints include: logical relationship constraints between the start-up and shutdown states, start-up actions, and shutdown actions of the combined cooling, heating, and power (CCHP) units and the electrolysis unit; minimum start-up time and minimum shutdown time constraints; and maximum daily cycle count constraints.

[0020] In some possible implementations, energy conversion constraints include: gas-to-electricity / heat / cooling conversion constraints for combined cooling, heating and power (CCHP) units, electricity-to-hydrogen / heating conversion constraints for electrolysis units, hydrogen-to-electricity / heating conversion constraints for fuel cells, heat-to-cooling conversion constraints for absorption chillers, and electricity-to-heating conversion constraints for electric boilers; each conversion constraint is described by the corresponding conversion efficiency or energy efficiency ratio.

[0021] Among some possible implementations, the multi-energy flow balance constraints include: electric power balance constraints, thermal power balance constraints, cold power balance constraints, gas power balance constraints, and hydrogen power balance constraints; each constraint indicates that the total supply and demand of the corresponding energy type are equal in real time at each time period.

[0022] Secondly, embodiments of this application provide a hydrogen-containing multi-energy microgrid dispatching device, comprising: an acquisition module for acquiring historical operating data of the hydrogen-containing multi-energy microgrid; a processing module for constructing a multi-level polyhedral uncertainty set of uncertain parameters in the hydrogen-containing multi-energy microgrid based on the acquired historical operating data; the uncertain parameters include renewable energy output power and demand for multiple types of loads such as electricity, heat, cooling, gas, and hydrogen; the processing module is further configured to establish a stochastic-robust optimization day-ahead dispatching model of the hydrogen-containing multi-energy microgrid based on the multi-level polyhedral uncertainty set; the model aims to minimize the weighted expectation of the worst-case comprehensive cost under each level of uncertainty set, satisfying unit combination constraints, energy conversion constraints, and multi-energy flow balance constraints; the processing module is further configured to solve the stochastic-robust optimization day-ahead dispatching model using an improved column and constraint generation algorithm to obtain the day-ahead optimal dispatching plan; the processing module is further configured to use a model predictive control method to solve the real-time economic dispatching model on a rolling basis according to time periods based on the day-ahead optimal dispatching plan, and determine the intraday real-time dispatching decision. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a schematic diagram of an H-MEMG system structure provided in an embodiment of this application; Figure 2 This is a schematic flowchart of a hydrogen-containing multi-energy microgrid scheduling method provided in an embodiment of this application; Figure 3 This is a schematic diagram of a multi-layered uncertain set provided in an embodiment of this application; Figure 4 This is a schematic diagram of a multi-layer uncertain set construction provided in an embodiment of this application; Figure 5 This is a schematic diagram illustrating the construction process of a multi-layered uncertain set provided in an embodiment of this application; Figure 6 This is a schematic diagram of the SRO model solution process based on the NC&CG algorithm provided in an embodiment of this application; Figure 7 This application provides a prediction curve for photovoltaic and wind power in a three-layer uncertainty interval. Figure 8 This is a load power prediction curve provided in an embodiment of this application; Figure 9 This application provides a U-type embodiment. 1 The corresponding worst-case power balance; Figure 10 This application provides a U-type embodiment. 2 The corresponding worst-case power balance; Figure 11 This application provides a U-type embodiment. 3 The corresponding worst-case power balance; Figure 12 This is a schematic diagram of the structure of a hydrogen-containing multi-energy microgrid dispatching device provided in an embodiment of this application. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0027] To facilitate understanding of the embodiments of this application, the following will provide further explanation and description with reference to the accompanying drawings and specific embodiments. These embodiments do not constitute a limitation on the embodiments of the present invention.

[0028] Figure 1This is a schematic diagram of an H-MEMG system structure provided in an embodiment of this application. For example... Figure 1 As shown, the H-MEMG system includes wind turbines (WT) and photovoltaic (PV) arrays that absorb renewable energy; battery energy storage (BES) systems to balance electrical loads; thermal energy storage (HES) systems to balance thermal loads; combined cooling, heating, and power (CCHP) units that can simultaneously output electricity, heat, and cooling power; hydrogen energy storage systems including electrolyzers (ED), hydrogen storage tanks (HT), and fuel cells (FC); and components such as electric boilers (EB) and absorption chillers (AC) for converting heterogeneous energy sources to balance loads. The entire microgrid system obtains electricity through connection to the upper-level grid, and the gas pipeline network supplies gas to meet the system's heating and cooling loads and gas consumption needs, while also exchanging hydrogen energy with the external hydrogen market. During operation, the system must meet the needs of multiple energy sources, including electricity, heat, cooling, gas, and hydrogen, while absorbing a high proportion of renewable energy to achieve carbon emission reduction goals.

[0029] This application provides a hydrogen-containing multi-energy microgrid scheduling method that combines the advantages of stochastic optimization and robust optimization. Based on historical data of uncertain parameters, a multi-level uncertainty set can be constructed to more accurately describe the distribution characteristics of uncertain parameters. A stochastic-robust optimization scheduling model is established, enabling the scheduling plan to simultaneously satisfy multiple loads of electricity, heat, cooling, gas, and hydrogen, taking into account both economic efficiency and resistance to source-load fluctuations.

[0030] For example, Figure 2 A schematic flowchart of a hydrogen-containing multi-energy microgrid scheduling method provided in an embodiment of this application is shown. Figure 2 As shown, the method may include the following steps:

[0031] S21: Based on the historical operation data of the hydrogen-containing multi-energy microgrid, construct a multi-level polyhedral uncertainty set of uncertainty parameters in the hydrogen-containing multi-energy microgrid; the uncertainty parameters include the output power of renewable energy and the demand for multiple types of loads such as electricity, heat, cooling, gas and hydrogen.

[0032] In this embodiment, during the actual operation of a hydrogen-containing multi-energy microgrid, the output of renewable energy sources such as wind power and photovoltaics, as well as the demand for various loads including electricity, heat, cooling, gas, and hydrogen, all exhibit significant randomness and volatility. These uncertainties directly affect the feasibility and economy of the scheduling plan. In robust optimization models, setting a reasonable uncertainty set can improve the optimization effect; therefore, uncertainty modeling is crucial. For classic robust optimization, a single-layer polyhedral uncertainty set, as shown in the following equation, is typically used to describe it:

[0033]

[0034] In the formula, Characterization parameters Uncertain set, Characterizing various uncertainties, Characterizing the uncertainty parameter at time t The actual value, Characterizing the parameters at time t The predicted value, Characterizing the parameters at time t The largest prediction deviation, Characterizing the parameters at time t The largest prediction bias, and Characterizes the direction (positive and negative) of the actual deviation. Characterizes the scheduling period, Characterization parameters Uncertain budget parameters. Among them, lie in[ , Between ] . By changing The size is adjustable to account for the degree of uncertainty. Specifically, when... When, it indicates the parameter Without uncertainty, robust optimization degenerates into a deterministic optimization model. When, it indicates the parameter Uncertainty exists throughout the entire scheduling cycle, and the polyhedral uncertainty set is transformed into a box-shaped uncertainty set.

[0035] However, based on the uncertainty set of a single-layer polyhedron, the worst-case scenario found often takes the boundary value in certain periods and the predicted value in other periods, but such a worst-case scenario has little significance for actual scheduling. Historical data shows that the actual values ​​of the uncertain parameters mostly fluctuate within an interval near the predicted values, with only a few periods showing large deviations, thus exhibiting excessive conservatism. In contrast, stochastic optimization methods are less conservative, but the incomplete robustness of the scheduling strategy and the difficulty in obtaining the actual distribution of uncertain parameters limit their application in the context of source-load uncertainty.

[0036] Therefore, in this embodiment, the multi-level polyhedral uncertainty set constructed using finite historical data is represented as:

[0037]

[0038]

[0039]

[0040]

[0041] This represents the uncertainty set of a single-layer polyhedron. In the case of dividing the uncertainty into B layers, where b represents the b-th layer, then... Characterization parameters The uncertainty set of the b-th layer increases with each layer, and the uncertainty set of the outermost layer is the same as the uncertainty set of a single layer. Same, that is Equivalent to . Characterization parameters belong The worst-case probability.

[0042] For example, please refer to Figure 3 , Figure 3 A schematic diagram of the partitioning of multi-level uncertain sets is shown. For ease of explanation, in Figure 3 Parameters omitted This will be illustrated using a four-layer uncertain set as an example. Figure 3 In, the set of all uncertainties , , and All are centered at the same point, which is the predicted value of the uncertain parameter. The smaller set is completely contained within the larger set, i.e. And their worst-case probabilities The sum is 1.

[0043] The worst-case probability differs from the standard probability of an uncertainty set. For example, parameters The probability realized in this set is the standard probability, which is 100% in this example. However... Worst-case scenario occurs (when) The probability of the occurrence of the boundary (of the boundary) is the "worst-case probability", which is much less than 1. Table 1 shows an example of the prediction bias and probability of the uncertain parameters.

[0044] Table 1

[0045]

[0046] Using historical data, uncertain sets can be constructed relatively quickly. For example, Figure 4 An example of constructing a multi-level uncertain set is shown below; please refer to [link / reference]. Figure 4 With wind power output For example, and Let represent the negative and positive deviations of the predicted value for time period t, respectively. This is based on the given worst-case probability. This allows us to obtain the distribution interval corresponding to time period t and the maximum positive and negative deviations of the interval. Figure 5 A schematic diagram illustrating the construction process of multi-level uncertain sets is shown. Please refer to... Figure 4 and Figure 5 After inputting historical wind power processing data, input the number of layers for uncertainty set partitioning and the worst-case probability. Based on the worst-case probability... The number of data points within the interval is calculated using the following formula: In the formula, Characterization The number of data points contained in it This represents the total number of historical wind turbine output power points at time t. All data points are sorted to obtain the sorted data sequence at time t. Traverse from the beginning of the sequence to the end, and find the sequence containing... The shortest interval of data points ,in , This represents the maximum reverse deviation value for this interval during time period t. This represents the maximum positive deviation. By taking t = 1 ~ T and b = 1 ~ B respectively, the wind power output can be obtained. each layer of uncertainty set This section uses wind power output as an example to illustrate how this method can be used to determine the multilevel uncertainty sets of other uncertain parameters. After determining the uncertainty sets of each uncertain parameter, their summation yields the improved multilevel uncertainty set: .in, The b-th layer of uncertainty is represented by the aggregated data.

[0047] S22: Based on multi-level polyhedral uncertainty sets, a stochastic-robust day-ahead scheduling model for hydrogen-containing multi-energy microgrids is established. The model aims to minimize the weighted expectation of the worst-case comprehensive cost under each level of uncertainty set, and satisfies unit combination constraints, energy conversion constraints, and multi-energy flow balance constraints.

[0048] In this embodiment, after constructing a multi-level polyhedral uncertainty set, a stochastic-robust optimized day-ahead scheduling model for a hydrogen-containing multi-energy microgrid can be established based on this multi-level polyhedral uncertainty set. The scheduling optimization objective of the H-MEMG system is to minimize the overall cost, including operation and maintenance costs, energy consumption costs, and unit start-up, shutdown, and operation costs, within a scheduling cycle, while simultaneously satisfying various constraints, including unit combination constraints, component operation constraints, and energy flow balance constraints. Please continue to refer to... Figure 1The H-MEMG system structure in this paper establishes a day-ahead scheduling model based on stochastic-robust optimization (SRO) using a multi-level polyhedral uncertainty set. This model combines the advantages of traditional stochastic optimization strategies with robust optimization strategies. The decision variables in the model are divided into two categories: one is the 0-1 integer decision variables in the day-ahead stage, denoted as vector x, which includes the start-up and shutdown states, start-up actions, and shutdown actions of the combined cooling, heating, and power (CCHP) units and electrolysis units. These variables need to be determined in advance in the day-ahead stage and remain unchanged during day-ahead operation. The other category consists of continuous variables and 0-1 variables during the intraday period, denoted as vectors y and z, respectively. y includes continuous variables such as the charging and discharging power of various energy storage devices, the input power of combined cooling, heating and power (CCHP) units, the input and output power of electric boilers and absorption chillers, the power exchanged with the upstream power grid, and the power exchanged with the hydrogen energy market. z is used to characterize the charging and discharging state of energy storage devices such as batteries, thermal energy storage, and hydrogen energy storage. This variable is a 0-1 variable, indicating whether the energy storage device is in charging mode or discharging mode at a certain time.

[0049] The objective function is expressed as follows:

[0050]

[0051] In the formula, The feasible region of the variable x. The transpose of the cost coefficient vector with the same dimension as x. Characterization and Transpose of a cost coefficient vector of the same dimension The right-hand side of the day-ahead stage constraint is represented by the following terms. The right-hand side of the constraint condition during the intraday phase is represented by... Characterizing various uncertainties in a given x and at level b. Intraday scheduling adjusts decision variables ( , The feasible region of (). B, C, D, E, F are constant coefficient matrices of various constraints during operation, and c, d are constant vectors. It is worth noting that in this application, unless otherwise specified, superscripts and subscripts usually refer to the same variable, such as and Both refer to the intraday continuous variable y in the b-th layer.

[0052] Regarding the three constraints, the first constraint represents the linear inequality constraints of the day-ahead phase, such as the minimum start-up and shutdown time constraint, the maximum number of daily cycles constraint, and start-up and shutdown state logic constraints. The second constraint represents all constraints of the intraday phase, including energy conversion constraints (such as the relationship between the output power and input power of the CCHP), multi-energy flow balance constraints (the supply and demand balance of the five energy sources: electricity, heat, cooling, gas, and hydrogen), and energy storage equipment operation constraints (such as capacity upper and lower limits, charge and discharge rate limits, etc.). , For decision variables. In the third constraint, This indicates that the decision variable x is an integer variable between 0 and 1, such as the start-up and shutdown status of the unit (1 indicates operation, 0 indicates shutdown). Indicates intraday continuous decision variables It is a real number variable that can take values ​​continuously within a certain range. Indicates the intraday energy storage charging and discharging state variables These are 0-1 integer variables. Therefore, the optimization objective of this stochastic-robust optimized day-ahead scheduling model is the day-ahead start / stop cost. This involves minimizing the sum of the worst-case intraday costs across all uncertainty levels. Specifically, for the b-th level uncertainty level, the worst-case uncertainty is chosen for realization. In the worst-case scenario, intraday scheduling selects the optimal continuous decision-making and energy storage charging / discharging state to minimize intraday operating costs. The lowest. Or, for each set of uncertainties, we can obtain the factor that maximizes the intraday scheduling cost under a given daily schedule x. , In this way, the SRO model can find a solution that minimizes the expected total cost within a multi-layered uncertainty set, thus determining the optimal plan for the current day. This reduces the conservatism of the scheduling scheme and its dependence on the completeness of uncertain parameter information, while also taking into account the economic requirements of scheduling.

[0053] For example, the constraints for CCHP unit combination include: the logical relationship between start / stop state variables and operating state variables, meaning the unit can only enter the operating state after startup and can only be shut down while in the operating state; and the minimum start / stop time constraint, meaning the unit must run continuously for at least a certain number of time periods after startup and must be continuously shut down for at least a certain number of time periods after shutdown. Additionally, there is a daily maximum cycle count constraint, meaning the total number of times the unit starts or shuts down in a day cannot exceed a set upper limit to avoid frequent start / stop cycles damaging equipment lifespan. In a scheduling cycle T, the CCHP unit combination constraint formula is expressed as:

[0054]

[0055]

[0056]

[0057]

[0058] In the formula, u t CCHP v t CCHP r t CCHP N represents the 0-1 variables representing the CCHP unit's startup, shutdown, and operating states at time t. CCHP UT represents the maximum number of daily cycles of the equipment. CCHP DT CCHP These represent the minimum start-up time and minimum shutdown time of the equipment, respectively. This constraint indicates that for any time t ( The change in the unit's current operating status relative to the previous period is exactly equal to the number of startup actions minus shutdown actions in this period.

[0059] For example, energy conversion constraints describe the physical relationships between the inputs and outputs of various energy conversion devices. CCHP units, electrolysis units, fuel cells, absorption chillers, and electric boilers respectively achieve energy flow conversions from gas to electricity / heat / cold, electricity to hydrogen / heat, hydrogen to electricity / heat, heat to cold, and electricity to heat. Specifically, a combined cooling, heating, and power (CCHP) unit uses natural gas as input and outputs electricity, heat, and cold energy; its output power is equal to the input power multiplied by the corresponding conversion efficiency. An electrolysis unit uses electricity as input and outputs hydrogen and heat energy; its output power is equal to the input power multiplied by the electrolysis efficiency and the heat production efficiency, respectively. A fuel cell uses hydrogen as input and outputs electricity and heat energy. An electric boiler uses electricity as input and outputs heat energy; its output heat power is equal to the input electrical power multiplied by the electricity-heat conversion efficiency. An absorption chiller uses heat energy as input and outputs cold energy; its output cold power is equal to the input heat power multiplied by the energy efficiency ratio. The energy conversion constraint is expressed as:

[0060]

[0061]

[0062]

[0063]

[0064]

[0065] Among them, P,H,Q,P NG ,P H These represent electrical, thermal, cooling, gas, and hydrogen power, respectively, with subscripts indicating the type of equipment. The superscripts e, h, Q, NG, and H represent the efficiency of various devices in converting energy into electricity, heat, cold, gas, and hydrogen, respectively; COP represents the energy efficiency ratio of an absorption chiller.

[0066] For example, regarding energy flow balance constraints, H-MEMG needs to simultaneously meet the real-time balance of five different types of loads during operation, while also accommodating the consumption of renewable energy sources such as wind and solar power. During each dispatch period, the total supply and demand of the five energy sources—electricity, heat, cooling, gas, and hydrogen—must be equal in real time. This constraint can be expressed as:

[0067]

[0068]

[0069]

[0070]

[0071]

[0072] Among them, L t e , L t h , L t Q , L t NG , L t H These represent the uncertain load demands for electricity, heat, cooling, gas, and hydrogen, respectively.

[0073] In the stochastic-robust optimization day-ahead scheduling model for a hydrogen-containing multi-energy microgrid constructed in this embodiment, the day-ahead plan x is a pre-determined intraday variable, while y and z can be flexibly adjusted according to actual uncertainties. When the uncertainty parameters change arbitrarily within the b-th level uncertainty set, the model guarantees the existence of a feasible intraday scheduling scheme, meaning the system can meet all load demands without load shedding or curtailment of wind and solar power. Furthermore, since different levels of uncertainty sets assign different probability weights, the model automatically adjusts the weights of inner-layer scenarios with higher probabilities when optimizing the day-ahead plan.

[0074] S23: An improved column and constraint generation algorithm is used to solve the stochastic-robust optimization day-ahead scheduling model to obtain the day-ahead optimal scheduling plan.

[0075] In this embodiment, the constructed SRO model exhibits a complex min-max-min form, and since the operating state of the energy storage device can be switched within a day, the innermost min problem involves 0-1 variables. The non-convexity of the optimization problem significantly increases the difficulty of solving it. While traditional nested column and constraint generation algorithms can solve this problem, they rely on enumerating integer variables and require two layers of iterative solutions, resulting in high computational time and failing to meet the efficiency requirements of practical engineering applications. Therefore, this invention employs an improved Column and Constraint Generation (NC&CG) algorithm to solve the problem. The original problem is decomposed into a relaxation master problem (MP) and a subproblem (SP). Through iterative interaction between the master and subproblems, the optimal solution to the original problem is gradually approximated. The relaxation strategy avoids the iterative solution of the inner SP layer, resulting in faster convergence and significantly improved computational efficiency.

[0076] Specifically, the main part of the computational process is solving MP and SP. For the main problem, auxiliary variables can be introduced. This replaces the optimal objective value of each subproblem, thus transforming the original problem into an equivalent combination of a main problem and B independent subproblems. The main problem includes the current decision variable x and auxiliary variables. The objective function is to minimize the sum of the day-ahead start / stop costs and the weighted sum of the auxiliary variables at each level. The constraints of the main problem include: the feasible region constraint of the day-ahead decision variable x itself, i.e., x∈X; and for each level b and the worst-case scenario determined in each iteration, corresponding feasibility constraints are added. These constraints require that a feasible intraday scheduling scheme exists under the worst-case scenario, and the corresponding intraday operating cost does not exceed the auxiliary variables. It should be noted that in the main problem, all variables related to intraday scheduling are considered as specific implementations under a given scenario. Therefore, the main problem is a mixed-integer linear programming problem that can be solved efficiently.

[0077] The main problem formula is expressed as:

[0078]

[0079] In the formula, n represents the iteration count of the current loop. This represents the worst-case scenario set at level b, determined by solving subsequent subproblems. After solving MP, the obtained x is used as a known parameter x. * Input into SP.

[0080] The subproblem SP can be represented as:

[0081]

[0082] in, This is the day-ahead crewing plan obtained by solving MP in the nth iteration. Since the problems under different levels of uncertainty are independent, SP can be solved in parallel. For ease of explanation later, the subscript 'b' is omitted, and only the case of a single-level uncertainty is described. The 0-1 variable z in the subproblem SP restricts the solution; the idea behind the NC&CG algorithm is to relax it into a continuous variable and determine a preliminary worst-case scenario. Then, based on this, we substitute it into the original problem to finally obtain the accurate worst-case scenario. .

[0083] The relaxon problem RSP is represented as:

[0084]

[0085] In RSP, the inner-layer min problem only includes continuous variables; therefore, duality theory can be used to transform the max-min problem into a single-layer max problem. The result is denoted as... .Will When these known quantities are brought back to SP, SP is transformed into a single-layer MIP problem SP-u, which can quickly solve for the optimal charge / discharge state of the energy storage device. Finally, the charge / discharge state is fixed. The SP problem is transformed into a max-min LP problem SP-z, and then recalculated to obtain the final worst-case scenario. and cost .

[0086] For example, Figure 6 A schematic diagram of the entire solution process for the SRO model based on the NC&CG algorithm is shown. Figure 6 As shown, solving the SRO model may include the following steps:

[0087] During the algorithm's initialization phase, the iteration counter n=1, the lower bound LB is set to negative infinity, the upper bound UB is set to positive infinity, and the convergence gap δ is set to a sufficiently small positive number, for example, δ=0.01. Simultaneously, the worst-case scenario set for each layer is initialized to an empty set.

[0088] Based on the established worst-case scenario set at layer b. Above, we solve the main problem and obtain the optimal plan determined in the nth iteration. With objective function ,renew .

[0089] The optimal plan As input, solve the subproblem SP in parallel. bFor each subproblem at each level, the relaxation subproblem RSP is solved first. This is because the relaxation subproblem RSP incorporates the 0-1 variables of the energy storage charging and discharging states in the inner-level minimization problem. Relaxing the variables to be continuous, that is, expanding their range from {0,1} to the interval [0,1], transforms the inner-layer problem into a linear programming problem containing only continuous variables. Using strong duality theory, this is further transformed into a single-layer maximization problem, which can be solved exactly in polynomial time. The optimal plan... As input, solve RSP b This allows us to obtain a preliminary worst-case scenario. .

[0090] Next, we will present the initial worst-case scenario. Substitute these parameters into the original subproblem. The subproblem then degenerates into an intraday scheduling optimization problem under a fixed uncertainty scenario, aiming to minimize intraday operating costs given the daily plans and the realization of uncertainties. This problem involves continuous variables. and 0-1 variables This is a mixed-integer linear programming problem that can be solved directly. Solving this problem yields the optimal charge / discharge state of the energy storage device. .

[0091] Optimal charge and discharge state With the parameters fixed as known, we resubstitute them into the subproblem. At this point, the inner-level min problem within the subproblem only contains continuous variables. This again becomes a linear programming problem. Utilizing duality theory, this subproblem is transformed into a single-level max problem, and solving it yields the final worst-case scenario. and the corresponding daily operating costs This allows us to obtain the worst-case scenario and corresponding worst-case cost for the b-th level subproblem under the current day-ahead plan. Since the subproblems at each level are independent, they can be executed in parallel on different computational cores. After all the B-level subproblems have been solved, the worst-case costs from each level are summed, a weighted sum is calculated, and the expected worst-case cost under the current day-ahead plan is obtained. The upper bound is then updated accordingly. .

[0092] Finally, check if the convergence condition |UB-LB|≤δ holds. If it does, the algorithm terminates and outputs the result at this point. As the optimal plan If this is not true, let n = n + 1, then... Add ,Right now Add new variables in MP New constraints and Return to the previous step and solve MP again.

[0093] The solution obtained by the NC&CG algorithm This is the optimal daily plan under the stochastic-robust optimization model, which can serve as a basis for guiding subsequent intraday real-time economic scheduling.

[0094] S24: Based on the day-ahead optimal scheduling plan, an enhanced model predictive control method is adopted to solve the real-time economic scheduling model on a rolling basis for each time period to determine the intraday real-time scheduling decision.

[0095] In this embodiment, based on the determined day-ahead optimal scheduling plan, an enhanced model predictive control method is adopted to solve the real-time economic scheduling model on a rolling basis for each time period, thereby determining the intraday real-time scheduling decision. The day-ahead scheduling plan provides a baseline scheme for the intraday operation of H-MEMG. However, since the actual values ​​of renewable energy output and load demand often deviate from the predicted values ​​in real-time operation, it is necessary to dynamically adjust the scheduling decision based on the latest acquired real-time information to ensure the safe and economical operation of the system.

[0096] During scheduling in time period t, decisions need to be made using the latest revealed uncertainty parameters for time period t and the predicted uncertainty parameters for the next T' look-ahead time periods. To overcome the short-sightedness of traditional MPC, this invention employs an enhanced MPC method, extending the look-ahead time period to the end of the scheduling cycle, i.e., time period T. This method improves the depth of scheduling while aligning with the time-series coupling characteristics of energy storage devices. Based on optimal planning... The real-time intraday scheduling model for time period t is expressed as:

[0097]

[0098] In the formula, This represents the scheduling decision to be implemented during time period t. This refers to scheduling decisions made in the future time period (τ>t) under predicted conditions. The scheduling decisions completed up to this point are denoted as follows: The uncertain parameters in time period t are denoted as... The prediction of uncertain parameters for the future time period is denoted as . Represents all deterministic scheduling constraints for time period t. This indicates that the scheduling decision depends on the previous time period. and prediction parameters The predictive scheduling constraints are given by T, where T is the scheduling period.

[0099] Taking a regional-level H-MEMG as an example, its structure is as follows: Figure 1 As shown in Table 2, the typical values ​​of the core parameters for the optimized scheduling of this microgrid system are presented.

[0100] Table 2

[0101]

[0102] Based on historical data, worst-case probabilities p1=0.6, p2=0.3, and p3=0.1 are selected to construct a three-layer uncertainty set for the source-load uncertainty parameters, corresponding to U. 1 U 2 U 3 The prediction curves for photovoltaic and wind power and the uncertainty interval of the 3rd layer are as follows: Figure 7 As shown. The prediction curves for the five load categories are as follows. Figure 8 As shown. The uncertainty budget parameter for the output of the two types of renewable energy is set to 8, and the uncertainty budget parameter for the five types of load is set to 4. The convergence gap of the NC&CG algorithm is set to 1%. Figure 9 , Figure 10 , Figure 11 The power balance of the H-MEMG system is shown under the worst-case scenarios corresponding to the three-level uncertainty sets. In each figure, (a) represents the balance of electrical power, and (b) represents the balance of thermal power. As shown in the figures, in the worst-case scenario, the load reaches its upper limit during the higher predicted periods (7-8 am, 9-10 am, 12-1 pm, and 2-3 pm), while the output of renewable energy reaches its lower limit during the lower predicted periods. The combined effect of these two factors increases operating costs. In the day-ahead dispatch plan, the ED (Electric Power Generation) unit is always in operation. The CCHP (Concentrated Concentration Power Generation) unit starts operation after 7 am when electricity demand is high and electricity prices are relatively high, minimizing the unit's online costs. The H-MEMG system can effectively resist the impact of uncertainties and reliably ensure the absorption of renewable energy. Using the optimal day-ahead dispatch plan, the intraday robust dispatch costs for the worst-case scenarios determined by the three-level uncertainty sets are calculated to be RMB 14,935.1, RMB 15,788.8, and RMB 17,563.5, respectively. After comprehensive weighting, the objective function of the scheduling model is 15454.1 yuan. The scheduling results show that if the traditional single-layer polyhedral uncertainty set is used, the robust comprehensive scheduling cost is maximized by U. 3 The cost under the uncertain set is 17563.5 yuan. However, in reality, only U is considered. 3 The results obtained from the set are more conservative.

[0103] To further verify the advantages of the enhanced MPC-based real-time economic scheduling model in intraday real-time economic scheduling, 1000 out-of-sample scenarios were generated using the Monte Carlo method. The results of the three different models in the 1000 test scenarios are summarized in Table 3.

[0104] Table 3

[0105]

[0106] Where F represents the total cost; "feasibility rate" represents the proportion of scenarios that are ultimately feasible using the corresponding daily schedules from 1000 test scenarios. The analysis shows that while the traditional stochastic optimization model has a cost advantage, it cannot guarantee feasibility across all scenarios. In extreme scenarios, it may require load shedding or wind / solar curtailment, resulting in greater economic losses. Both the traditional robust optimization model and the stochastic-robust optimization model proposed in this invention are robust, but the traditional robust model is more expensive, verifying the conclusion that considering only the largest uncertainty set leads to an overly conservative scheduling strategy.

[0107] The above is an introduction to the hydrogen-containing multi-energy microgrid dispatching method provided in the embodiments of this application. First, the day-ahead dispatching problem is solved using a day-ahead dispatching model. Then, the real-time economic dispatching problem is calculated sequentially time-by-time, allowing for the optimal dispatching decision for time period t based on the actual renewable energy output and load conditions. This invention achieves optimal operation of H-MEMGs with renewable energy and energy storage systems, while ensuring the robustness of the solution to uncertainties.

[0108] It is understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. Furthermore, in some possible implementations, each step in the above embodiments may be selectively executed according to actual circumstances; it may be partially or fully executed, without limitation here. All or part of any feature of any embodiment of this application can be freely and arbitrarily combined without contradiction. The combined technical solutions are also within the scope of this application.

[0109] Based on the methods in the above embodiments, this application also provides a hydrogen-containing multi-energy microgrid dispatching device. For example, Figure 12 A schematic diagram of a hydrogen-containing multi-energy microgrid dispatching device according to an embodiment of this application is shown. This device is deployed in a hydrogen-containing multi-energy microgrid. Figure 12 As shown, the hydrogen-containing multi-energy microgrid dispatching device 1200 includes an acquisition module 1201 and a processing module 1202.

[0110] Among them, the acquisition module 1201 is used to acquire historical operating data of hydrogen-containing multi-energy microgrids.

[0111] The processing module 1202 is used to construct a multi-level polyhedral uncertainty set of uncertainty parameters in the hydrogen-containing multi-energy microgrid based on the historical operation data of the acquired hydrogen-containing multi-energy microgrid; the uncertainty parameters include the output power of renewable energy and the demand for multiple types of loads such as electricity, heat, cooling, gas and hydrogen.

[0112] The processing module 1202 is also used to establish a stochastic-robust optimization day-ahead scheduling model for a hydrogen-containing multi-energy microgrid based on the multi-level polyhedral uncertainty set; the model aims to minimize the weighted expectation of the worst-case comprehensive cost under each level of uncertainty set, and satisfies unit combination constraints, energy conversion constraints and multi-energy flow balance constraints.

[0113] The processing module 1202 is also used to solve the stochastic-robust optimization day-ahead scheduling model using an improved column and constraint generation algorithm to obtain the day-ahead optimal scheduling plan;

[0114] The processing module 1202 is also used to determine the intraday real-time scheduling decision by using the model predictive control method to solve the real-time economic scheduling model on a rolling basis according to the day-ahead optimal scheduling plan.

[0115] It should be understood that the above-described device is used to execute the methods in the above embodiments. The implementation principle and technical effect of the corresponding program modules in the device are similar to those described in the above methods. The working process of the device can be referred to the corresponding process in the above methods, and will not be repeated here.

[0116] It is understood that the various numerical designations used in the embodiments of this application are merely for descriptive convenience and are not intended to limit the scope of the embodiments of this application.

Claims

1. A method for dispatching hydrogen-containing multi-energy microgrids, characterized in that, The method includes: Based on the historical operating data of the hydrogen-containing multi-energy microgrid, a multi-level polyhedral uncertainty set of uncertainty parameters in the hydrogen-containing multi-energy microgrid is constructed. The uncertainty parameters include renewable energy output and demand for multiple types of loads, including electricity, heat, cooling, gas, and hydrogen. Constructing the multi-level polyhedral uncertainty set of uncertainty parameters in the hydrogen-containing multi-energy microgrid includes: dividing the original uncertainty set into B-layer nested sub-uncertain sets, satisfying… Each layer of uncertainty corresponds to a preset worst-case probability, and the sum of the worst-case probabilities of all layers is 1. For each scheduling period and each uncertainty parameter, based on the historical data points of that period, the upper and lower deviation boundaries of the b-th layer of uncertainty are determined by the shortest interval containing a preset number of data points; the preset number is determined by the product of the worst-case probability and the total number of historical data points. Based on the aforementioned multi-level polyhedral uncertainty set, a stochastic-robust day-ahead scheduling model for a hydrogen-containing multi-energy microgrid is established. The model aims to minimize the weighted expectation of the worst-case comprehensive cost under each level of uncertainty set, satisfying unit combination constraints, energy conversion constraints, and multi-energy flow balance constraints. The stochastic-robust day-ahead scheduling model is expressed as follows: In the formula, x represents the 0-1 integer decision variable of the current day stage. Characterize the feasible region of x. Characterizing the start-up and shutdown costs associated with the current-day decision variables, Characterizing the worst-case probability of the uncertainty set at level b, Characterizing the uncertainty parameter values ​​within the b-th level uncertainty set. Characterizing continuous decision variables during intraday phases, Characterizing 0-1 decision variables during the intraday phase, Characterizes the overall cost of intraday operations. The feasible region characterizes the intraday decision variables under given daily plans and uncertainties. An improved column and constraint generation algorithm is used to solve the stochastic-robust optimization day-ahead scheduling model to obtain the day-ahead optimal scheduling plan; Based on the aforementioned day-ahead optimal scheduling plan, a model predictive control method is used to solve the real-time economic scheduling model on a rolling basis for each time period to determine the intraday real-time scheduling decision.

2. The method according to claim 1, characterized in that, The multi-level polyhedral uncertainty set of the uncertainty parameters in the hydrogen-containing multi-energy microgrid is expressed by the following formula: In the formula, This represents the total number of floors. Characterizing Uncertainty Parameters The uncertain set of layer b, Characterizing the uncertainty parameter at time t The actual value of layer b. Characterizing the uncertainty parameter at time t The predicted value, The direction representing the actual deviation is positive. Characterizing the parameters at time t The maximum prediction bias in layer b The direction representing the actual deviation is negative. Characterizing the parameters at time t The maximum prediction down bias in layer b Characterizes the scheduling period, Characterization parameters Uncertain budget parameters.

3. The method according to claim 1, characterized in that, feasible region Constrained by the following formula: In the formula, C, D, E, F are constant coefficient matrices, d is a constant vector, and the constraints include energy conversion constraints, multi-energy flow balance constraints, and energy storage device operation constraints.

4. The method according to claim 1, characterized in that, The improved column and constraint generation algorithm includes: The original problem is decomposed into a relaxation master problem and subproblems, and an iteration counter, lower bound, upper bound, and convergence gap are set. Solve the relaxation master problem to obtain the day-ahead optimal plan for the current iteration, and update the lower bound; Using the aforementioned optimal plan as known parameters, we solve the sub-problems at each level in parallel. For each sub-problem, the 0-1 variables of the energy storage charging and discharging state are relaxed to continuous variables. The relaxation sub-problem is transformed into a single-layer maximization problem using duality theory. The initial worst-case scenario is obtained by solving the problem. Using the initial worst-case scenario as known quantities, we solve the sub-problems under the fixed scenario to obtain the optimal charging and discharging state of the energy storage device. By fixing the charging and discharging states, the subproblem is solved again to obtain the final worst-case scenario and corresponding cost, and the upper bound is updated. Determine if the difference between the upper and lower bounds is less than the convergence gap. If so, terminate the iteration and output the current day-to-day optimal plan; otherwise, add the constraint corresponding to the current worst-case scenario to the main problem and continue to the next iteration.

5. The method according to claim 4, characterized in that, The formula for the main problem is expressed as follows: In the formula, n represents the iteration count of the current loop, and N is the total number of iterations. This represents the worst-case scenario set at level b, determined by the solutions to subsequent subproblems. and These represent the intraday continuous decision variables and the 0-1 decision variables of the energy storage charging and discharging state, respectively, introduced under the worst-case scenario of the b-th layer reference in the nth iteration. As an auxiliary variable, it is used to characterize the worst-case intraday operating cost under the uncertainty set of level b. The worst-case scenario for the b-th layer, determined by the subproblems, is given in the nth iteration.

6. The method according to claim 1, characterized in that, The unit combination constraints include: logical relationship constraints between the start-up and shutdown states, start-up actions and shutdown actions of the combined cooling, heating and power (CCHP) units and the electrolysis unit; minimum start-up time and minimum shutdown time constraints; and maximum daily cycle count constraints.

7. The method according to claim 1, characterized in that, The energy conversion constraints include: gas-to-electricity / heat / cooling conversion constraints for combined cooling, heating and power (CCHP) units, electricity-to-hydrogen / heating conversion constraints for electrolysis units, hydrogen-to-electricity / heating conversion constraints for fuel cells, heat-to-cooling conversion constraints for absorption chillers, and electricity-to-heating conversion constraints for electric boilers; each conversion constraint is described by the corresponding conversion efficiency or energy efficiency ratio.

8. The method according to claim 1, characterized in that, The multi-energy flow balance constraints include: electric power balance constraints, thermal power balance constraints, cold power balance constraints, gas power balance constraints, and hydrogen power balance constraints; each constraint indicates that the total supply and demand of the corresponding energy type are equal in real time at each time period.

Images