Robust optimization method for multi-energy microgrid considering clustering information of uncertain scenarios

Abstract

Aiming at the high uncertainty of renewable energy output in multi-energy microgrids, the present invention develops a robust optimization method for multi-energy microgrids that considers clustering information of uncertain scenarios; a fuzzy set based on Wasserstein distance that considers scenario information is designed to describe the uncertainty of photovoltaic output, and a k-means time series clustering algorithm based on dynamic time warping is used to capture typical scenario information of photovoltaic output. The method of the present invention can significantly improve the computational efficiency of distributed robust optimization, effectively resist the fluctuation of photovoltaic output in multi-energy microgrids, and maintain good economic benefits.

Description

Technical Field

[0001] The present invention belongs to the technical field of microgrid operation and dispatching, and develops a multi-energy microgrid robust optimization method considering uncertain scenario clustering information for a multi-energy microgrid with a high proportion of photovoltaic renewable energy. Background Art

[0002] A multi-energy microgrid consists of renewable energy generators, controllable power sources, combined cooling, heating and power units, thermal energy storage systems and loads. It can be connected to the main grid through a substation or operate in an island mode. A multi-energy microgrid can coordinate the conversion and utilization of various energy sources to meet the power and heat load needs of users, playing an important role in improving energy efficiency and system economic benefits. However, with the gradual penetration of renewable energy sources such as photovoltaic panels and wind turbines, the high uncertainty and intermittency of renewable energy output make it challenging to maintain the supply and demand balance of multi-energy microgrids. Therefore, it is of great significance to study robust methods for achieving stable economic dispatch of multi-energy microgrids under uncertain conditions.

[0003] The key issue is how to use mathematical methods to characterize uncertain variables. Currently, there are two main methods for dealing with uncertainty: stochastic programming and robust optimization. Stochastic programming usually generates a series of random scenarios based on the probability distribution information of uncertainty for solution. However, it is difficult to estimate the probability distribution of uncertain quantities through multidimensional integrals, the computational complexity is high, and it is usually difficult to obtain the accurate probability distribution of uncertain quantities in practice. Robust optimization defines an uncertainty set to represent the fluctuation range of random variables, thereby eliminating the need for an accurate probability distribution of random variables. However, robust optimization looks for the worst case among all possible realizations within the uncertainty set, resulting in overly conservative decisions.

[0004] Distributed robust optimization combines the advantages of stochastic optimization and robust optimization. It does not require precise probability distribution, but uses a fuzzy set consisting of a series of probability distributions to characterize uncertainty. Distributed robust optimization constructs a fuzzy set of probability distribution of uncertain quantities through historical data of uncertain quantities, and searches for the worst-case probability distribution in this fuzzy set, rather than the worst-case scenario that may not occur. This can effectively address the problem of over-conservative robust optimization solutions. However, the traditional distributed robust optimization model directly uses the prediction results of uncertain quantities, so the accuracy of the prediction will affect the reliability of the optimization decision.

[0005] One way to solve the above problems is to incorporate more historical data and expand the scope of fuzzy sets. As the number of scenarios increases, the performance of the model may improve slightly. However, the attendant disadvantages may include a significant increase in solution time and the risk of overfitting. In order to reduce the complexity of the model and improve computational efficiency, the present invention uses a clustering algorithm to reduce the number of uncertainty scenarios and obtain typical scenarios that can reflect effective information about uncertainty. Dynamic time warping (DTW) can measure the similarity between time series and has good anti-interference ability. There is a lack of research on using DTW-based k-means time series clustering to obtain scenario probability information contained in uncertain quantities to drive distributed robust optimization of multi-energy microgrids. Summary of the invention

[0006] Aiming at the high uncertainty of renewable energy output in multi-energy microgrids, the present invention develops a robust optimization method for multi-energy microgrids that considers clustering information of uncertain scenarios. The multi-energy microgrid consists of photovoltaic panels, diesel generators, combined heat and power units, gas boilers, heat storage tanks and heating coils, and can trade electricity with the main power grid. The distributed robust optimization method includes a first-stage day-ahead optimization problem and a second-stage intraday optimization problem. The method is committed to characterizing the uncertainty of photovoltaic output with an improved fuzzy set based on Wasserstein distance, adding scene information of uncertain quantities to the fuzzy set; and then is committed to obtaining typical scene information of uncertain quantities in the fuzzy set using a k-means time series clustering algorithm based on DTW. The method of the present invention can significantly improve the computational efficiency of distributed robust optimization, effectively resist the fluctuation of photovoltaic output in multi-energy microgrids, while maintaining good economic benefits, and provide a reference for the planning, operation and scheduling of multi-energy microgrids.

[0007] In order to achieve the purpose of the present invention, the present invention provides a multi-energy microgrid robust optimization method considering uncertain scenario clustering information, comprising the following steps:

[0008] Step 1: Describe the architecture of the multi-energy microgrid; the multi-energy microgrid operates in grid-connected mode and trades electricity with the main grid. It consists of photovoltaic panels, diesel generators, combined heat and power units, gas boilers, heat storage tanks and heating coils; the combined heat and power units include micro gas turbines, heat recovery devices and absorption chillers; the micro gas turbine generates electricity by consuming natural gas, and the waste heat generated is recovered by the heat recovery device to generate heat energy; the gas boiler and heat storage tank are used to supplement the shortfall in heat demand; the gas boiler converts natural gas into heat energy; the heat storage tank stores or releases heat energy according to the heat load demand; the absorption chiller collects waste heat for evaporation and condensation to meet the cooling demand;

[0009] Step 2: Build a multi-energy microgrid model; establish a power generation device model, which includes a diesel generator model and a combined heat and power unit model; establish a power grid flow model, which includes an active power flow model, a reactive power flow model, a node voltage model, and a power grid security constraint model; establish a heat generation device model, which includes a heat recovery device model and a gas boiler model; establish a thermal energy storage system model, which includes a heat storage tank model; establish a production and consumption balance constraint model, which includes a thermal energy production and consumption balance model; establish a load model, which includes a heat load model and a cold load model;

[0010] Step 3: Propose a distributionally robust optimization method that considers the clustering information of uncertain scenarios. Establish a distributionally robust optimization model, which includes a first-stage day-ahead optimization problem and a second-stage intraday optimization problem; establish an uncertainty model, which is characterized by an improved fuzzy set based on Wasserstein distance, and the improved fuzzy set based on Wasserstein distance adds scene information of uncertainty, which is obtained by a k-means time series clustering algorithm based on DTW.

[0011] The multi-energy microgrid robust optimization method considering uncertain scenario clustering information, wherein, in step 2, the power generation device model includes a diesel generator model and a combined heating, cooling and power unit model;

[0012] The diesel generator model is:

[0013] (1)

[0014] (2)

[0015] (3)

[0016] in, For the Diesel generators during the period The binary start and stop state, Indicates opening, Indicates shutdown; For the Diesel generators during the period The meritorious efforts within, For the Diesel generators during the period The meritorious efforts within, For the Diesel generators during the period The reactive power output inside, For the Diesel generators during the period Apparent power within; For the The minimum active power generation of a diesel generator, For the Maximum active power generation of a diesel generator; For the The ramp rate of a diesel generator.

[0017] The multi-energy microgrid robust optimization method considering uncertain scenario clustering information, wherein the expression of the combined cooling, heating and power unit model is as follows:

[0018] (4)

[0019] (5)

[0020] (6)

[0021] in, For the The combined heat and power unit of Taiwan The meritorious contribution; For the The minimum active power generation of the combined heat and power unit of Taiwan Cooling, For the The maximum active power generation of the cooling, heating and power cogeneration unit; For the The ramp rate of the combined heating, cooling and power units; For the The micro gas turbine in the combined heat and power unit of Taiwan of natural gas-fired power generation; The energy conversion efficiency of the micro gas turbine; A scheduling interval.

[0022] The multi-energy microgrid robust optimization method considering uncertain scenario clustering information, wherein, in step 2, the power grid flow model includes an active power flow model, a reactive power flow model, a node voltage model and a power grid security constraint model;

[0023] (7)

[0024] (8)

[0025] (9)

[0026] (10)

[0027] in, For multi-energy microgrid in time period Purchase electricity from the main grid, For multi-energy microgrid in time period Sales of electricity to the main grid; For Node In the period Active power of For Node The crew is on time Auxiliary variables related to the active power generated; For Node In the period Active load demand; node Binary status of the upper diesel generator, For Node The binary status of the combined cooling, heating and power unit, For Node The binary state of the PV unit, Representation Node There are diesel generators on the Representation Node There are combined cooling, heating and power units on the Representation Node There are photovoltaic units on the Representation Node There are no diesel generators on the Representation Node There is no combined cooling, heating and power unit. Representation Node There are no photovoltaic units on the ground; For the PV units in the period The meritorious contribution; Indicates the serial number of the combined cooling, heating and power unit. Indicates the serial number of the photovoltaic unit. Indicates the serial number of the diesel generator.

[0028] The multi-energy microgrid robust optimization method considering uncertain scenario clustering information, wherein the expression of the reactive power flow model is:

[0029] (11)

[0030] (12)

[0031] in, For Node In the period The reactive power, For Node In the period Reactive power; For Node The crew is on time Auxiliary variables related to the reactive power generated; For Node In the period Reactive load demand; For the PV inverters in the period Reactive output.

[0032] The multi-energy microgrid robust optimization method considering uncertain scenario clustering information, wherein the expression of the node voltage model is:

[0033] (13)

[0034] in, For Node In the period Voltage; It is the voltage reference of the substation; For Node The resistance, For Node reactance.

[0035] The multi-energy microgrid robust optimization method considering uncertain scenario clustering information, wherein the expression of the grid security constraint model is as follows:

[0036] (14)

[0037] (15)

[0038] in, For Node In the period The apparent capacity of is the maximum allowable voltage deviation.

[0039] The multi-energy microgrid robust optimization method considering uncertain scene clustering information, wherein the heat recovery device model in step 2 is:

[0040] (16)

[0041] (17)

[0042] in, For the Heat recovery devices in the period Recovered heat energy; The energy conversion efficiency of the heat recovery device; For the Maximum heat output of a heat recovery unit.

[0043] The multi-energy microgrid robust optimization method considering uncertain scenario clustering information, wherein the expression of the gas boiler model in step 2 is as follows:

[0044] (18)

[0045] (19)

[0046] (20)

[0047] in, Respectively Gas boilers in the period heat production; Energy conversion efficiency of gas boilers; For the Gas boilers in the period of natural gas-fired electricity generation; For the Maximum heat output of a gas boiler; is the unit natural gas power generation; For the period Domestic natural gas usage; It is a collection of combined cooling, heating and power units.

[0048] The multi-energy microgrid robust optimization method considering uncertain scene clustering information, wherein the heat storage tank model in step 2 is:

[0049] (twenty one)

[0050] (twenty two)

[0051] (twenty three)

[0052] (twenty four)

[0053] (25)

[0054] in, For the The heat storage tank is in the period The binary hot charging state, For the The heat storage tank is in the period The binary exothermic state of For the The heat storage tank is in the period The heat charge, For the The heat storage tank is in the period The heat release; For the The maximum heat capacity of the heat storage tank is For the Maximum heat release of the heat storage tank; For the The heat storage tank is in the period The stored energy, For the The heat storage tank is in the period stored energy; For the The minimum storage capacity of a heat storage tank, For the Maximum storage capacity of a heat storage tank; The initial time of scheduling Storage capacity of heat storage tanks; The last time for scheduling Storage capacity of heat storage tanks; is the heat transfer coefficient of the heat storage tank; is the charging efficiency of the heat storage tank, is the heat release efficiency of the heat storage tank.

[0055] The multi-energy microgrid robust optimization method considering uncertain scene clustering information, wherein the production and consumption balance model expression of thermal energy in step 2 is as follows:

[0056] (26)

[0057] in, For the Heating coils in the period Heat absorbed; For the Absorption chillers in the period Absorbed heat.

[0058] The multi-energy microgrid robust optimization method considering uncertain scene clustering information, wherein the expressions of the heat load model and the cold load model in step 2 are as follows:

[0059] (27)

[0060] (28)

[0061] in, Heat load demand, is the cooling load demand; is the energy conversion coefficient of the heating coil; is the energy conversion coefficient of the absorption chiller.

[0062] The multi-energy microgrid robust optimization method considering uncertain scenario clustering information, wherein the objective function expression of the first stage day-ahead optimization problem in step 3 is as follows:

[0063] (29)

[0064] (30)

[0065] (31)

[0066] (32)

[0067] in, is the day-ahead decision variable; is the day-ahead operating cost; is the random variable of photovoltaic uncertainty; is the expected operating cost within the day; is the supremum function; The start-up and shutdown costs of diesel generators; is the operating cost of the heat storage tank; For the The start-stop cost coefficient of a diesel generator; For the Operation and maintenance cost coefficient of each heat storage tank; is the set of scheduling periods, For the collection of diesel generators.

[0068] The multi-energy microgrid robust optimization method considering uncertain scene clustering information, wherein the objective function and constraint condition expressions of the second-stage intraday optimization problem are as follows:

[0069] (33)

[0070] (34)

[0071] (35)

[0072] (36)

[0073] (37)

[0074] (38)

[0075] (39)

[0076] (40)

[0077] (41)

[0078] (42)

[0079] (43)

[0080] (44)

[0081] (45)

[0082] (46)

[0083] (47)

[0084] in, is the intraday objective function; The fuel cost for the multi-energy microgrid; The operation and maintenance cost of the units in the multi-energy microgrid; The electricity transaction cost of multi-energy microgrid; The cost factor for purchasing natural gas; is the fossil fuel cost factor; are the operation and maintenance cost coefficients of diesel generators respectively; is the operation and maintenance cost coefficient of the combined cooling, heating and power unit; is the operation and maintenance cost coefficient of the photovoltaic panel; For the period The purchase price of electricity; For the period The electricity selling price; For node 1 in period The active power, that is, the purchased power of the grid-connected node; For Node In the period Active power of A collection of photovoltaic units.

[0085] The multi-energy microgrid robust optimization method considering uncertain scenario clustering information, wherein in step 3, the uncertainty model is:

[0086] (48)

[0087] in, For the PV units in the period The output forecast value; For the PV units in the period Random prediction error of output;

[0088] The improved fuzzy set based on Wasserstein distance is:

[0089] (49)

[0090] (50)

[0091] in, It is an improved fuzzy set based on Wasserstein distance; is the support set of the uncertain quantity; is the forecast error random variable; is the Wasserstein radius auxiliary variable; represents the forecast error random variable and The joint distribution of represents the set of true distributions; The dimension is The set of real numbers; Indicates the total number of historical samples; represents auxiliary variables associated with typical scenarios of PV uncertainty, Represents the index of all typical scenes, Represents the index set of all typical scenes, Represents a typical scenario Auxiliary variables of Represents a typical scenario Wasserstein radius under ; Represents each typical scenario The following historical data; Represents each typical scenario Auxiliary variable of Wasserstein radius under ; Represents a typical scenario The prediction error random variable.

[0092] The multi-energy microgrid robust optimization method considering uncertain scene clustering information, wherein the typical scene information of the uncertain quantity is obtained by a k-means time series clustering algorithm based on DTW;

[0093] The DTW distance expression is:

[0094] (51)

[0095] in, and are any two PV output prediction error time series, for No. data, for No. data, which are mapped to construct Dynamically regularized matrix of ; Represents the DTW distance, which is the last element of the dynamic warping matrix; Indicates the dynamic distortion matrix Line List elements;

[0096] Based on the DTW k-means time series clustering algorithm, the objective function is to minimize the DTW distance between the photovoltaic output prediction errors, as shown below:

[0097] (52)

[0098] (53)

[0099] in, is the cluster center of each scene; The meaning of To achieve the minimum solution value;

[0100] Typical scenario information of uncertain quantity is:

[0101] (54)

[0102] (55)

[0103] in, is the total number of historical samples; For clustering The number of historical samples in ; For a typical scenario Probability of occurrence; For a typical scenario The corresponding Wasserstein radius.

[0104] Compared with the prior art, the present invention can achieve at least the following beneficial effects:

[0105] The present invention develops a robust optimization method for a multi-energy microgrid considering clustering information of uncertain scenarios. In terms of system architecture, the multi-energy microgrid in the present invention is composed of photovoltaic panels, diesel generators, combined heat and power units, gas boilers, heat storage tanks and heating coils, and can trade electricity with the main power grid. The combined heat and power units, photovoltaic panels and diesel generators work together to generate electricity to meet the demand for electrical loads; the heat recovery device, heat storage tank, gas boiler, heating coil and absorption refrigerator work together to meet the heating and cooling needs. In terms of scheduling methods, the present invention proposes a robust optimization method for a multi-energy microgrid considering clustering information of uncertain scenarios, uses a k-means time series clustering algorithm based on DTW to obtain typical scenario information of photovoltaic output, and constructs an improved fuzzy set based on Wasserstein distance to characterize photovoltaic uncertainty. Therefore, the present invention is suitable for multi-energy microgrids, helps to resist fluctuations in photovoltaic renewable energy output, takes into account the robustness and economy of multi-energy microgrid operation, significantly improves the computational efficiency of distributed robust optimization, and provides a reference for the planning, operation and scheduling of multi-energy microgrids. BRIEF DESCRIPTION OF THE DRAWINGS

[0106] Figure 1 The present invention provides a multi-energy microgrid system architecture.

[0107] Figure 2 The Wasserstein sphere of scene information is considered for the present invention. DETAILED DESCRIPTION

[0108] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0109] The present invention provides a multi-energy microgrid robust optimization method considering uncertain scenario clustering information, including:

[0110] Step 1: Describe the multi-energy microgrid architecture.

[0111] See also Figure 1 ,The present paper considers a multi-energy microgrid operating in grid-connected mode that can trade electricity with the main grid, which consists of photovoltaic panels, diesel generators, cogeneration units, gas boilers, heat storage tanks and heating coils.

[0112] Specifically, the combined heat and power unit usually includes a micro gas turbine, a heat recovery device and an absorption chiller. The micro gas turbine generates electricity by consuming natural gas, and the waste heat generated is recovered by the heat recovery device to generate thermal energy. Therefore, due to the coupling of electrical energy and thermal energy, the electrical power or heat output of the combined heat and power unit is mutually restricted. Under the operation strategy of determining heat by electricity, the combined heat and power unit generates electricity according to the power demand, which may lead to a mismatch between heat output and demand. When the heat generation is greater than the heat load, it may not be solved due to the equation constraint of heat balance. Gas boilers and heat storage tanks supplement the lack of heat demand. Gas boilers convert natural gas into thermal energy. The heat storage tank stores or releases thermal energy according to the heat load demand. In addition, the absorption chiller collects waste heat for evaporation and condensation to meet the cooling demand.

[0113] Here, the combined cooling, heating and power units, photovoltaic panels and diesel generators work together to generate electricity to meet the electrical load demand. At the same time, the heat recovery device, heat storage tank, gas boiler, heating coil and absorption chiller work together to meet the heating and cooling needs.

[0114] Step 2: Build a multi-energy microgrid model.

[0115] In some embodiments of the present invention, the multi-energy microgrid model includes a power generation device model, a power grid flow model, a heat generation device model, a thermal energy storage system model, a production and consumption balance constraint model, and a load model. The power generation device model includes a diesel generator model and a combined heat and power unit model; the power grid flow model includes an active power flow model, a reactive power flow model, a node voltage model, and a power grid security constraint model; the heat generation device model includes a heat recovery device model and a gas boiler model; the thermal energy storage system model includes a heat storage tank model; the production and consumption balance constraint model includes a thermal energy production and consumption balance model; and the load model includes a heat load model and a cold load model.

[0116] In the present invention, step 2 specifically includes the following sub-steps:

[0117] Step 2.1: Build a power generation device model.

[0118] In some embodiments of the present invention, the power generation device model includes a diesel generator model and a combined heat and power generation model. The combined heat and power generation unit includes a micro gas turbine, a heat recovery device and an absorption chiller, and operates under the strategy of determining heat by electricity. Due to the randomness, intermittency and volatility of renewable energy output, the multi-energy microgrid deploys diesel generators to ensure reliable power supply.

[0119] Diesel generators are controllable backup power sources for multi-energy microgrids. Their power generation costs are high and are accompanied by carbon emissions. When scheduling, it is necessary to take into account both economy and reliability to ensure stable power supply and reduce operating costs and environmental impacts. Formula (1) ensures that the active output of the diesel generator must be higher than the lower limit. The ramp limit of the active output of the diesel generator is shown in formula (2). The active power and reactive power output of the diesel generator are limited by the visible power capacity, as shown in formula (3).

[0120] (1)

[0121] (2)

[0122] (3)

[0123] in, For the Diesel generators during the period The binary start and stop state, Indicates opening, Indicates shutdown; For the Diesel generators during the period The meritorious efforts within, For the Diesel generators during the period The meritorious efforts within, For the Diesel generators during the period The reactive power output inside, For the Diesel generators during the period Apparent power within; For the The minimum active power generation of a diesel generator, For the Maximum active power generation of a diesel generator; For the The ramp rate of a diesel generator.

[0124] The combined heat and power unit includes a micro gas turbine, a heat recovery device and an absorption chiller. The micro gas turbine generates electricity by consuming natural gas, and the waste heat generated is recovered by the heat recovery device to generate heat energy. The absorption chiller collects the waste heat for evaporation and condensation to meet the cooling demand. Formula (4) defines the maximum and minimum active power generation of the combined heat and power unit. The ramp limit of the active output of the combined heat and power unit is shown in formula (5). The power generation efficiency of the combined heat and power unit, that is, the ratio of the power generation of the micro gas turbine to the total output of the combined heat and power unit (the power generation of the micro gas turbine plus the heat output of the heat recovery device), is expressed by formula (6).

[0125] (4)

[0126] (5)

[0127] (6)

[0128] in, For the The combined heat and power unit of Taiwan The meritorious contribution; For the The minimum active power generation of the combined heat and power unit of Taiwan Cooling, For the The maximum active power generation of the cooling, heating and power cogeneration unit; For the The ramp rate of the combined heating, cooling and power units; For the The micro gas turbine in the combined heat and power unit of Taiwan of natural gas-fired power generation; The energy conversion efficiency of the micro gas turbine; A scheduling interval.

[0129] Step 2.2: Establish a power grid flow model.

[0130] In the multi-energy microgrid proposed in the present invention, the power grid flow model includes an active power flow model, a reactive power flow model, a node voltage model and a power grid safety constraint model.

[0131] Active power flow is the steady-state distribution of active power in each branch. Equation (7) calculates the power from the substation, Equation (8) limits the transmission power between the multi-energy microgrid and the main grid, and Equation (9) follows the DistFlow linearized power flow model to model the active power of each node. In order to unify and simplify the model, auxiliary variables are used in Equation (10) to represent the active power generated by the unit.

[0132] (7)

[0133] (8)

[0134] (9)

[0135] (10)

[0136] in, For multi-energy microgrid in time period Purchase electricity from the main grid, For multi-energy microgrid in time period Sales of electricity to the main grid; For Node In the period Active power of For Node The crew is on time Auxiliary variables related to the active power generated; For Node In the period Active load demand; node Binary status of the upper diesel generator, For Node The binary status of the combined cooling, heating and power unit, For Node The binary state of the PV unit, Representation Node There are diesel generators on the Representation Node There are combined cooling, heating and power units on the Representation Node There are photovoltaic units on the Representation Node There are no diesel generators on the Representation Node There is no combined cooling, heating and power unit. Representation Node There are no photovoltaic units on the ground; For the PV units in the period The meritorious contribution; Indicates the serial number of the combined cooling, heating and power unit. Indicates the serial number of the photovoltaic unit. Indicates the serial number of the diesel generator.

[0137] Reactive power flow is the steady-state distribution of reactive power in each branch. Equation (11) follows the DistFlow linearized power flow model to model the reactive power of each node. In order to unify and simplify the model, auxiliary variables are used in Equation (12) to represent the reactive power generated by the unit.

[0138] (11)

[0139] (12)

[0140] in, For Node In the period The reactive power, For Node In the period Reactive power; For Node The crew is on time Auxiliary variables related to the reactive power generated; For Node In the period Reactive load demand; For the PV inverters in the period Reactive output.

[0141] The voltage model of each node is expressed by equation (13).

[0142] (13)

[0143] in, For Node In the period Voltage; It is the voltage reference of the substation; For Node The resistance, For Node reactance.

[0144] The grid security constraint model includes branch transmission capacity constraints and voltage deviation constraints. The network security constraint is expressed by (14) and the voltage amplitude constraint is expressed by (15).

[0145] (14)

[0146] (15)

[0147] in, For Node In the period The apparent capacity of is the maximum allowable voltage deviation.

[0148] Step 2.3: Build a heat generating device model.

[0149] In some embodiments of the present invention, the heat generating device model includes a heat recovery device model and a gas boiler model.

[0150] The heat recovery device recovers the waste heat generated by the micro gas turbine. The heat generation of the heat recovery device is calculated by equation (16), and equation (17) constrains the upper and lower bounds of the heat generation of the heat recovery device.

[0151] (16)

[0152] (17)

[0153] in, For the Heat recovery devices in the period Recovered heat energy; The energy conversion efficiency of the heat recovery device; For the Maximum heat output of a heat recovery unit.

[0154] The gas boiler converts natural gas into thermal energy to supplement the shortage of heat demand; the heat production of the gas boiler is calculated by formula (18), and the upper and lower bounds of the heat production of the gas boiler are constrained by formula (19). The total amount of natural gas consumed by the multi-energy microgrid is calculated in formula (20);

[0155] (18)

[0156] (19)

[0157] (20)

[0158] in, Respectively Gas boilers in the period heat production; Energy conversion efficiency of gas boilers; For the Gas boilers in the period of natural gas-fired electricity generation; For the Maximum heat output of a gas boiler; is the unit natural gas power generation; For the period Domestic natural gas usage; It is a collection of combined cooling, heating and power units.

[0159] Step 2.4: Build a thermal energy storage system model.

[0160] In some embodiments of the present invention, the thermal energy storage system model includes a thermal storage tank model.

[0161] The heat storage tank stores heat when there is excess heat and releases the stored heat when there is insufficient heat. Formula (21) uses binary variables to represent the heat charging and heat releasing states of the heat storage tank. Formula (22) limits the thermal power of the heat storage tank so that it does not exceed its capacity. The thermal energy storage capacity of the heat storage tank is calculated by formula (23). Formula (24) constrains the storage capacity of the heat storage tank. Formula (25) ensures that the capacity of the heat storage tank is equal at the start and end of the scheduling, which is helpful for cyclic scheduling.

[0162] (twenty one)

[0163] (twenty two)

[0164] (twenty three)

[0165] (twenty four)

[0166] (25)

[0167] in, For the The heat storage tank is in the period The binary hot charging state, For the The heat storage tank is in the period The binary exothermic state of For the The heat storage tank is in the period The heat charge, For the The heat storage tank is in the period The heat release; For the The maximum heat capacity of the heat storage tank is For the Maximum heat release of the heat storage tank; For the The heat storage tank is in the period The stored energy, For the The heat storage tank is in the period stored energy; For the The minimum storage capacity of a heat storage tank, For the Maximum storage capacity of a heat storage tank; The initial time of scheduling Storage capacity of heat storage tanks; The last time for scheduling Storage capacity of heat storage tanks; is the heat transfer coefficient of the heat storage tank; is the charging efficiency of the heat storage tank, is the heat release efficiency of the heat storage tank.

[0168] Step 2.5: Establish a production and consumption balance constraint model.

[0169] In some embodiments of the present invention, the production and consumption balance constraint model includes a production and consumption balance model of thermal energy. Formula (26) ensures the balance between the supply and demand of thermal energy.

[0170] (26)

[0171] in, For the Heating coils in the period Heat absorbed; For the Absorption chillers in the period Absorbed heat.

[0172] Step 2.6: Build load model.

[0173] In some embodiments of the present invention, the load of the multi-energy microgrid includes a thermal load model and a cooling load model.

[0174] Formula (27) shows that the heat from the heating coil needs to meet the heating load demand, and formula (28) shows that the heat recovered by the absorption chiller needs to meet the cooling load demand.

[0175] (27)

[0176] (28)

[0177] in, Heat load demand, is the cooling load demand; is the energy conversion coefficient of the heating coil; is the energy conversion coefficient of the absorption chiller.

[0178] Step 3: Propose a distributionally robust optimization method that considers clustering information of uncertain scenarios.

[0179] In some embodiments of the present invention, the distributed robust optimization model includes a first-stage day-ahead optimization problem and a second-stage intraday optimization problem, and the uncertainty model is characterized by an improved fuzzy set based on Wasserstein distance, and the improved fuzzy set based on Wasserstein distance adds scene information of uncertainty, and the scene information of uncertainty is obtained by a k-means time series clustering algorithm based on DTW.

[0180] In the present invention, step 3 specifically includes the following sub-steps:

[0181] Step 3.1: Formulate the first-stage day-ahead optimization problem.

[0182] The first stage is the day-ahead problem, which determines the state of the diesel generator and the decision of the heat storage tank. The goal is to minimize the day-ahead operating cost. and expected daily operating costs The sum of the cost of the day-ahead operation is shown in formula (29). Formula (30) represents the day-ahead operation cost The start-stop cost of diesel generators and the operating costs of the thermal storage tank Formula (31) calculates the start-stop cost of the diesel generator, and formula (32) calculates the operating cost of the heat storage tank.

[0183] (29)

[0184] (30)

[0185] (31)

[0186] (32)

[0187] in, is the day-ahead decision variable; is the day-ahead operating cost; is the random variable of photovoltaic uncertainty; is the expected operating cost within the day; is the supremum function; The start-up and shutdown costs of diesel generators; is the operating cost of the heat storage tank; For the The start-stop cost coefficient of a diesel generator; For the Operation and maintenance cost coefficient of each heat storage tank; is the set of scheduling periods, For the collection of diesel generators.

[0188] Step 3.2: Establish the second stage intraday optimization model.

[0189] After the uncertainty is realized, the second stage of the intraday problem determines the optimal pursuit decision, with the goal of minimizing the intraday operating cost. The intraday objective function under The expectation of represents the worst-case distribution of the photovoltaic uncertainty on the fuzzy set. is given and The cost function in this case is shown in formula (33). The intraday objective function includes fuel cost, unit operation and maintenance cost, and power transaction cost with the main grid. Formula (34) is the fuel cost, formula (35) is the unit operation and maintenance cost, and formula (36) is the power transaction cost between the multi-energy microgrid and the main grid.

[0190] Formula (37) is the diesel generator operation constraint, formula (38) is the combined heat and power unit operation constraint, formula (39) is the active power flow constraint, formula (40) is the reactive power flow constraint, formula (41) is the node voltage constraint, formula (42) is the power grid security constraint, formula (43) is the heat recovery device operation constraint, formula (44) is the gas boiler operation constraint, formula (45) is the heat storage tank operation constraint, formula (46) is the heat energy production and consumption balance constraint, and formula (47) is the heating and cooling load constraint.

[0191] (33)

[0192] (34)

[0193] (35)

[0194] (36)

[0195] (37)

[0196] (38)

[0197] (39)

[0198] (40)

[0199] (41)

[0200] (42)

[0201] (43)

[0202] (44)

[0203] (45)

[0204] (46)

[0205] (47)

[0206] in, is the intraday objective function; The fuel cost for the multi-energy microgrid; The operation and maintenance cost of the units in the multi-energy microgrid; The electricity transaction cost of multi-energy microgrid; The cost factor for purchasing natural gas; is the fossil fuel cost factor; are the operation and maintenance cost coefficients of diesel generators respectively; is the operation and maintenance cost coefficient of the combined cooling, heating and power unit; is the operation and maintenance cost coefficient of the photovoltaic panel; For the period The purchase price of electricity; For the period The electricity selling price; For node 1 in period The active power, that is, the purchased power of the grid-connected node; For Node In the period Active power of A collection of photovoltaic units.

[0207] Step 3.3: Establish the improved fuzzy set based on Wasserstein distance.

[0208] In some embodiments of the present invention, the two-stage distributed robust optimization model for multi-energy microgrid scheduling takes into account the uncertainty of photovoltaic power generation, as shown in formula (48).

[0209] (48)

[0210] in, For the PV units in the period The output forecast value; For the PV units in the period The random prediction error of the output.

[0211] The high uncertainty of photovoltaic output seriously affects the power balance of multi-energy microgrids. The present invention uses fuzzy sets based on Wasserstein probability distance to characterize the photovoltaic output prediction error. distribution. The Wasserstein fuzzy set is a sphere with the empirical distribution of the uncertainty as the center and the Wasserstein distance as the radius. It measures the distance between the empirical distribution and the actual distribution, described by the Euclidean norm. The key is to find the minimum radius so that the Wasserstein sphere contains as much real distribution as possible. However, traditional modeling methods ignore the scene information of photovoltaic uncertainty, and decisions may be difficult to accurately reflect actual needs. Therefore, the present invention proposes an improved fuzzy set based on Wasserstein distance, which uses clustering information to mine the potential correlation between prediction errors and typical scenarios, expressed as (49). The support set of the uncertainty can be expressed as formula (50), which restricts With its historical data In order to describe the deviation, the present invention adopts the 1-norm distance which is easy to linearize.

[0212] (49)

[0213] (50)

[0214] in, It is an improved fuzzy set based on Wasserstein distance; is the support set of the uncertain quantity; is the forecast error random variable; is the Wasserstein radius auxiliary variable; represents the forecast error random variable and The joint distribution of represents the set of true distributions; The dimension is The set of real numbers; Indicates the total number of historical samples; represents auxiliary variables associated with typical scenarios of PV uncertainty, Represents the index of all typical scenes, Represents the index set of all typical scenes, Represents a typical scenario Auxiliary variables of Represents a typical scenario Wasserstein radius under ; Represents each typical scenario The following historical data; Represents each typical scenario Auxiliary variable of Wasserstein radius under ; Represents a typical scenario The prediction error random variable.

[0215] The introduction of scenario variables classifies photovoltaic output into typical scenarios, each of which has a corresponding Wasserstein radius. and probability It is worth noting that when there is only one scenario, the fuzzy set It can be equivalent to the form of the classic Wasserstein fuzzy set, that is, no additional scene information is provided.

[0216] In order to construct a fuzzy set that considers scene information, we first need to obtain the scene parameters in the fuzzy set, that is, the probability and Wasserstein radius Therefore, we need to obtain a reasonable set of scenarios , so that the empirical distribution is closest to the real distribution in different scenarios. Since there are many scenarios and their statistical characteristics are similar, it is suitable to apply clustering algorithms for scene classification.

[0217] Step 3.4: Obtain the uncertainty scene clustering information in the fuzzy set.

[0218] K-means is a commonly used clustering algorithm that uses Euclidean distance as the basic metric. However, anomalies such as missing data may occur during energy data collection. In practice, the applicability of this metric is usually limited in cases where the data is time-shifted. In order to overcome the problem of time dimension misalignment, it is necessary to use a distance metric specifically for time series, such as DTW distance.

[0219] The DTW-based k-means time series clustering algorithm is widely used in data mining, information retrieval, and many other fields because it can handle nonlinear deformations of time series in the time dimension. It uses the DTW distance as a similarity measure, which improves the k-means time series clustering results. The core idea is to offset and scale any two time series in the time dimension to obtain the shortest regular path. Therefore, it searches for a meaningful average shape regardless of the temporal changes in the time series.

[0220] set up and are any two PV output forecast error time series. They are mapped to construct A dynamically regularized matrix. The DTW distance represented by is the last element of the dynamic warping matrix, and the remaining elements are calculated by formula (51).

[0221] (51)

[0222] in, and are any two PV output prediction error time series, for No. data, for No. data, which are mapped to construct Dynamically regularized matrix of ; Represents the DTW distance, which is the last element of the dynamic warping matrix; Indicates the dynamic distortion matrix Line Column elements.

[0223] Then, all time series are traversed and clustered, and the objective function is to minimize the DTW distance between the PV output prediction errors, as shown in Equation (52). Equation (53) represents the calculation method of the scene cluster center.

[0224] (52)

[0225] (53)

[0226] in, is the cluster center of each scene; The meaning of To achieve the minimum solution value.

[0227] By using the DTW-based k-means clustering algorithm, each cluster Aggregate similar shapes Historical data. Scenario probability It can be calculated based on the size of each cluster, as shown in formula (54). Correspondingly, the Wasserstein radius of each scene is Defined by formula (55).

[0228] (54)

[0229] (55)

[0230] in, is the total number of historical samples; For clustering The number of historical samples in ; For a typical scenario Probability of occurrence; For a typical scenario The corresponding Wasserstein radius.

[0231] The uncertainty is described according to the probability distribution and Wasserstein radius of each scene. The Wasserstein sphere considering the scene information is as follows: Figure 2 The introduction of scene clustering information divides historical data into three categories, each with a corresponding Wasserstein radius If the scene clustering information is not considered, the classic Wasserstein sphere is obtained ( Figure 2 The center of the sphere is based on the empirical distribution of historical data, and the radius is a single Wasserstein radius. As the scene scale increases, the range of the improved fuzziness set gradually approaches that of the classic fuzziness set.

[0232] Compared with traditional fuzzy sets, the improved fuzzy sets with scene clustering information have smaller Wasserstein radius. The improved Wasserstein sphere aims to avoid overly conservative worst-case distribution by learning scene information from historical data. , while also including the true distribution It covers the potential correlation between prediction errors and typical scenarios, which helps to improve the effective use of historical data and reduce the conservatism of the model. Therefore, random variables in multi-energy microgrid systems, such as the prediction error of photovoltaic output, can be classified according to the scenario clustering information to fully explore their correlation.

[0233] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A robust optimization method for multi-energy microgrids considering uncertain scenario clustering information, characterized in that: The following steps are involved: Step 1: Describe the architecture of the multi-energy microgrid; the multi-energy microgrid operates in grid-connected mode and trades electricity with the main grid. It consists of photovoltaic panels, diesel generators, combined heat and power units, gas boilers, heat storage tanks and heating coils; the combined heat and power units include micro gas turbines, heat recovery devices and absorption chillers; the micro gas turbine generates electricity by consuming natural gas, and the waste heat generated is recovered by the heat recovery device to generate heat energy; the gas boiler and heat storage tank are used to supplement the shortfall in heat demand; the gas boiler converts natural gas into heat energy; the heat storage tank stores or releases heat energy according to the heat load demand; the absorption chiller collects waste heat for evaporation and condensation to meet the cooling demand; Step 2: Build a multi-energy microgrid model; establish a power generation device model, which includes a diesel generator model and a combined heat and power unit model; establish a power grid flow model, which includes an active power flow model, a reactive power flow model, a node voltage model, and a power grid security constraint model; establish a heat generation device model, which includes a heat recovery device model and a gas boiler model; establish a thermal energy storage system model, which includes a heat storage tank model; establish a production and consumption balance constraint model, which includes a thermal energy production and consumption balance model; establish a load model, which includes a heat load model and a cold load model; Step 3: Propose a distributed robust optimization method considering the clustering information of uncertain scenarios; establish a distributed robust optimization model, which includes a first-stage day-ahead optimization problem and a second-stage intraday optimization problem; establish an uncertainty model, which is characterized by an improved fuzzy set based on Wasserstein distance, and the improved fuzzy set based on Wasserstein distance adds scene information of uncertain quantity, and the scene information of uncertain quantity is obtained by a k-means time series clustering algorithm based on DTW; In step 3, the uncertainty model is: （48） in, For the PV units in the period Output forecast value; For the PV units in the period Random prediction error of output; For the PV units in the period The meritorious contribution; The improved fuzzy set based on Wasserstein distance is: （49） （50） in, It is an improved fuzzy set based on Wasserstein distance; is the support set of the uncertain quantity; is the forecast error random variable; is the upper limit of the forecast error random variable, is the lower limit of the prediction error random variable; is the Wasserstein radius auxiliary variable; represents the forecast error random variable and The joint distribution of represents the set of true distributions; The dimension is The set of real numbers; Indicates the total number of historical samples; represents auxiliary variables associated with typical scenarios of PV uncertainty, Represents the index of all typical scenes, Represents the index set of all typical scenes, Represents a typical scenario Auxiliary variables of Represents a typical scenario Wasserstein radius under ; Represents each typical scenario The following historical data; Represents each typical scenario Auxiliary variable of Wasserstein radius under ; Represents a typical scenario The prediction error random variable of For a typical scenario Probability of occurrence; represents the expectation of the intraday forecast error random variable.

2. The multi-energy microgrid robust optimization method considering uncertain scenario clustering information according to claim 1 is characterized in that: In step 2, the power generation device model includes a diesel generator model and a combined cooling, heating and power unit model; The diesel generator model is: （1） （2） （3） in, For the Diesel generators during the period The binary start and stop state, Indicates opening, Indicates shutdown; For the Diesel generators during the period The meritorious efforts within, For the Diesel generators during the period The meritorious efforts within, For the Diesel generators during the period The reactive power output inside, For the Diesel generators during the period Apparent power within; For the The minimum active power generation of a diesel generator, For the Maximum active power generation of a diesel generator; For the The ramp rate of a diesel generator.

3. The multi-energy microgrid robust optimization method considering uncertain scenario clustering information according to claim 2 is characterized in that: The expression of the combined cooling, heating and power unit model is as follows: （4） （5） （6） in, For the The combined heat and power unit of Taiwan The meritorious contribution; For the The combined heat and power unit of Taiwan The meritorious contribution; For the The minimum active power generation of the combined heat and power unit of Taiwan Cooling, For the The maximum active power generation of the cooling, heating and power cogeneration unit; For the The ramp rate of the combined heating, cooling and power units; For the The micro gas turbine in the combined heat and power unit of Taiwan of natural gas-fired power generation; The energy conversion efficiency of the micro gas turbine; A scheduling interval.

4. The multi-energy microgrid robust optimization method considering uncertain scenario clustering information according to claim 3 is characterized in that: In step 2, the power grid flow model includes an active power flow model, a reactive power flow model, a node voltage model and a power grid security constraint model; The active power flow model is: （7） （8） （9） （10） in, For multi-energy microgrid in time period Purchase electricity from the main grid, For multi-energy microgrid in time period Sales of electricity to the main grid; For Node In the period Active power of For Node In the period t +1 active power; For Node The crew is on time Auxiliary variables related to the active power generated; For Node In the period Active load demand; node Binary status of the upper diesel generator, For Node The binary status of the upper CHP unit, For Node The binary state of the PV unit, Representation Node There are diesel generators on the Representation Node There are combined cooling, heating and power units on the Representation Node There are photovoltaic units on the Representation Node There are no diesel generators on the Representation Node There is no combined cooling, heating and power unit. Representation Node There are no photovoltaic units on the ground; Indicates the serial number of the combined cooling, heating and power unit. Indicates the serial number of the photovoltaic unit. Indicates the serial number of the diesel generator.

5. The multi-energy microgrid robust optimization method considering uncertain scenario clustering information according to claim 4 is characterized in that: The expression of the reactive power flow model is: （11） （12） in, For Node In the period The reactive power, For Node In the period Reactive power; For Node The crew is on time Auxiliary variables related to the reactive power generated; For Node In the period Reactive load demand; For the PV inverters in the period Reactive output.

6. The multi-energy microgrid robust optimization method considering uncertain scenario clustering information according to claim 5 is characterized in that: The expression of the node voltage model is: （13） in, For Node In the period Voltage; It is the voltage reference of the substation; For Node The resistance, For Node reactance.

7. The multi-energy microgrid robust optimization method considering uncertain scenario clustering information according to claim 6 is characterized in that: The expression of the power grid security constraint model is as follows: （14） （15） in, For Node In the period The apparent capacity of is the maximum allowable voltage deviation.

8. The multi-energy microgrid robust optimization method considering uncertain scenario clustering information according to claim 7 is characterized in that: In step 2, the heat recovery device model is: （16） （17） in, For the Heat recovery devices in the period Recovered heat energy; The energy conversion efficiency of the heat recovery device; For the Maximum heat output of a heat recovery unit.

9. The multi-energy microgrid robust optimization method considering uncertain scenario clustering information according to claim 8 is characterized in that: The expression of the gas boiler model in step 2 is as follows: （18） （19） （20） in, Respectively Gas boilers in the period heat production; Energy conversion efficiency of gas boilers; For the Gas boilers in the period of natural gas-fired electricity generation; For the Maximum heat output of a gas boiler; is the unit natural gas power generation; For the period Domestic natural gas usage; It is a collection of combined cooling, heating and power units.

10. The multi-energy microgrid robust optimization method considering uncertain scenario clustering information according to claim 9, characterized in that: In step 2, the heat storage tank model is: （21） （22） （23） （24） （25） in, For the The heat storage tank is in the period The binary hot charging state, For the The heat storage tank is in the period The binary exothermic state of For the The heat storage tank is in the period The heat charge, For the The heat storage tank is in the period The heat release; For the The maximum heat capacity of the heat storage tank is For the Maximum heat release of the heat storage tank; For the The heat storage tank is in the period The stored energy, For the The heat storage tank is in the period stored energy; For the The minimum storage capacity of a heat storage tank, For the Maximum storage capacity of a heat storage tank; The scheduling initial time Storage capacity of heat storage tanks; The last time for scheduling Storage capacity of heat storage tanks; is the heat transfer coefficient of the heat storage tank; is the charging efficiency of the heat storage tank, is the heat release efficiency of the heat storage tank.

11. The multi-energy microgrid robust optimization method considering uncertain scenario clustering information according to claim 10, characterized in that: The expression of the production and consumption balance model of thermal energy in step 2 is as follows: （26） in, For the Heating coils in the period Heat absorbed; For the Absorption chillers in the period Absorbed heat.

12. The multi-energy microgrid robust optimization method considering uncertain scenario clustering information according to claim 11, characterized in that: In step 2, the expressions of the heat load model and the cold load model are as follows: （27） （28） in, Heat load demand, is the cooling load demand; is the energy conversion coefficient of the heating coil; is the energy conversion coefficient of the absorption chiller.

13. The multi-energy microgrid robust optimization method considering uncertain scenario clustering information according to claim 12, characterized in that: In step 3, the objective function expression of the first stage day-ahead optimization problem is as follows: （29） （30） （31） （32） in, is the day-ahead decision variable; is the day-ahead operating cost; is the random variable of photovoltaic uncertainty; is the expected operating cost within the day; is the supremum function; The start-stop cost of the diesel generator; is the operating cost of the heat storage tank; For the The start-stop cost coefficient of a diesel generator; For the Operation and maintenance cost coefficient of each heat storage tank; is the set of scheduling periods, For the collection of diesel generators, is a fuzzy set based on Wasserstein distance; For the Diesel generators during the period t -1 binary start / stop state.

14. The multi-energy microgrid robust optimization method considering uncertain scenario clustering information according to claim 13, characterized in that: The objective function and constraint expressions of the second-stage intraday optimization problem are as follows: （33） （34） （35） （36) (37) (38) (39) (40) (41) (42) (43) (44) (45) (46) (47) in, is the intraday objective function; The fuel cost for the multi-energy microgrid; The operation and maintenance cost of the units in the multi-energy microgrid; The electricity transaction cost of multi-energy microgrid; The cost factor for purchasing natural gas; is the fossil fuel cost factor; are the operation and maintenance cost coefficients of diesel generators respectively; is the operation and maintenance cost coefficient of the combined cooling, heating and power unit; is the operation and maintenance cost coefficient of the photovoltaic panel; For the period The purchase price of electricity; For the period The electricity selling price; For node 1 in period The active power, that is, the purchased power of the grid-connected node; For Node In the period Active power of A collection of photovoltaic units.

15. The multi-energy microgrid robust optimization method considering uncertain scenario clustering information according to claim 14, characterized in that: The typical scene information of the uncertain quantity is obtained by a k-means time series clustering algorithm based on DTW; The DTW distance expression is: （51） in, and are any two PV output prediction error time series, for No. data, for No. data, which are mapped to construct Dynamically regularized matrix of ; Represents the DTW distance, which is the last element of the dynamic warping matrix; Indicates the dynamic distortion matrix Line List elements; Based on the DTW k-means time series clustering algorithm, the objective function is to minimize the DTW distance between the photovoltaic output prediction errors, as shown below: （52） （53） in, is the cluster center of each scene; The meaning of To achieve the minimum solution value; Represents each typical scenario Clustering data under Typical scenario information of uncertain quantity is: （54） （55） in, is the total number of historical samples; For clustering The number of historical samples in ; For a typical scenario The corresponding Wasserstein radius.

Images