Multi-microgrid system distributed real-time scheduling method and system

Abstract

The invention belongs to the related technical field of electrical engineering, and discloses a distributed real-time scheduling method and system for a multi-micro-grid system, and the method comprises the steps: building a centralized scheduling model of the multi-micro-grid system according to the technical parameters of all elements of the multi-micro-grid system and a Bellman optimality principle; taking each superior power grid node on the main line as a splitting node, splitting an alternating current power flow constraint in the centralized scheduling model by adopting an ADMM algorithm, and establishing a distributed scheduling model of the multi-microgrid system; in combination with an energy transaction mechanism, an optimization problem corresponding to the scheduling model of each micro-grid is further split into a superior power grid interaction sub-problem and a micro-grid internal optimization sub-problem, and a distributed scheduling model for enhancing privacy protection is obtained; and solving by adopting an ADP random optimization algorithm to obtain an optimal real-time scheduling decision of a full scheduling domain. By means of the method, distributed scheduling solving is achieved, the data privacy can be improved, and the calculation cost is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field related to electrical engineering, and more specifically, relates to a distributed real-time scheduling method and system for a multi-microgrid system. Background Art

[0002] With the development of distributed generation technology and demand-side response, microgrids have become an important form of efficient energy utilization. In practical applications, multiple microgrids with similar geographical locations can form a multi-microgrid system to further improve the safety and economy of system operation. However, the operation of multi-microgrids faces the influence of external physical information attacks and internal operation randomness (new energy generation, load, electricity price, etc.). Therefore, it is of great significance to study a decentralized real-time scheduling method for multi-microgrid systems, improve the privacy of multi-microgrid system operation, deal with randomness, and ensure the safe and economic operation of multi-microgrid systems.

[0003] As a stochastic optimization method, the approximate dynamic programming (ADP) method is widely used in the real-time scheduling of multi-microgrids. However, the ADP method is currently a centralized scheduling method. When applied to the decentralized scheduling of multi-microgrids, the centralized scheduling method faces the problems of high communication and computing costs, poor system privacy and reliability. Summary of the invention

[0004] In view of the above defects or improvement needs of the prior art, the present invention provides a decentralized real-time scheduling method and system for a multi-microgrid system, which aims to realize decentralized scheduling planning for multiple microgrids to improve privacy protection and reduce computing costs.

[0005] To achieve the above object, according to one aspect of the present invention, a decentralized real-time scheduling method for a multi-microgrid system is provided, which comprises:

[0006] S1. According to the technical parameters of each component of the multi-microgrid system and the Bellman optimality principle, a centralized dispatching model of the multi-microgrid system is established;

[0007] S2. Taking each upper grid node on the main line as a splitting node, the ADMM algorithm is used to split the AC power flow constraints in the centralized scheduling model, and a decentralized scheduling model of a multi-microgrid system is established. Different microgrids correspond to independent scheduling models. The upper grid nodes include external grid nodes, microgrid nodes, and intermediate connection nodes.

[0008] S3. Combined with the energy trading mechanism, the optimization problem corresponding to the dispatch model of each microgrid is further split into the upper grid interaction sub-problem and the microgrid internal optimization sub-problem, and a decentralized dispatch model with enhanced privacy protection is obtained;

[0009] S4. Use the ADP random optimization algorithm to solve the decentralized scheduling model with enhanced privacy protection to obtain the optimal real-time scheduling decision in the entire scheduling domain.

[0010] Optionally, in S1, the objective function of the centralized dispatching model of the multi-microgrid system is:

[0011] Where J represents the objective function, is the operating cost of the mth microgrid at time t, is the post-decision value function and post-decision time state variable of microgrid m at time t, p t is the electricity price at time t, is the active transaction power between multiple microgrids and the external grid at time t, m is the sequence number of the microgrid, M is the set of microgrids, is the power generation of gas generator set g at time t, a g 、b g 、c g is the power generation cost coefficient of gas-fired generator set g, G m is the set of gas generator sets in the mth microgrid, C w , C s is the penalty coefficient for abandoning wind and solar power, is the available power generation of wind turbine w and photovoltaic unit s at time t, is the actual power generation of wind turbine w and photovoltaic unit s at time t, W m , S m , B m is the set of wind turbines, photovoltaic units, and energy storage in the mth microgrid, It is the approximate value function related to the energy storage b at time t and the state of charge after the decision.

[0012] Optionally, in S1, the constraints of the centralized dispatching model of the multi-microgrid system include: AC power flow constraints, node voltage upper and lower limit constraints, gas turbine unit output upper and lower limit constraints, gas turbine unit climbing constraints, wind power output upper and lower limit constraints, photovoltaic output upper and lower limit constraints, multi-microgrid system and external power grid transaction power upper and lower limit constraints, energy storage operation constraints, gas power, wind power, photovoltaic, external power grid, and energy storage reactive output upper and lower limit constraints.

[0013] Optionally, in S2, when the ADMM algorithm is used to split the AC power flow constraints in the centralized scheduling model, the x update subproblem, y update subproblem, λ update problem and convergence judgment condition related to each upper grid node in the decentralized scheduling model of the multi-microgrid system are as follows:

[0014] x update subproblem:

[0015]

[0016] x t Update the decision variables of the subproblem for time x, y t Update the decision variable of the subproblem y at time t, λ t is the Lagrange multiplier, and is the decision variable of the λ-updating subproblem at time t;

[0017] is the augmented Lagrangian function in the ADMM algorithm at time t, is the feasible domain of the x-update subproblem in the ADMM algorithm at time t; the superscript ns is the number of iterations of the ADMM algorithm;

[0018] is the set of neighbor nodes of node i in the upper power grid node, is the total active and reactive power generation of the upper grid node i, ρ is a positive number close to 0, is the set of decision variables of the upper grid node i, for Replicated variables at node j;

[0019] is the total power generation cost of the upper grid node i, for the microgrid node, For external grid nodes, For the intermediate connection nodes,

[0020] is the square of the voltage and current at node i at time t, is the active and reactive power flow of line i formed by node i and its neighboring nodes at time t, is the lower and upper limits of the square voltage of node i, is the set of nodes in the upper power grid; is the internal constraint convex set of the node i subproblem. For the microgrid node, Includes: internal AC power flow constraints of microgrids, node voltage upper and lower limit constraints, gas turbine output upper and lower limit constraints, gas turbine ramp constraints, wind power output upper and lower limit constraints, photovoltaic output upper and lower limit constraints, energy storage operation constraints, gas power, wind power, photovoltaic, energy storage reactive output upper and lower limit constraints, for external grid nodes, Includes: upper and lower limits of active and reactive power transactions between multi-microgrid systems and external power grids, and for intermediate connection nodes: The active and reactive power of the node is required to be 0;

[0021] y update subproblem:

[0022]

[0023] in, is the feasible domain of the y update subproblem in the ADMM algorithm at time t, r i 、x i is the resistance and reactance of line i, is the collection of upper power grid lines, with superscript A i represents the parent node of node i, C i is the set of child nodes of node i;

[0024] λ Update Problem:

[0025]

[0026] Convergence judgment conditions:

[0027]

[0028] Among them, r t 、s t are the initial residual and dual residual at time t.

[0029] Optionally, in S3, the goal of solving the microgrid internal optimization subproblem is to obtain the total active and reactive output power of the entire microgrid, and its objective function is a subset of the objective function of the x-update subproblem, Contains only and The corresponding constraints include the microgrid internal constraints in the x-update subproblem;

[0030] The goal of solving the microgrid upper grid interaction subproblem is to optimize the communication between adjacent nodes to obtain power flow information, so as to satisfy the AC power flow constraints in a decentralized manner. Its objective function is a subset of the objective function of the x-update subproblem. Contains only The corresponding constraints include the AC power flow constraints and node voltage upper and lower limit constraints in the x-update subproblem.

[0031] Optionally, in S4, the ADP stochastic optimization algorithm uses a piecewise linear function to fit the value function in the Bellman optimality principle, and the proposed piecewise linear function expression is as follows:

[0032]

[0033] Where PW is the total number of segments of the piecewise linear function, is the slope of the piecewise linear function of energy storage b at the pw segment at time t, is the SOC value of energy storage b in the pw segment at time t, are the maximum and minimum charge states of energy storage b.

[0034] Optionally, in S4, before using the ADP random optimization algorithm to solve the scheduling model, the slope of the piecewise linear function in the ADP random optimization algorithm is first trained as follows:

[0035] S4-1, initialize the slope of the piecewise linear function and set the number of training times, let n = 1;

[0036] S4-2, based on the day-ahead forecast values ​​of wind power, photovoltaic power, load, and electricity price, a set of training scenarios are generated using the Monte Carlo method;

[0037] S4-3, let t = Δt;

[0038] S4-4, determine the time state variables of the multi-microgrid system at time t, solve the decentralized scheduling model with enhanced privacy protection, and obtain the decision at time t;

[0039] S4-5. Add a disturbance △SOC to the SOC of energy storage and re-solve the decentralized scheduling model with enhanced privacy protection;

[0040] S4-6. If the output of each component of the multi-microgrid system obtained by S4-5 is consistent with that obtained by S4-4, the slope of the value function is updated in the first way; if the output of each component of the multi-microgrid system obtained by S4-5 is different from that obtained by S4-4, the slope of the value function is updated in the second way; the first way is to calculate the sampled estimated value of the slope by the marginal influence of the energy storage SOC on the value function of the energy storage itself, and then update the slope; the second way is to calculate the sampled estimated value of the slope by decentralized calculation of the negative node marginal electricity price on the energy storage side of the microgrid, and then update the slope;

[0041] S4-7, calculating the state of the multi-microgrid system at the next moment;

[0042] S4-8, let t = t + Δt, repeat steps S4-4 to S4-7 until t = T;

[0043] S4-9, let n=n+1, repeat steps S4-2 to S4-8 until n=N;

[0044] S4-10. Output the slope of the trained piecewise linear function.

[0045] Optionally, in S4, the step of solving the decentralized scheduling model with enhanced privacy protection includes:

[0046] S2S3-1, initialize the state of the multi-microgrid system;

[0047] S2S3-2, for the x-update subproblem, solve the microgrid internal optimization problem and the upper grid interaction subproblem in sequence;

[0048] S2S3-3, solve the y update subproblem;

[0049] S2S3-4, solve the λ update problem;

[0050] S2S3-4: Perform convergence judgment. If the initial residual and the dual residual are not within the allowable error range, return to S2S3-2 to continue the distributed solution; if the initial residual and the dual residual are both within the allowable error range, output the distributed scheduling decision of multiple microgrids.

[0051] Optionally, in S4, the obtaining of the optimal real-time scheduling decision of the entire scheduling domain includes solving the optimal real-time scheduling decision at each moment in turn to form the optimal real-time scheduling decision of the entire scheduling domain;

[0052] Among them, the decision process at time t is:

[0053] Based on the real information of wind power, photovoltaic power, load and electricity price at time t, the real time state S of the multi-microgrid system at time t is determined. t tem ;

[0054] Determine the piecewise linear function according to the slope of the trained piecewise linear function;

[0055] Solve the decentralized scheduling model with enhanced privacy protection and obtain the optimal scheduling decision at time t.

[0056] According to another aspect of the present invention, there is provided a decentralized real-time dispatching system for a multi-microgrid system, comprising a memory and a processor, wherein the memory stores a computer program, wherein the processor implements the steps of any of the above methods when executing the computer program.

[0057] In general, compared with the prior art, the above technical solutions conceived by the present invention mainly have the following beneficial effects:

[0058] The present invention provides a decentralized real-time scheduling method for a multi-microgrid system. According to the technical parameters of each component of the multi-microgrid system and the Bellman optimality principle, a scheduling model for the multi-microgrid system is established; and based on the ADMM algorithm, the AC power flow constraints are split into multiple microgrids to establish a decentralized scheduling model for the multi-microgrid system; further combined with the energy trading mechanism, the optimization problem of each microgrid is split to strengthen privacy protection and improve computing efficiency; finally, the ADP algorithm is used to solve the decentralized scheduling model to obtain an approximate global optimal real-time scheduling solution for the multi-microgrid system. The present invention constructs a decentralized scheduling model and solves it. Compared with the centralized scheduling model, the decentralized scheduling model has higher data privacy protection and less computation.

[0059] Furthermore, the present invention also proposes a decentralized value function slope training method, which can embed the empirical information of the random factors of the entire system based only on the transaction energy information and the internal information of the microgrid, thereby strengthening privacy protection in ADP offline training. At the same time, the trained ADP value function can cope with randomness in real-time scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 A flow chart of a distributed real-time scheduling method for a multi-microgrid system provided by an embodiment of the present invention;

[0061] Figure 2 A structural diagram of a multi-microgrid system used in an embodiment of the present invention;

[0062] Figure 3 The output of wind power, photovoltaic power, load and electricity price in a multi-microgrid system used in an embodiment of the present invention in 24 hours, wherein (a) (b) (c) (d) correspond to wind power, photovoltaic power, load and electricity price respectively;

[0063] Figure 4 The convergence of the multi-microgrid decentralized dispatching model based on ADMM provided in the embodiment of the present invention;

[0064] Figure 5 The difference between the slope of the piecewise linear function obtained by the decentralized calculation of the ADP algorithm provided in the embodiment of the present invention and the slope obtained by the centralized calculation;

[0065] Figure 6 This is a comparison chart of the effects of the fully decentralized ADP algorithm (FD-ADP) and decentralized myopic (FD-Myopic) and decentralized model predictive control (FD-MPC) provided in the real-time example of the present invention. DETAILED DESCRIPTION

[0066] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0067] like Figure 1 The figure is a flow chart of a distributed real-time scheduling method for a multi-microgrid system provided by an embodiment of the present invention, and the steps thereof are described in detail below.

[0068] S1. According to the technical parameters of each component of the multi-microgrid system and the Bellman optimality principle, a centralized scheduling model of the multi-microgrid system is established.

[0069] Specifically, the technical parameters of each component include:

[0070] 1) Number of microgrid nodes N bus , the number of lines N of multiple microgrids branch , the node numbers at the beginning and end of the line, and the node range included in each microgrid;

[0071] 2) The node number where the gas-fired power plant is located and the upper and lower limits of the output of the gas-fired power plant and Gas-generator unit climbing rate and the cost coefficient a of the gas-fired generator set g 、b g and c g ;

[0072] 3) The node number of the external power grid, and the upper limit of active and reactive power exchange between the multi-microgrid system and the external power grid

[0073] 4) The nodes where wind turbines, loads and photovoltaic units are located, as well as the available output of wind turbines and photovoltaic units at the current moment and The load P and Q at the current moment;

[0074] 5) The node where the energy storage unit is located, the theoretical maximum charging and discharging power of the energy storage Upper and lower limits of energy storage unit capacity and And the initial capacity of the energy storage unit

[0075] like Figure 2 As shown, the structure diagram of the multi-microgrid system used in the embodiment of the present invention. In this embodiment, the upper and lower limits of the output of the gas-fired generator set are 60kW, 10kW (microgrid 1-microgrid 3) and 80kW, 20kW (microgrid 4), respectively. The ramp rates are 40kW / h (microgrid 1-microgrid 3) and 50kW / h (microgrid 4), respectively. The cost coefficients are 0.00012$ / kWh^2, 0.034$ / kWh, and 0.04$ (microgrid 1-microgrid 3). Grid 3) and 0.00008$ / kWh^2, 0.038$ / kWh, 0.08$ (microgrid 4); the wind turbine capacities of microgrids 1, 3, and 4 are 30kW, 40kW, and 40kW; the photovoltaic capacities of microgrids 2, 3, and 4 are 30kW, 30kW, and 40kW; the maximum charging and discharging power of the energy storage unit in microgrid 4 is 45kW, the efficiency is 0.95, and the initial, maximum, and minimum power are 90kWh, 180kWh, and 0kWh.

[0076] Specifically, the objective function of the centralized dispatching model of multi-microgrid system based on Bellman equation is:

[0077]

[0078] Where J represents the objective function, T is the dispatch domain of the multi-microgrid system, and C t is the operating cost of the multi-microgrid system at time t, are the state variables and decision variables of the system at time t, is the operating cost of the mth microgrid at time t, p t is the electricity price at time t, is the active transaction power between multiple microgrids and the external grid at time t, is the power generation of gas generator set g at time t, a g 、b g 、c g is the power generation cost coefficient of gas-fired generator set g, G m is the set of gas generator sets in the mth microgrid, C w , C s is the penalty coefficient for abandoning wind and solar power, is the available power generation of wind turbine w and photovoltaic unit s at time t, is the actual power generation of wind turbine w and photovoltaic unit s at time t, W m , S m , B m is the set of wind turbines, photovoltaic units, and energy storage in the mth microgrid, is the post-decision value function and post-decision time state variable of microgrid m at time t, It is the approximate value function related to the energy storage b at time t and the state of charge after the decision.

[0079] Specifically, the constraints of the centralized dispatching model of multi-microgrid systems include AC power flow constraints, node voltage upper and lower limit constraints, gas turbine output upper and lower limit constraints, gas turbine ramp constraints, wind power output upper and lower limit constraints, photovoltaic output upper and lower limit constraints, multi-microgrid system and external power grid transaction power upper and lower limit constraints, energy storage operation constraints, gas power, wind power, photovoltaic, external power grid, and energy storage reactive output upper and lower limit constraints.

[0080] AC power flow constraints include upper grid layer AC power flow constraints and microgrid layer AC power flow constraints:

[0081]

[0082] Node voltage upper and lower limit constraints (upper grid layer and microgrid layer):

[0083]

[0084] The upper and lower limits of gas-generator output and climbing constraints within the microgrid:

[0085]

[0086] in, is the downward climbing rate and upward climbing rate of thermal power unit g.

[0087] Upper and lower limits of wind power and photovoltaic output within the microgrid:

[0088]

[0089] The upper and lower limits of reactive power output of generator sets within the microgrid:

[0090]

[0091] in, is the reactive power of generator set o at time t. Generator set o includes thermal generator set g, wind generator set w, photovoltaic generator set s and energy storage generator set b. It is the lower and upper limits of the output of the generator set o.

[0092] Microgrid internal energy storage operation constraints:

[0093]

[0094] in, is the charging and discharging power of energy storage b at time t, is the charge and discharge efficiency of energy storage b, is the upper limit of charging and discharging power of energy storage b, is the initial charge state of the energy storage.

[0095] The upper and lower limits of active and reactive power trading between multi-microgrid system and external power grid:

[0096]

[0097] S2. Taking each upper-level grid node on the main line as a splitting node, the ADMM algorithm is used to split the AC power flow constraints in the centralized scheduling model, and a decentralized scheduling model of a multi-microgrid system is established. Different microgrids correspond to independent scheduling models. The upper-level grid nodes include external grid nodes, microgrid nodes, and intermediate connection nodes.

[0098] Specifically, Figure 2 For example, nodes 0, 1, 2, 3, 4, 5, and 6 are all upper-level power grid nodes.

[0099] The centralized dispatching model constructed in S1 is split into an external power grid model, each microgrid model and an intermediate connection node model, thereby splitting the centralized dispatching model into a decentralized dispatching model.

[0100] Since the AC power flow constraint is a continuous variable, the main purpose is to split the AC power flow constraint. The present invention adopts the ADMM algorithm for splitting. The core of the algorithm involves the construction of x update subproblem, y update subproblem, λ update problem and convergence judgment condition.

[0101] x update subproblem:

[0102]

[0103]

[0104] x t Update the decision variables of the subproblem for time x, y t Update the decision variable of the subproblem y at time t, λ t is the Lagrange multiplier, and is the decision variable of the λ-updating subproblem at time t;

[0105] is the augmented Lagrangian function in the ADMM algorithm at time t, is the feasible domain of the x-update subproblem in the ADMM algorithm at time t; the superscript ns is the number of iterations of the ADMM algorithm;

[0106] is the set of neighbor nodes of node i in the upper power grid node, is the total active and reactive power generation of the upper grid node i, ρ is a positive number close to 0, is the set of decision variables of the upper grid node i, for Replicated variables at node j;

[0107] is the total power generation cost of the upper grid node i, for the microgrid node, For external grid nodes, For the intermediate connection nodes,

[0108] is the square of the voltage and current at node i at time t, is the active and reactive power flow of line i formed by node i and its neighboring nodes at time t, is the lower and upper limits of the square voltage of node i, is the set of nodes in the upper power grid; is the internal constraint convex set of the node i subproblem. For the microgrid node, Includes: internal AC power flow constraints of microgrids, node voltage upper and lower limit constraints, gas turbine output upper and lower limit constraints, gas turbine ramp constraints, wind power output upper and lower limit constraints, photovoltaic output upper and lower limit constraints, energy storage operation constraints, gas power, wind power, photovoltaic, energy storage reactive output upper and lower limit constraints, for external grid nodes, Includes: upper and lower limits of active and reactive power transactions between multi-microgrid systems and external power grids, and for intermediate connection nodes: The active and reactive power of the node is required to be 0;

[0109] y update subproblem:

[0110]

[0111] in, is the feasible domain of the y update subproblem in the ADMM algorithm at time t, r i 、x i is the resistance and reactance of line i, is the collection of upper power grid lines, with superscript A i represents the parent node of node i, C i is the set of child nodes of node i;

[0112] λ Update Problem:

[0113]

[0114] Convergence judgment conditions:

[0115]

[0116] Among them, r t 、s t are the initial residual and dual residual at time t.

[0117] Through the ADMM algorithm, the centralized scheduling model can be initially split to form a decentralized scheduling model.

[0118] S3. Combined with the energy trading mechanism, the optimization problem corresponding to the dispatching model of each microgrid is further split into the upper-level grid interaction sub-problem and the microgrid internal optimization sub-problem, and a decentralized dispatching model with enhanced privacy protection is obtained.

[0119] Specifically, the goal of solving the microgrid internal optimization subproblem is to obtain the total active and reactive output power of the entire microgrid, and its objective function is a subset of the objective function of the x-update subproblem, Contains only and The corresponding constraints include the internal constraints of the microgrid in the x-update subproblem.

[0120] The internal optimization subproblem of the microgrid can be specifically expressed as:

[0121]

[0122] The goal of solving the microgrid parent grid interaction subproblem is to optimize the communication between adjacent nodes to obtain power flow information, so as to satisfy the AC power flow constraints in a decentralized manner. Its objective function is a subset of the objective function of the x-update subproblem. Contains only The corresponding constraints include the AC power flow constraints in the x-update subproblem.

[0123] The microgrid upper grid interaction subproblem can be specifically expressed as:

[0124]

[0125] Through this step, each microgrid model can be further split to strengthen data privacy protection.

[0126] S4. The ADP random optimization algorithm is used to solve the decentralized scheduling model with enhanced privacy protection and obtain the optimal real-time scheduling decision in the entire scheduling domain.

[0127] Specifically, after the splitting steps of S2 and S3, a decentralized scheduling model with very strong data privacy can be obtained. The present invention adopts the ADP random optimization algorithm for solving, and can obtain the optimal real-time scheduling decision of the entire scheduling domain.

[0128] In a specific embodiment, the ADP stochastic optimization algorithm uses a piecewise linear function to fit the value function in the Bellman optimality principle. The proposed piecewise linear function expression is as follows:

[0129]

[0130]

[0131] Where PW is the total number of segments of the piecewise linear function, is the slope of the piecewise linear function of energy storage b at the pw segment at time t, is the SOC value of energy storage b in the pw segment at time t, are the maximum and minimum charge states of energy storage b.

[0132] It is understandable that the slope of the piecewise linear function needs to be trained in advance, such as Figure 3 The figure shows the output of wind power, photovoltaic power, load and electricity price in a multi-microgrid system in 24 hours used for training the slope in an embodiment of the present invention. In this embodiment, the slope of the piecewise linear function in the ADP random optimization algorithm is trained as follows:

[0133] S4-1, initialize the slope of the piecewise linear function and set the number of training times, let n = 1;

[0134] S4-2, based on the day-ahead forecast values ​​of wind power, photovoltaic power, load, and electricity price, a set of training scenarios are generated using the Monte Carlo method;

[0135] S4-3, let t = Δt;

[0136] S4-4. Determine the time state variables of the multi-microgrid system at time t, solve the decentralized scheduling model with enhanced privacy protection, and obtain the decision at time t.

[0137] S4-5. Add a disturbance △SOC to the SOC of energy storage and re-solve the decentralized scheduling model with enhanced privacy protection;

[0138] S4-6. If the output of each component of the multi-microgrid system obtained by S4-5 is consistent with that obtained by S4-4, the slope of the value function is updated in the first way; if the output of each component of the multi-microgrid system obtained by S4-5 is different from that obtained by S4-4, the slope of the value function is updated in the second way; the first way is to calculate the sampled estimated value of the slope by the marginal influence of the energy storage SOC on the value function of the energy storage itself, and then update the slope; the second way is to calculate the sampled estimated value of the slope by decentralized calculation of the negative node marginal electricity price on the energy storage side of the microgrid, and then update the slope;

[0139] S4-7, calculating the state of the multi-microgrid system at the next moment;

[0140] S4-8, let t = t + Δt, repeat steps S4-4 to S4-7 until t = T;

[0141] S4-9, let n=n+1, repeat steps S4-2 to S4-8 until n=N;

[0142] S4-10. Output the slope of the trained piecewise linear function.

[0143] Among them, the first method is to calculate the sampled estimated value of the slope through the marginal influence of the energy storage SOC on the energy storage value function itself, which can be specifically expressed as:

[0144]

[0145] Among them, the superscript n is the number of training times of the ADP algorithm. is the estimated value of the slope sampled by the energy storage b at time t in the pw′ segment, where the segment pw′ represents the piecewise linear function The segment where α is located, and α is the iteration step size.

[0146] The second method is to calculate the sampled estimated value of the slope of the negative node marginal electricity price on the energy storage side of the microgrid in a decentralized manner, and then update the slope, which can be specifically expressed as:

[0147]

[0148] in, is the node marginal electricity price of microgrid energy storage node b at time t, is the set of nodes on the path from source node i to energy storage node b in microgrid m.

[0149] In one embodiment, in steps S4-4 and S4-5, the decentralized scheduling model for enhancing privacy protection is re-solved, specifically:

[0150] S2S3-1, initialize the state of the multi-microgrid system;

[0151] S2S3-2, for the x-update subproblem, solve the microgrid internal optimization problem and the upper grid interaction subproblem in sequence;

[0152] S2S3-3, solve the y update subproblem;

[0153] S2S3-4, solve the λ update problem;

[0154] S2S3-4: Perform convergence judgment. If the initial residual and the dual residual are not within the allowable error range, return to S2S3-2 to continue the distributed solution; if the initial residual and the dual residual are both within the allowable error range, output the distributed scheduling decision of multiple microgrids.

[0155] After the slope is trained, the scheduling decision can be solved at each moment, specifically:

[0156] Based on the real information of wind power, photovoltaic power, load and electricity price at time t, determine the real time state of the multi-microgrid system at time t According to the slope of the trained piecewise linear function, the multi-microgrid decentralized dispatch optimization framework based on ADMM is used to solve the optimal dispatch decision at time t. According to the real information of wind power, photovoltaic power, load and electricity price at the next moment, the real state of the multi-microgrid system at the next moment is obtained, and then the approximate optimal scheduling decision of the multi-microgrid system at the next moment is obtained; the solutions are solved in sequence to obtain the approximate optimal scheduling strategy of the multi-microgrid system in the entire scheduling domain.

[0157] Correspondingly, the present invention also protects a multi-microgrid system decentralized real-time dispatching system, comprising a memory and a processor, wherein the memory stores a computer program, wherein the processor implements the steps of the above method when executing the computer program.

[0158] Specifically, the system can be loaded on computing devices such as desktop computers, notebooks, PDAs and cloud servers. The processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The memory can be used to store computer programs and / or modules, and the processor can run or execute computer programs and / or modules stored in the memory, and call data stored in the memory, and various functions of the electronic device.

[0159] In order to prove the effectiveness of this scheme, the scheme is verified as follows, in which the method of this scheme is defined as a completely decentralized ADP algorithm.

[0160] like Figure 4 The figure shows the convergence of the multi-microgrid distributed scheduling model based on ADMM provided in an embodiment of the present invention. It can be seen from the figure that in the 24-hour scheduling, the ADMM algorithm can converge after very few iterations, and the difference between the objective function obtained by distributed calculation and the objective function of centralized scheduling is on the order of 1e-5, indicating that the multi-microgrid distributed scheduling model based on ADMM can obtain centralized global optimal scheduling only by relying on local information.

[0161] Figure 5 The difference between the slope of the piecewise linear function calculated by the fully decentralized ADP algorithm provided in the embodiment of the present invention and the slope obtained by the centralized calculation. It can be seen from the figure that the error between the slope obtained by the decentralized calculation and the slope obtained by the centralized calculation is on the order of 1e-4, indicating that the decentralized slope update method in the fully decentralized ADP algorithm can achieve the same effect as the centralized update.

[0162] Figure 6 The comparison diagram of the effects of the fully decentralized ADP algorithm (FD-ADP) and the decentralized myopic (FD-Myopic) and decentralized model predictive control (FD-MPC) provided for the real-time example of the present invention shows that the optimization error of the FD-ADP algorithm is the smallest, and the error order of magnitude is 1e-3, indicating that the fully decentralized ADP algorithm can obtain the approximate optimal scheduling decision in the entire scheduling domain.

[0163] In general, the present invention provides a decentralized real-time scheduling method for a multi-microgrid system. According to the technical parameters of each component of the multi-microgrid system and the Bellman optimality principle, a scheduling model of the multi-microgrid system is established; and based on the ADMM algorithm, the AC power flow constraints are split into multiple microgrids to establish a decentralized scheduling model for the multi-microgrid system; further combined with the energy trading mechanism, the optimization problem of each microgrid is split to strengthen privacy protection and improve computing efficiency; finally, a fully decentralized ADP algorithm is proposed to solve the scheduling model, a piecewise linear function approximates the value function, an ADP decentralized value function slope update method is designed, and the trained decentralized value function is applied to real-time scheduling to assist in obtaining an approximate global optimal real-time scheduling scheme for the multi-microgrid system.

[0164] Moreover, the decentralized real-time scheduling method for multi-microgrid systems provided by the present invention can take into account the influence of AC power flow constraints, and achieve the optimal operation of the multi-microgrid system in a decentralized manner through the interactive trading of energy information between each microgrid, and has the advantages of privacy protection and convenient calculation.

[0165] Furthermore, the present invention provides a decentralized real-time scheduling method for a multi-microgrid system, which derives the physical meaning of the slope of the ADP piecewise linear value function under AC power flow constraints, and designs a decentralized value function slope update method, which can embed the empirical information of the random factors of the entire system based only on transaction energy information and microgrid internal information, and strengthen privacy protection in ADP offline training. At the same time, the trained ADP value function can cope with randomness in real-time scheduling.

[0166] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification. It should be noted that "in one embodiment", "for example", "for example", etc. of the present invention are intended to illustrate the present invention, rather than to limit the present invention.

[0167] The above embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the patent application. It should be pointed out that, for those of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention.

Claims

1. A distributed real-time scheduling method for a multi-microgrid system, characterized in that: include: S1. According to the technical parameters of each component of the multi-microgrid system and the Bellman optimality principle, a centralized dispatching model of the multi-microgrid system is established; S2. Taking each upper grid node on the main line as a splitting node, the ADMM algorithm is used to split the AC power flow constraints in the centralized scheduling model, and a decentralized scheduling model of a multi-microgrid system is established. Different microgrids correspond to independent scheduling models. The upper grid nodes include external grid nodes, microgrid nodes, and intermediate connection nodes. S3. Combined with the energy trading mechanism, the optimization problem corresponding to the dispatch model of each microgrid is further split into the upper grid interaction sub-problem and the microgrid internal optimization sub-problem, and a decentralized dispatch model with enhanced privacy protection is obtained; S4. Use the ADP random optimization algorithm to solve the decentralized scheduling model with enhanced privacy protection to obtain the optimal real-time scheduling decision in the entire scheduling domain.

2. The distributed real-time dispatching method for a multi-microgrid system according to claim 1, characterized in that: In S1, the objective function of the centralized dispatching model of the multi-microgrid system is: Where J represents the objective function, is the operating cost of the mth microgrid at time t, V t m,a , is the post-decision value function and post-decision time state variable of microgrid m at time t, p t is the electricity price at time t, is the active transaction power between multiple microgrids and the external grid at time t, m is the sequence number of the microgrid, M is the set of microgrids, is the power generation of gas generator set g at time t, a g 、b g 、c g is the power generation cost coefficient of gas-fired generator set g, G m is the set of gas generator sets in the mth microgrid, C w , C s is the penalty coefficient for abandoning wind and solar power, is the available power generation of wind turbine w and photovoltaic unit s at time t, is the actual power generation of wind turbine w and photovoltaic unit s at time t, W m , S m , B m is the set of wind turbines, photovoltaic units, and energy storage in the mth microgrid, It is the approximate value function related to the energy storage b at time t and the state of charge after the decision.

3. The distributed real-time dispatching method for a multi-microgrid system according to claim 1, characterized in that: In S1, the constraints of the centralized dispatching model of the multi-microgrid system include: AC power flow constraints, node voltage upper and lower limit constraints, gas turbine output upper and lower limit constraints, gas turbine climbing constraints, wind power output upper and lower limit constraints, photovoltaic output upper and lower limit constraints, multi-microgrid system and external power grid transaction power upper and lower limit constraints, energy storage operation constraints, gas power, wind power, photovoltaic, external power grid, and energy storage reactive output upper and lower limit constraints.

4. The distributed real-time dispatching method for a multi-microgrid system according to claim 2, characterized in that: In S2, when the ADMM algorithm is used to split the AC power flow constraints in the centralized scheduling model, the x-update subproblem, y-update subproblem, λ-update problem and convergence judgment conditions related to each upper-level grid node in the decentralized scheduling model of the multi-microgrid system are as follows: x update subproblem: x t Update the decision variables of the subproblem for time x, y t Update the decision variable of the subproblem y at time t, λ t is the Lagrange multiplier, which is the decision variable of the λ update subproblem at time t; is the augmented Lagrangian function in the ADMM algorithm at time t, is the feasible domain of the x-update subproblem in the ADMM algorithm at time t; the superscript ns is the number of iterations of the ADMM algorithm; is the set of neighbor nodes of node i in the upper power grid node, is the total active and reactive power generation of the upper grid node i, ρ is a positive number close to 0, is the set of decision variables of the upper grid node i, for Replicated variables at node j; is the total power generation cost of the upper grid node i, for the microgrid node, For external grid nodes, For the intermediate connection nodes, is the square of the voltage and current at node i at time t, is the active and reactive power flow of line i formed by node i and its neighboring nodes at time t, is the lower and upper limits of the square of the voltage at node i, is the set of nodes in the upper power grid; is the internal constraint convex set of the node i subproblem. For the microgrid node, Includes: internal AC power flow constraints of microgrids, node voltage upper and lower limit constraints, gas turbine output upper and lower limit constraints, gas turbine ramp constraints, wind power output upper and lower limit constraints, photovoltaic output upper and lower limit constraints, energy storage operation constraints, gas power, wind power, photovoltaic, energy storage reactive output upper and lower limit constraints, for external grid nodes, Includes: upper and lower limits of active and reactive power transactions between multi-microgrid systems and external power grids, and for intermediate connection nodes: The active and reactive power of the node is required to be 0; y update subproblem: in, is the feasible domain of the y update subproblem in the ADMM algorithm at time t, r i 、x i is the resistance and reactance of line i, is the collection of upper power grid lines, with superscript A i represents the parent node of node i, C i is the set of child nodes of node i; λ Update Problem: Convergence judgment conditions: Among them, r t 、s t are the initial residual and dual residual at time t.

5. The distributed real-time dispatching method for a multi-microgrid system according to claim 4, characterized in that: In S3, the goal of solving the microgrid internal optimization subproblem is to obtain the total active and reactive output power of the entire microgrid, and its objective function is a subset of the objective function of the x-update subproblem, Contains only and The corresponding constraints include the microgrid internal constraints in the x-update subproblem; The goal of solving the microgrid upper grid interaction subproblem is to optimize the communication between adjacent nodes to obtain power flow information, so as to satisfy the AC power flow constraints in a decentralized manner. Its objective function is a subset of the objective function of the x-update subproblem. Contains only The corresponding constraints include the AC power flow constraints and node voltage upper and lower limit constraints in the x-update subproblem.

6. The distributed real-time dispatching method for a multi-microgrid system according to claim 2, characterized in that: In S4, the ADP stochastic optimization algorithm uses a piecewise linear function to fit the value function in the Bellman optimality principle. The proposed piecewise linear function expression is as follows: Where PW is the total number of segments of the piecewise linear function, is the slope of the piecewise linear function of energy storage b at the pw segment at time t, is the SOC value of energy storage b in the pw segment at time t, are the maximum and minimum charge states of energy storage b.

7. The distributed real-time dispatching method for a multi-microgrid system according to claim 6, characterized in that: In S4, before using the ADP random optimization algorithm to solve the scheduling model, the slope of the piecewise linear function in the ADP random optimization algorithm is trained as follows: S4-1, initialize the slope of the piecewise linear function and set the number of training times, let n = 1; S4-2, based on the day-ahead forecast values ​​of wind power, photovoltaic power, load, and electricity price, a set of training scenarios are generated using the Monte Carlo method; S4-3, let t = Δt; S4-4, determine the time state variables of the multi-microgrid system at time t, solve the decentralized scheduling model with enhanced privacy protection, and obtain the decision at time t; S4-5. Add a disturbance △SOC to the SOC of energy storage and re-solve the decentralized scheduling model with enhanced privacy protection; S4-6. If the output of each component of the multi-microgrid system obtained by S4-5 is consistent with that obtained by S4-4, the slope of the value function is updated in the first way; if the output of each component of the multi-microgrid system obtained by S4-5 is different from that obtained by S4-4, the slope of the value function is updated in the second way; the first way is to calculate the sampled estimated value of the slope by the marginal influence of the energy storage SOC on the value function of the energy storage itself, and then update the slope; the second way is to calculate the sampled estimated value of the slope by decentralized calculation of the negative node marginal electricity price on the energy storage side of the microgrid, and then update the slope; S4-7, calculating the state of the multi-microgrid system at the next moment; S4-8, let t = t + Δt, repeat steps S4-4 to S4-7 until t = T; S4-9, let n=n+1, repeat steps S4-2 to S4-8 until n=N; S4-10. Output the slope of the trained piecewise linear function.

8. The distributed real-time dispatching method for a multi-microgrid system according to claim 7, characterized in that: In S4, the steps of solving the decentralized scheduling model with enhanced privacy protection include: S2S3-1, initialize the state of the multi-microgrid system; S2S3-2, for the x-update subproblem, solve the microgrid internal optimization problem and the upper grid interaction subproblem in sequence; S2S3-3, solve the y update subproblem; S2S3-4, solve the λ update problem; S2S3-4: Perform convergence judgment. If the initial residual and the dual residual are not within the allowable error range, return to S2S3-2 to continue the distributed solution; if the initial residual and the dual residual are both within the allowable error range, output the distributed scheduling decision of multiple microgrids.

9. The distributed real-time dispatching method for a multi-microgrid system according to claim 1, characterized in that: In S4, the obtaining of the optimal real-time scheduling decision of the entire scheduling domain includes solving the optimal real-time scheduling decision at each moment in turn to form the optimal real-time scheduling decision of the entire scheduling domain; Among them, the decision process at time t is: Based on the real information of wind power, photovoltaic power, load and electricity price at time t, determine the real time state of the multi-microgrid system at time t Determine the piecewise linear function according to the slope of the trained piecewise linear function; Solve the decentralized scheduling model with enhanced privacy protection and obtain the optimal scheduling decision at time t.

10. A distributed real-time dispatching system for a multi-microgrid system, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 9 are implemented.

Images