A multi-energy coupling power distribution network fault recovery method, system, device and medium

Abstract

A multi-energy coupling power distribution network fault recovery method, system, device and medium, comprising: acquiring parameters of a power distribution network layer and parameters of an energy subsystem of a multi-energy coupling power distribution network; bringing the parameters of the power distribution network layer and the parameters of the energy subsystem into a pre-constructed double-layer optimization model, solving by using a deep reinforcement learning algorithm to obtain a loss load power evaluation index and a target function value; the fault recovery scheme corresponding to the loss load power evaluation index and the target function value is taken as the fault recovery scheme of the multi-energy coupling power distribution network; the deep reinforcement learning model is an upper layer optimization model constructed based on the parameters of the power distribution network layer and optimization criteria, and a lower layer optimization model constructed based on the parameters of the energy subsystem and optimization criteria combined with a deep learning algorithm. The present application adopts a double-layer optimization model and solves by combining a deep reinforcement learning algorithm, improves the optimization efficiency, and overcomes the problem of low optimization efficiency of the prior art which only processes a single-layer model.

Description

Technical Field

[0001] The present invention relates to the field of distribution network fault recovery, and in particular to a multi-energy coupling distribution network fault recovery method, system, equipment and medium. Background Art

[0002] Against the backdrop of the comprehensive advancement of the Energy Internet, interconnected energy distribution networks have become a key development direction for smart distribution networks. Distribution network power facilities are often exposed to the natural environment and are highly susceptible to low-probability, high-loss events, causing failures in lines, switchgear, and distribution and transformation equipment, leading to power outages. Furthermore, the power system exhibits a multi-energy coupling pattern of source-grid-load-storage. There is an urgent need to break away from the traditional top-down fault recovery power supply path and prioritize the restoration of power to critical loads through distributed resources, achieving rapid, coordinated, and bidirectional recovery across the main and distribution networks. Therefore, rapid fault recovery is a must-have function for distribution networks, and multi-energy coupled distribution network fault recovery technology is one of the challenges that must be overcome.

[0003] Current multi-energy coupled distribution network fault recovery technologies have the following shortcomings: Current distribution network fault recovery strategies first perform overall optimization at the distribution network level, then optimize the underlying device layers after obtaining the overall optimization results. This serial recovery strategy can only handle the optimization problem of a single model per unit time, resulting in low optimization efficiency. When a large number of optimization parameters are involved, the real-time performance of the distribution network fault recovery strategy generation is severely affected. Summary of the Invention

[0004] In order to solve the problem that the existing technology uses a serial recovery strategy that can only process the optimization problem of a single-layer model per unit time, the optimization efficiency is low, and when there are many optimization parameters involved, the real-time performance of the distribution network fault recovery strategy generation will be seriously affected. The present invention proposes a multi-energy coupling distribution network fault recovery method, including:

[0005] Obtain the parameters of the distribution layer and energy subsystem of the multi-energy coupled distribution network;

[0006] The parameters of the distribution network layer and the parameters of the energy subsystem are introduced into a pre-built two-layer optimization model, and a deep reinforcement learning algorithm is used to solve the problem to obtain the load loss power evaluation index and the objective function value;

[0007] The parameters of the distribution network layer and the energy subsystem corresponding to the load loss power evaluation index and the objective function value are used as a fault recovery plan for the multi-energy coupling distribution network;

[0008] Among them, the deep reinforcement learning model is constructed based on the upper-level optimization model constructed based on the parameters and optimization criteria of the distribution network layer, and the lower-level optimization model constructed based on the parameters and optimization criteria of the energy subsystem combined with the deep learning algorithm.

[0009] Optionally, the construction of the two-layer optimization model includes:

[0010] The upper layer optimization model is constructed based on the Markov state of the distribution network layer, the interruption action of the distribution network layer switch, and the minimum load loss power within the minimum fault time as the distribution network layer optimization criterion;

[0011] The lower-level optimization model is constructed based on the Markov state of the energy subsystem, the interruption action of the energy subsystem switch, and the minimum power recovery of the multi-energy coupling distribution network as the optimization criterion of the energy subsystem;

[0012] The upper-layer optimization model and the lower-layer optimization model are combined to construct a deep reinforcement learning model.

[0013] Optionally, the parameters of the network distribution layer include:

[0014] Decision-making on the interruption power of each load node at each moment, the number of load nodes, and the switch interruption action of each load node at each decision moment;

[0015] The parameters of the energy subsystem include: power generation value, power purchase and sale to the upper power grid, natural gas system exchange power, operating power of traditional generator sets and power variation.

[0016] Optionally, the parameters of the distribution network layer and the parameters of the energy subsystem are brought into a pre-built two-layer optimization model, and a deep reinforcement learning algorithm is used to solve the model to obtain a load loss power evaluation index and an objective function value, including:

[0017] S1 initializes the parameters of the two-layer optimization model and the parameters of the deep learning algorithm;

[0018] S2 selects the distribution network switch adjustment state behavior based on the Markov state of the upper optimization model and the Markov behavior selection probability distribution table of the upper optimization model to form a fault recovery plan, and transmits the fault recovery plan to the lower optimization model;

[0019] S3: selecting a Markov decision behavior action of the lower-layer optimization model based on the fault recovery solution and the Markov behavior selection probability distribution table of the lower-layer optimization model to form the power output of each device of the access node;

[0020] S4 calculates the load loss power evaluation index for executing the decision-making behavior of the upper-layer optimization model in the current round based on the fault recovery plan, the attenuation factor of the optimization criterion of the upper-layer optimization model and the distribution network layer optimization criterion;

[0021] S5 calculates the objective function value of the decision-making behavior of the lower-level optimization model in the current round based on the power output of each device of the node and the attenuation factor of the optimization criterion of the lower-level optimization model in combination with the optimization criterion of the lower-level optimization model;

[0022] S6: updating the selection probability distribution table of the lower-level optimization model based on the classic roulette wheel selection method, and using the updated selection probability distribution table of the lower-level optimization model as the selection probability distribution table of the lower-level optimization model; and updating the selection probability distribution table of the upper-level optimization model based on the classic roulette wheel selection method, and using the updated selection probability distribution table of the upper-level optimization model as the selection probability distribution table of the upper-level optimization model;

[0023] S7 determines whether the current round has reached the maximum iteration round of deep learning. If not, the current round is incremented by 1 and returned to S2. Otherwise, the load-loss power evaluation index and objective function value of the current round are output, and the optimization decision process ends.

[0024] The deep learning algorithm parameters include: the maximum iteration round of deep reinforcement learning, the Markov behavior selection probability distribution table of the upper optimization model and the attenuation factor of the optimization criterion, the Markov behavior selection probability distribution table of the lower optimization model and the attenuation factor value of the optimization criterion;

[0025] The parameters of the two-layer optimization model include: distribution network operation parameters, distribution network topology parameters, distribution network switch state parameters, and equipment operation state parameters of each energy subsystem of the multi-energy coupling distribution network.

[0026] Optionally, the optimization criterion of the upper-layer optimization model is as follows:

[0027]

[0028] Where V cl It is the load-loss power evaluation index; is the node importance weight; is the interruption action of the switch at the i-th load node at the decision time k; is the interruption power of the i-th load node at the decision time k; n mg is the total amount of load nodes; T k is the kth moment; k is the time; i is the device number.

[0029] Optionally, the optimization criterion of the lower optimization model is calculated as follows:

[0030]

[0031] Where V me The lower layer objective function value for multi-energy coupled distribution network fault recovery; T kis the kth moment; k is the moment; N is the number of generator sets; is the operating cost of equipment i; is the status of device i; is the action of device i.

[0032] In another aspect, the present invention further provides a multi-energy coupled distribution network fault recovery system, comprising:

[0033] An acquisition module is used to obtain parameters of the distribution network layer and energy subsystem of the multi-energy coupling distribution network;

[0034] A calculation module is used to bring the parameters of the distribution network layer and the parameters of the energy subsystem into a pre-built two-layer optimization model, and solve it using a deep reinforcement learning algorithm to obtain a load loss power evaluation index and an objective function value;

[0035] A solution determination module is used to use the parameters of the distribution network layer and the energy subsystem corresponding to the load loss power evaluation index and the objective function value as a fault recovery solution for the multi-energy coupling distribution network;

[0036] Among them, the deep reinforcement learning model is constructed based on the upper-level optimization model constructed based on the parameters and optimization criteria of the distribution network layer, and the lower-level optimization model constructed based on the parameters and optimization criteria of the energy subsystem combined with the deep learning algorithm.

[0037] Optionally, a two-layer optimization model building module is also included, which is used to build an upper-layer optimization model based on the distribution network layer Markov state, the distribution network layer switch interruption action, and the minimum load loss power within the minimization of the fault time as the distribution network layer optimization criterion;

[0038] The lower-level optimization model is constructed based on the Markov state of the energy subsystem, the interruption action of the energy subsystem switch, and the minimum power recovery of the multi-energy coupling distribution network as the optimization criterion of the energy subsystem;

[0039] The upper-layer optimization model and the lower-layer optimization model are combined to construct a deep reinforcement learning model.

[0040] Optionally, the calculation module is specifically used to:

[0041] S1 initializes the parameters of the two-layer optimization model and the parameters of the deep learning algorithm;

[0042] S2 selects the distribution network switch adjustment state behavior based on the Markov state of the upper optimization model and the Markov behavior selection probability distribution table of the upper optimization model to form a fault recovery plan, and transmits the fault recovery plan to the lower optimization model;

[0043] S3: selecting a Markov decision behavior action of the lower-layer optimization model based on the fault recovery solution and the Markov behavior selection probability distribution table of the lower-layer optimization model to form the power output of each device of the access node;

[0044] S4 calculates the load loss power evaluation index for executing the decision-making behavior of the upper-layer optimization model in the current round based on the fault recovery plan, the attenuation factor of the optimization criterion of the upper-layer optimization model and the distribution network layer optimization criterion;

[0045] S5 calculates the objective function value of the decision-making behavior of the lower-level optimization model in the current round based on the power output of each device of the node and the attenuation factor of the optimization criterion of the lower-level optimization model in combination with the optimization criterion of the lower-level optimization model;

[0046] S6: updating the selection probability distribution table of the lower-level optimization model based on the classic roulette wheel selection method, and using the updated selection probability distribution table of the lower-level optimization model as the selection probability distribution table of the lower-level optimization model; and updating the selection probability distribution table of the upper-level optimization model based on the classic roulette wheel selection method, and using the updated selection probability distribution table of the upper-level optimization model as the selection probability distribution table of the upper-level optimization model;

[0047] S7 determines whether the current round has reached the maximum iteration round of deep learning. If not, the current round is incremented by 1 and returned to S2. Otherwise, the load-loss power evaluation index and objective function value of the current round are output, and the optimization decision process ends.

[0048] The deep learning algorithm parameters include: the maximum iteration round of deep reinforcement learning, the Markov behavior selection probability distribution table of the upper optimization model and the attenuation factor of the optimization criterion, the Markov behavior selection probability distribution table of the lower optimization model and the attenuation factor value of the optimization criterion;

[0049] The parameters of the two-layer optimization model include: distribution network operation parameters, distribution network topology parameters, distribution network switch state parameters, and equipment operation state parameters of each energy subsystem of the multi-energy coupling distribution network.

[0050] Optionally, the optimization criterion of the upper-layer optimization model is as follows:

[0051]

[0052] Where V cl It is the load-loss power evaluation index; is the node importance weight; is the interruption action of the switch at the i-th load node at the decision time k; is the interruption power of the i-th load node at the decision time k; n mg is the total amount of load nodes; Tk is the kth moment; k is the time; i is the device number.

[0053] Optionally, the optimization criterion of the lower optimization model is calculated as follows:

[0054]

[0055] Where V me The lower layer objective function value for multi-energy coupled distribution network fault recovery; T k is the kth moment; k is the moment; N is the number of generator sets; is the operating cost of equipment i; is the status of device i; is the action of device i.

[0056] In another aspect, the present application further provides a computing device, comprising: one or more processors;

[0057] a processor for executing one or more programs;

[0058] When the one or more programs are executed by the one or more processors, the multi-energy coupling distribution network fault recovery method as described above is implemented.

[0059] On the other hand, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed, the multi-energy coupling distribution network fault recovery method as described above is implemented.

[0060] Compared with the prior art, the present invention has the following beneficial effects:

[0061] The present invention provides a multi-energy coupled distribution network fault recovery method, comprising obtaining parameters of a distribution network layer and parameters of an energy subsystem of the multi-energy coupled distribution network; introducing the parameters of the distribution network layer and the parameters of the energy subsystem into a pre-constructed two-layer optimization model, solving the problem using a deep reinforcement learning algorithm to obtain a load loss power evaluation index and an objective function value; using the fault recovery solution corresponding to the load loss power evaluation index and the objective function value as the fault recovery solution for the multi-energy coupled distribution network; the deep reinforcement learning model is constructed by combining an upper-layer optimization model constructed based on the parameters and optimization criteria of the distribution network layer and a lower-layer optimization model constructed based on the parameters and optimization criteria of the energy subsystem with a deep learning algorithm. The present invention adopts a two-layer optimization model and combines it with a deep reinforcement learning algorithm for solving the problem that the existing technology only processes a single-layer model and has low optimization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is a flow chart of a multi-energy coupling distribution network fault recovery method of the present invention;

[0063] Figure 2 A schematic diagram of a two-layer model for fault recovery of a multi-energy coupled distribution network according to the present invention;

[0064] Figure 3 This is a flowchart for optimizing the deep reinforcement learning model of the present invention;

[0065] Figure 4 A schematic diagram of a multi-energy coupling distribution network fault recovery system according to the present invention;

[0066] Figure 5 Schematic diagram of the topological structure of the power distribution network of the present invention;

[0067] Figure 6 Schematic diagram of comparison of optimized operating power of the system during fault period of the present invention. DETAILED DESCRIPTION

[0068] The present invention proposes a multi-energy coupling distribution network fault recovery method, establishes a multi-layer deep reinforcement learning model for multi-energy coupling distribution network fault recovery, establishes a Markov model that can parallel process optimization problems for the fault recovery switch state and the optimized operation plan of each energy system, uses the reinforcement learning algorithm to process the Markov model, realizes the parallel optimization solution of the distribution network layer and the equipment layer, and forms an efficient multi-energy coupling distribution network fault recovery system.

[0069] Example 1:

[0070] A multi-energy coupled distribution network fault recovery method, such as Figure 1 As shown, including:

[0071] Step 1: Obtain the parameters of the distribution network layer and the parameters of the energy subsystem of the multi-energy coupling distribution network;

[0072] Step 2: The parameters of the distribution network layer and the parameters of the energy subsystem are introduced into a pre-built two-layer optimization model, and a deep reinforcement learning algorithm is used to solve the model to obtain the load loss power evaluation index and the objective function value;

[0073] Step 3: Parameters of the distribution network layer and parameters of the energy subsystem corresponding to the load loss power evaluation index and the objective function value are used as a fault recovery plan for the multi-energy coupling distribution network;

[0074] Among them, the deep reinforcement learning model is constructed based on the upper-level optimization model constructed based on the parameters and optimization criteria of the distribution network layer, and the lower-level optimization model constructed based on the parameters and optimization criteria of the energy subsystem combined with the deep learning algorithm.

[0075] The technical solution of the present invention to solve the above technical problems is: a multi-energy coupling distribution network fault recovery method, comprising the following steps:

[0076] Before step 1, it also includes: constructing a two-layer model for fault recovery of a multi-energy coupled distribution network, such as Figure 2 As shown, specifically including:

[0077] The first step is to establish an upper-level optimization model for fault recovery switch states in multi-energy coupled distribution networks.

[0078] The upper-level optimization model of the fault recovery switch state of the multi-energy coupled distribution network includes the distribution network layer Markov state, distribution network layer action and distribution network layer optimization criteria.

[0079] Distribution network Markov state of the upper layer optimization model of distribution network It can be expressed as:

[0080]

[0081] in, is the interruption power of the i-th load node at the decision time k, n mg is the total amount of load nodes, is the interruption power of the first load node at the decision time k, is the interruption power of the second load node at the decision time k.

[0082] Furthermore, the Markov actions of the distribution network layer of the upper optimization model for:

[0083]

[0084] in, is the switch interruption action of the i-th load node at the decision time k.

[0085] Furthermore, the Markov optimization criterion of the distribution network layer of the upper optimization model is to minimize the load loss power within the fault time, specifically:

[0086]

[0087] Among them, V cl It is the load-loss power evaluation index; is the node importance weight; is the interruption action of the switch at the i-th load node at the decision time k; is the interruption power of the i-th device at the load node at the decision time k; n mg is the total amount of load nodes; T k is the kth moment; k is the time; i is the device number.

[0088] The second step is to establish a lower-level optimization model for fault recovery of multi-energy coupled distribution networks.

[0089] Furthermore, the underlying optimization model for fault recovery of multi-energy coupled distribution networks includes the Markov states, actions, and optimization criteria of energy subsystems.

[0090] Furthermore, the Markov state of the energy subsystem of the lower-level optimization model for multi-energy coupled distribution network fault recovery is:

[0091]

[0092] in, is the power generation capacity of the micro gas turbine; is the power generation capacity of the fuel cell; To purchase and sell electricity to the upper power grid; is the photovoltaic power generation power; is the load interruption power; Exchange power for thermal energy systems; exchanging power for natural gas systems; is the operating power of the traditional generator set i; is the operating power of the traditional generator set 1; is the operating power of the traditional generator set 2.

[0093] Furthermore, the Markov actions of the energy subsystem of the lower-level optimization model for multi-energy coupled distribution network fault recovery are:

[0094]

[0095] in, It is a Markov behavior model of energy subsystems in multi-energy coupled distribution network; is the power generation change of the micro gas turbine; is the change in power generation of the fuel cell; The change in power purchased and sold to the upper power grid; is the change in photovoltaic power generation power; is the load interruption power change; Exchange power variation for thermal energy systems; Exchange power variation for natural gas systems; is the operating power variation of the traditional generator set i; is the power change of the energy storage system; is the operating power variation of the traditional generator set 1; is the operating power variation of the traditional generator set 2.

[0096] Furthermore, the optimization criteria for the energy subsystem of the multi-energy coupling operation optimization model under the multi-energy coupling distribution network fault recovery are as follows:

[0097]

[0098] Among them, V me is the objective function value of the lower layer of multi-energy coupled distribution network fault recovery; T k is the kth moment; k is the moment; is the operating cost of equipment i; is the status of device i; is the action of device i.

[0099] Step 3: Combine the upper-level optimization model constructed in the first step and the lower-level optimization model constructed in the second step to construct a two-level optimization model for fault recovery of multi-energy coupled distribution networks.

[0100] Step 1: Obtain the parameters of the distribution network layer and the parameters of the energy subsystem of the multi-energy coupling distribution network, specifically including: the parameters of the distribution network layer and the parameters of the energy subsystem.

[0101] The parameters of the distribution network layer include:

[0102] Decision-making on the interruption power of each load node at each moment, the number of load nodes, and the switch interruption action of each load node at each decision moment;

[0103] The parameters of the energy subsystem include: the power generation power of the micro gas turbine, the power generation power of the fuel cell, the power purchased and sold to the upper-level power grid, the photovoltaic power generation power, the load interruption power, the thermal energy system exchange power, the natural gas system exchange power, the operating power of the traditional generator set, and the Markov action of the energy subsystem of the multi-energy coupled distribution network; the power generation power change of the micro gas turbine, the power generation power change of the fuel cell, the power purchase and sale to the upper-level power grid, the photovoltaic power generation power change, the load interruption power change, the thermal energy system exchange power change, the natural gas system exchange power change, the operating power change of the traditional generator set and the energy storage system power change.

[0104] Step 2: Bring the parameters of the distribution network layer and the parameters of the energy subsystem into the pre-built two-layer optimization model, use the deep reinforcement learning algorithm to solve, and obtain the load loss power evaluation index and objective function value. The specific steps are as follows: Figure 3 As shown, including:

[0105] Step 2.1: Initialize the parameters of the multi-energy coupled distribution network model, including distribution network operation parameters, distribution network topology parameters, distribution network switch state parameters, and equipment operation state parameters of each energy subsystem of the multi-energy coupled distribution network.

[0106] Step 2.2: Initialize the parameters of the deep reinforcement learning algorithm based on the Markov process model, including the maximum number of iterations of deep reinforcement learning, the selection probability distribution table of the two-layer Markov behavior model, and the attenuation factor value of the two-layer Markov optimization criterion.

[0107] Step 2.3: Based on the Markov state of the upper-level optimization model of the multi-energy coupling distribution network fault recovery switch state, the distribution network switch adjustment state behavior is selected in combination with the Markov behavior selection probability distribution table of the upper-level optimization model to form a fault recovery plan, and the switch adjustment state in the fault recovery plan is transmitted to the lower-level multi-energy coupling operation optimization model of the multi-energy coupling distribution network fault recovery.

[0108] Step 2.4: Based on the fault recovery plan formed by the upper-layer optimization model, the Markov behavior selection probability distribution table of the lower-layer optimization model is combined to select the Markov decision behavior action of the lower-layer optimization model to form the power output of each device at the access node.

[0109] Step 2.5: Based on the optimization criteria of the upper optimization model of the multi-energy coupling distribution network fault recovery switch state, calculate the load loss power evaluation index V for the decision behavior of the upper optimization model in this round cl ;.

[0110] Step 2.6: Based on the optimization criteria of the multi-energy coupling distribution network fault recovery lower-level multi-energy coupling operation optimization model, calculate the objective function value V of the decision-making behavior of the lower-level optimization model for executing this round me .

[0111] Step 2.7: Update the selection probability distribution table of the lower-level model for multi-energy coupled distribution network fault restoration;

[0112] Step 2.8: Update the selection probability distribution table of the upper-layer model for multi-energy coupled distribution network fault recovery;

[0113] Step 2.9: If the maximum number of iterations of deep reinforcement learning has not been reached, return to step 2.3; if the decision-making process is completed, the optimization decision-making process ends.

[0114] Step 3: The parameters of the distribution network layer and the parameters of the energy subsystem corresponding to the load loss power evaluation index and the objective function value are used as the fault recovery plan of the multi-energy coupling distribution network.

[0115] Example 2:

[0116] In order to make the purpose, technical solutions and advantages of the present invention more clear, the following Figure 5 , the present invention is described in further detail.

[0117] A multi-energy coupled distribution network is selected for case analysis. The multi-energy coupled distribution network is based on the IEEE 33-node standard distribution network system and includes a thermal energy network and a natural gas network. The distribution network topology is as follows: Figure 5As shown, there is a connecting branch between nodes 18-32, the microturbine (MT) is located at node 32, and the fuel cell (FC) is located at node 28. The fault mode is considered to be a permanent fault on branch 26-27.

[0118] Based on the model proposed in this invention, the fault recovery solution is to disconnect the switches at branches 28-29 and 29-30 and close the switches at the connecting branch 18-32. Figure 6 As shown. It can be seen that after the failure occurs, the output of the micro gas turbine located at node 32 increases, which can provide support to the node 18 after the transfer. The support level of the fuel cell located at node 28 is limited, so its operating power remains basically unchanged. It can be seen that the solution proposed in the present invention can effectively utilize multi-energy power generation resources to restore node power supply after a failure occurs. Table 1 shows the comparison of the operating indicators of the model proposed in the present invention and the operating indicators of the traditional optimization model. It can be seen that the model proposed in the present invention has better solution performance.

[0119] Table 1 Comparison of operating indicators of different models

[0120] Fault recovery method Fault load loss / kW Solution time / s Number of iterations Method of the present invention 492 15.4 127 Traditional methods 483 9.5 97

[0121] Example 3:

[0122] The present invention also provides a multi-energy coupling distribution network fault recovery system, such as Figure 4 As shown, including:

[0123] An acquisition module is used to obtain parameters of the distribution network layer and energy subsystem of the multi-energy coupling distribution network;

[0124] A calculation module is used to bring the parameters of the distribution network layer and the parameters of the energy subsystem into a pre-built two-layer optimization model, and solve it using a deep reinforcement learning algorithm to obtain a load loss power evaluation index and an objective function value;

[0125] A solution determination module is used to use the parameters of the distribution network layer and the energy subsystem corresponding to the load loss power evaluation index and the objective function value as a fault recovery solution for the multi-energy coupling distribution network;

[0126] Among them, the deep reinforcement learning model is constructed based on the upper-level optimization model constructed based on the parameters and optimization criteria of the distribution network layer, and the lower-level optimization model constructed based on the parameters and optimization criteria of the energy subsystem combined with the deep learning algorithm.

[0127] A multi-energy coupling distribution network fault recovery system also includes:

[0128] A two-layer optimization model construction module is used to construct an upper-layer optimization model based on the distribution network layer Markov state, the distribution network layer switch interruption action, and the distribution network layer optimization criterion of minimizing the load loss power within the fault time;

[0129] The lower-level optimization model is constructed based on the Markov state of the energy subsystem, the interruption action of the energy subsystem switch, and the minimum power recovery of the multi-energy coupling distribution network as the optimization criterion of the energy subsystem;

[0130] The upper-layer optimization model and the lower-layer optimization model are combined to construct a deep reinforcement learning model.

[0131] The calculation module is specifically used for:

[0132] S1 initializes the parameters of the two-layer optimization model and the parameters of the deep learning algorithm;

[0133] S2 selects the distribution network switch adjustment state behavior based on the Markov state of the upper optimization model and the Markov behavior selection probability distribution table of the upper optimization model to form a fault recovery plan, and transmits the fault recovery plan to the lower optimization model;

[0134] S3: selecting a Markov decision behavior action of the lower-layer optimization model based on the fault recovery solution and the Markov behavior selection probability distribution table of the lower-layer optimization model to form the power output of each device of the access node;

[0135] S4 calculates the load loss power evaluation index for executing the decision-making behavior of the upper-layer optimization model in the current round based on the fault recovery plan, the attenuation factor of the optimization criterion of the upper-layer optimization model and the distribution network layer optimization criterion;

[0136] S5 calculates the objective function value of the decision-making behavior of the lower-level optimization model in the current round based on the power output of each device of the node and the attenuation factor of the optimization criterion of the lower-level optimization model in combination with the optimization criterion of the lower-level optimization model;

[0137] S6: updating the selection probability distribution table of the lower-level optimization model based on the classic roulette wheel selection method, and using the updated selection probability distribution table of the lower-level optimization model as the selection probability distribution table of the lower-level optimization model; and updating the selection probability distribution table of the upper-level optimization model based on the classic roulette wheel selection method, and using the updated selection probability distribution table of the upper-level optimization model as the selection probability distribution table of the upper-level optimization model;

[0138] S7 determines whether the current round has reached the maximum iteration round of deep learning. If not, the current round is incremented by 1 and returned to S2. Otherwise, the load-loss power evaluation index and objective function value of the current round are output, and the optimization decision process ends.

[0139] The deep learning algorithm parameters include: the maximum iteration round of deep reinforcement learning, the Markov behavior selection probability distribution table of the upper optimization model and the attenuation factor of the optimization criterion, the Markov behavior selection probability distribution table of the lower optimization model and the attenuation factor value of the optimization criterion;

[0140] The parameters of the two-layer optimization model include: distribution network operation parameters, distribution network topology parameters, distribution network switch state parameters, and equipment operation state parameters of each energy subsystem of the multi-energy coupling distribution network.

[0141] The optimization criterion of the upper optimization model is as follows:

[0142]

[0143] Where V cl It is the load-loss power evaluation index; is the node importance weight; is the interruption action of the switch at the i-th load node at the decision time k; is the interruption power of the i-th load node at the decision time k; n mg is the total amount of load nodes; T k is the kth moment; k is the time; i is the device number.

[0144] The optimization criterion of the lower optimization model is calculated as follows:

[0145]

[0146] Where V me The lower layer objective function value for multi-energy coupled distribution network fault recovery; T k is the kth moment; k is the moment; N is the number of generator sets; is the operating cost of equipment i; is the status of device i; is the action of device i.

[0147] The Markov actions of the energy subsystem in the lower-level optimization model for multi-energy coupled distribution network fault recovery are:

[0148]

[0149] in, It is a Markov behavior model of energy subsystems in multi-energy coupled distribution network; is the power generation change of the micro gas turbine; is the change in power generation of the fuel cell; The change in power purchased and sold to the upper power grid; is the change in photovoltaic power generation power; is the load interruption power change; Exchange power variation for thermal energy systems; Exchange power variation for natural gas systems; is the operating power variation of the traditional generator set i; is the power change of the energy storage system; is the operating power variation of the traditional generator set 1; is the operating power variation of the traditional generator set 2.

[0150] Example 4:

[0151] Based on the same inventive concept, the present invention also provides a computer device, which includes a processor and a memory, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or may be 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. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the computer storage medium to implement the corresponding method flow or corresponding function, so as to implement the steps of a multi-energy coupled distribution network fault recovery method in the above embodiment.

[0152] Example 5:

[0153] Based on the same inventive concept, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It is understandable that the computer-readable storage medium here can include both built-in storage media in the computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space that stores the operating system of the terminal. In addition, the storage space also stores one or more instructions suitable for being loaded and executed by the processor. These instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the steps of a multi-energy coupled distribution network fault recovery method in the above embodiment.

[0154] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0155] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0156] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1The function specified in one or more boxes.

[0157] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0158] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are included in the scope of the claims of the present invention to be approved.

Claims

1. A multi-energy coupling distribution network fault recovery method, characterized in that: include: Obtain the parameters of the distribution layer and energy subsystem of the multi-energy coupled distribution network; The parameters of the distribution network layer and the parameters of the energy subsystem are introduced into a pre-built two-layer optimization model, and a deep reinforcement learning algorithm is used to solve the problem to obtain the load loss power evaluation index and the objective function value; The parameters of the distribution network layer and the energy subsystem corresponding to the load loss power evaluation index and the objective function value are used as a fault recovery plan for the multi-energy coupling distribution network; The two-layer optimization model is constructed by combining an upper-layer optimization model based on the parameters and optimization criteria of the distribution network layer and a lower-layer optimization model based on the parameters and optimization criteria of the energy subsystem with a deep learning algorithm. The parameters of the distribution network layer and the parameters of the energy subsystem are brought into the pre-built two-layer optimization model, and the deep reinforcement learning algorithm is used to solve the problem to obtain the load loss power evaluation index and the objective function value, including: S1 initializes the parameters of the two-layer optimization model and the parameters of the deep learning algorithm; S2 selects the distribution network switch adjustment state behavior based on the Markov state of the upper optimization model and the Markov behavior selection probability distribution table of the upper optimization model to form a fault recovery plan, and transmits the fault recovery plan to the lower optimization model; S3: selecting a Markov decision behavior action of the lower-layer optimization model based on the fault recovery solution and the Markov behavior selection probability distribution table of the lower-layer optimization model to form the power output of each device of the access node; S4 calculates the load loss power evaluation index for executing the decision-making behavior of the upper-layer optimization model in the current round based on the fault recovery plan, the attenuation factor of the optimization criterion of the upper-layer optimization model and the distribution network layer optimization criterion; S5 calculates the objective function value of the decision-making behavior of the lower-level optimization model in the current round based on the power output of each device of the node and the attenuation factor of the optimization criterion of the lower-level optimization model in combination with the optimization criterion of the lower-level optimization model; S6: updating the selection probability distribution table of the lower-level optimization model based on the classic roulette wheel selection method, and using the updated selection probability distribution table of the lower-level optimization model as the selection probability distribution table of the lower-level optimization model; and updating the selection probability distribution table of the upper-level optimization model based on the classic roulette wheel selection method, and using the updated selection probability distribution table of the upper-level optimization model as the selection probability distribution table of the upper-level optimization model; S7 determines whether the current round has reached the maximum iteration round of deep learning. If not, the current round is increased by 1 and returned to S2. Otherwise, the load-loss power evaluation index and objective function value of the current round are output, and the optimization decision process ends.

2. The method according to claim 1, wherein The construction of the two-layer optimization model includes: The upper layer optimization model is constructed based on the Markov state of the distribution network layer, the interruption action of the distribution network layer switch, and the minimum load loss power within the minimum fault time as the distribution network layer optimization criterion; The lower-level optimization model is constructed based on the Markov state of the energy subsystem, the interruption action of the energy subsystem switch, and the minimum power recovery of the multi-energy coupling distribution network as the optimization criterion of the energy subsystem; The upper-layer optimization model and the lower-layer optimization model are combined to construct a two-layer optimization model.

3. The method according to claim 1, wherein The parameters of the distribution network layer include: Decision-making on the interruption power of each load node at each moment, the number of load nodes, and the switch interruption action of each load node at each decision moment; The parameters of the energy subsystem include: power generation value, power purchase and sale to the upper power grid, natural gas system exchange power, operating power of traditional generator sets and power variation.

4. The method according to claim 1, wherein The parameters of the deep learning algorithm include: the maximum iteration round of deep reinforcement learning, the Markov behavior selection probability distribution table of the upper optimization model and the attenuation factor of the optimization criterion, the Markov behavior selection probability distribution table of the lower optimization model and the attenuation factor value of the optimization criterion; The parameters of the two-layer optimization model include: distribution network operation parameters, distribution network topology parameters, distribution network switch state parameters, and equipment operation state parameters of each energy subsystem of the multi-energy coupling distribution network.

5. The method according to claim 1, wherein The optimization criterion of the upper optimization model is as follows: ； Where, It is the load-loss power evaluation index; is the node importance weight; is the interruption action of the switch at the i-th load node at the decision time k; is the interruption power of the i-th device at the load node at the decision time k; is the total amount of load nodes; is the kth moment; k is the time; i is the device number.

6. The method according to claim 5, wherein The optimization criterion of the lower optimization model is calculated as follows: ； Where, is the lower layer objective function value of multi-energy coupled distribution network fault recovery; N is the number of generator sets; is the operating cost of the i-th device; For devices Status; For devices action.

7. A system for implementing the multi-energy coupling distribution network fault recovery method according to any one of claims 1 to 6, characterized in that: include: An acquisition module is used to obtain parameters of the distribution network layer and energy subsystem of the multi-energy coupling distribution network; A calculation module is used to bring the parameters of the distribution network layer and the parameters of the energy subsystem into a pre-built two-layer optimization model, and solve it using a deep reinforcement learning algorithm to obtain a load loss power evaluation index and an objective function value; A solution determination module is used to use the parameters of the distribution network layer and the energy subsystem corresponding to the load loss power evaluation index and the objective function value as a fault recovery solution for the multi-energy coupling distribution network; Among them, the two-layer optimization model is constructed based on the parameters and optimization criteria of the distribution network layer, and the lower-layer optimization model constructed based on the parameters and optimization criteria of the energy subsystem combined with a deep learning algorithm.

8. The system according to claim 7, wherein: It also includes a two-layer optimization model construction module for constructing an upper-layer optimization model based on the distribution network layer Markov state, the distribution network layer switch interruption action, and the distribution network layer optimization criterion of minimizing the load loss power within the fault time; The lower-level optimization model is constructed based on the Markov state of the energy subsystem, the interruption action of the energy subsystem switch, and the minimum power recovery of the multi-energy coupling distribution network as the optimization criterion of the energy subsystem; The upper-layer optimization model and the lower-layer optimization model are combined to construct a two-layer optimization model.

9. The system according to claim 7, wherein: The calculation module is specifically used for: S1 initializes the parameters of the two-layer optimization model and the parameters of the deep learning algorithm; S2 selects the distribution network switch adjustment state behavior based on the Markov state of the upper optimization model and the Markov behavior selection probability distribution table of the upper optimization model to form a fault recovery plan, and transmits the fault recovery plan to the lower optimization model; S3: selecting a Markov decision behavior action of the lower-layer optimization model based on the fault recovery solution and the Markov behavior selection probability distribution table of the lower-layer optimization model to form the power output of each device of the access node; S4 calculates the load loss power evaluation index for executing the decision-making behavior of the upper-layer optimization model in the current round based on the fault recovery plan, the attenuation factor of the optimization criterion of the upper-layer optimization model and the distribution network layer optimization criterion; S5 calculates the objective function value of the decision-making behavior of the lower-level optimization model in the current round based on the power output of each device of the node and the attenuation factor of the optimization criterion of the lower-level optimization model in combination with the optimization criterion of the lower-level optimization model; S6: updating the selection probability distribution table of the lower-level optimization model based on the classic roulette wheel selection method, and using the updated selection probability distribution table of the lower-level optimization model as the selection probability distribution table of the lower-level optimization model; and updating the selection probability distribution table of the upper-level optimization model based on the classic roulette wheel selection method, and using the updated selection probability distribution table of the upper-level optimization model as the selection probability distribution table of the upper-level optimization model; S7 determines whether the current round has reached the maximum iteration round of deep learning. If not, the current round is incremented by 1 and returned to S2. Otherwise, the load-loss power evaluation index and objective function value of the current round are output, and the optimization decision process ends. The deep learning algorithm parameters include: the maximum iteration round of deep reinforcement learning, the Markov behavior selection probability distribution table of the upper optimization model and the attenuation factor of the optimization criterion, the Markov behavior selection probability distribution table of the lower optimization model and the attenuation factor value of the optimization criterion; The parameters of the two-layer optimization model include: distribution network operation parameters, distribution network topology parameters, distribution network switch state parameters, and equipment operation state parameters of each energy subsystem of the multi-energy coupling distribution network.

10. The system according to claim 9, wherein: The optimization criterion of the upper optimization model is as follows: ； Where, It is the load-loss power evaluation index; is the node importance weight; is the interruption action of the switch at the i-th load node at the decision time k; is the interruption power of the i-th load node at the decision time k; is the total amount of load nodes; is the kth moment; k is the moment; i is the load node number.

11. The system according to claim 10, wherein: The optimization criterion of the lower optimization model is calculated as follows: ； Where, is the lower layer objective function value of multi-energy coupled distribution network fault recovery; N is the number of generator sets; For equipment operating costs; For equipment Status; For equipment action.

12. A computer device, characterized in that: include: one or more processors; The processor is configured to store one or more programs; When the one or more programs are executed by the one or more processors, a multi-energy coupling distribution network fault recovery method according to any one of claims 1 to 6 is implemented.

13. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed, a multi-energy coupling distribution network fault recovery method according to any one of claims 1 to 6 is implemented.

Images