Source network load storage integrated optimization analysis method and system based on machine learning

Abstract

The invention relates to the technical field of power distribution network source network load storage optimization, in particular to a source network load storage integrated optimization analysis method and system based on machine learning, and the method comprises the following steps: collecting a plurality of operation scenes of a power distribution network and a plurality of optimization targets of power distribution network source network load storage integrated scheduling; performing optimization effect measurement and calculation according to each optimization target in each operation scene in sequence to obtain an effect index obtained under each optimization target in each operation scene; on the basis of the effect indexes, determining an optimal optimization target for power distribution network source network load storage integrated scheduling in each operation scene; and performing mapping training on the operation scene and the optimal optimization target by using a convolutional neural network to obtain an optimization target screening model for determining the optimal optimization target according to the operation scene. According to the method, the optimization target screening model of the optimal optimization target is determined according to the operation scene, the optimization target is objectively screened by using the model, and the power distribution network source network load storage integrated optimization scheduling efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power distribution network source-grid-load-storage optimization, and in particular to a power distribution network source-grid-load-storage integrated optimization analysis method and system based on machine learning. Background Art

[0002] In "source-grid-load-storage integration," intelligent technologies enable comprehensive monitoring, control, and coordinated operation of power sources, grids, charges, and energy storage to achieve efficient energy utilization and stable system operation. This integrated power system is more adaptable, enabling better integration of renewable energy, reducing grid volatility, and increasing the proportion of renewable energy.

[0003] The results of prior art optimization for integrated dispatching of distribution networks with power sources, grids, loads, and storage vary under different distribution network operating scenarios and under different optimization objectives. In other words, the results of optimization for integrated dispatching of distribution networks with power sources, grids, loads, and storage depend on the operating scenario and the optimization objective. Currently, while optimization for integrated dispatching of distribution networks with power sources, grids, loads, and storage takes different scenarios into account, the selection of optimization objectives tends to be subjective, making it easy for subjectively chosen optimization objectives to fail to achieve optimal dispatching of the integrated dispatching of distribution networks with power sources, grids, loads, and storage. Consequently, this subjective nature makes the dispatching of the integrated dispatching of distribution networks with power sources, grids, loads, and storage uncontrollable. Summary of the Invention

[0004] The purpose of the present invention is to provide a source-grid-load-storage integrated optimization analysis method and system based on machine learning, so as to solve the technical problem in the existing technology that the selection of optimization targets tends to be subjective and difficult to achieve the effect of optimal scheduling of the source-grid-load-storage integrated distribution network.

[0005] In order to solve the above technical problems, the present invention specifically provides the following technical solutions:

[0006] A machine learning-based optimization analysis method for source-grid-load-storage integration includes the following steps:

[0007] Collect multiple operating scenarios of the distribution network and multiple optimization objectives for the integrated dispatch of distribution network sources, grids, loads and storage;

[0008] Calculate the optimization effect with each optimization target in each operation scenario in turn, and obtain the effect index achieved under each optimization target in each operation scenario;

[0009] Based on the effect indicators, determine the optimal optimization target for the integrated dispatch of distribution network sources, grids, loads and storage under various operation scenarios;

[0010] A convolutional neural network is used to map the operating scenarios and the optimal optimization targets, and an optimization target screening model is obtained that determines the optimal optimization target according to the operating scenarios.

[0011] As a preferred solution of the present invention, the effect indicators include voltage stability, wind and solar reliability, and power supply quality, wherein the quantified formula for voltage stability is:

[0012]

[0013] Where GU is the voltage stability of the distribution network, U i,t is the voltage value of the ith node in the distribution network at the tth moment, avg(U i ) is the mean voltage of the i-th node in the distribution network participating in the source-grid-load-storage integrated dispatch, N is the total number of nodes in the distribution network, and T is the total operation time of the source-grid-load-storage integrated dispatch of the distribution network;

[0014] The quantitative formula of wind-solar reliability is:

[0015]

[0016] Where GP is the wind and solar reliability rate of the distribution network, P i,t is the power consumption of the i-th node in the distribution network at the t-th moment, P wt,t is the power generation of the wind system at the tth moment, P pv,t is the power generation of the photovoltaic system at the tth moment, N is the total number of nodes in the distribution network, and T is the total operation time of the integrated source-grid-load-storage scheduling of the distribution network;

[0017] The quantitative formula for the power supply quality is:

[0018]

[0019] Where GC is the power quality of the distribution network, P j,t is the power consumption of the jth load in the distribution network at the tth moment, {P j,t |t∈[1,T]} is the power consumption of the jth load in the distribution network from time 1 to time T, max({P j,t |t∈[1,T]}) is {P j,t |The maximum value in t∈[1,T]}, min({P j,t |t∈[1,T]}) is {P j,t |The minimum value in t∈[1,T]}, avg({P j,t |t∈[1,T]}) is {P j,t|t∈[1,T]}, M is the total number of loads in the distribution network, and T is the total operation time of the integrated dispatching of source, grid, load and storage in the distribution network.

[0020] As a preferred solution of the present invention, the method of measuring the optimization effect with each optimization target in each operation scenario includes:

[0021] Randomly select multiple running scenarios as sample scenarios;

[0022] In each sample scenario, multiple optimization objectives are randomly selected and combined as the target set for source-grid-load-storage integrated scheduling in that sample scenario;

[0023] Minimize each target set of each sample scenario one by one to obtain the source-grid-load-storage integrated scheduling data under the sample scenario;

[0024] Based on the source-grid-load-storage integrated scheduling data, the distribution network in the sample scenario is dispatched in an integrated manner, and the voltage stability, wind-solar reliability and power quality of the distribution network in the sample scenario are quantified between time 1 and time T;

[0025] The sample scenario and target set are used as the input of the BP neural network, and the voltage stability of the distribution network, the wind and solar reliability rate of the distribution network, and the power quality of the distribution network are used as the output of the BP neural network. The BP neural network is trained to obtain an optimization effect measurement model;

[0026] The optimization effect calculation model includes:

[0027] (GU,GP,GC)=BP(S,Z);

[0028] Where GU is the voltage stability of the distribution network, GP is the wind and solar reliability rate of the distribution network, GC is the power quality of the distribution network, S is the operation scenario, and Z is the target set;

[0029] Based on the optimization effect measurement model, the optimization effect of each target set formed by all optimization targets in each operation scenario is calculated to obtain the voltage stability of the distribution network, the wind and solar reliability rate of the distribution network, and the power supply quality of the distribution network corresponding to each target set formed by all optimization targets in each operation scenario.

[0030] As a preferred solution of the present invention, the method for determining the optimal optimization target includes:

[0031] In each operation scenario, the target set corresponding to the highest stability, the highest wind-solar reliability rate, and the highest power supply quality is selected as the optimal target set for that operation scenario;

[0032] All optimization targets in the optimal target set are taken as the optimal optimization targets in the operation scenario.

[0033] As a preferred embodiment of the present invention, the method for constructing the optimization target screening model includes:

[0034] The running scenario is used as the input of the convolutional neural network, and the optimal target set is used as the output of the convolutional neural network. The convolutional neural network is trained to obtain the optimization target screening model;

[0035] The optimization target screening model is:

[0036] Z best =CNN(S);

[0037] Where Z best is the optimal target set, S is the operating scenario, and CNN is the convolutional neural network.

[0038] As a preferred solution of the present invention, the operation scenario is formed by connecting the distribution network to a wind power system, a photovoltaic system, and an energy storage system.

[0039] As a preferred solution of the present invention, the optimization objectives include: the cost of purchasing electricity from the superior power grid, network loss cost, the cost of purchasing electricity from the wind system and photovoltaic system, the operation and maintenance cost of the wind system and photovoltaic system, and the operation and maintenance cost of the energy storage system;

[0040] Among them, the cost of purchasing electricity from the superior power grid is: Where FA is the cost of purchasing electricity from the upper grid, c A,t is the electricity purchase price at the distribution network location at the tth moment, P A,t is the power purchased by the distribution network from the upper power grid at the tth moment, and T is the total operation time of the distribution network's source-grid-load-storage integrated dispatching;

[0041] The network loss cost is: Where FB is the network loss cost, c B,t is the electricity price at the distribution network location at the tth moment, P B,t is the power loss of the distribution network at the tth moment;

[0042] The cost of purchasing electricity from wind power systems and photovoltaic systems is: Where FC is the electricity purchase cost from wind power system and photovoltaic system, c wt,C,t is the electricity price of the wind power system at the tth moment, c pv,C,t is the electricity price of the photovoltaic system at the tth moment, P wt,t is the power generation of the wind system at the tth moment, P pv,t is the power generation of the photovoltaic system at the tth moment;

[0043] The operation and maintenance cost of the energy storage system is: Where FD is the operation and maintenance cost of the energy storage system, c D is the operation and maintenance cost coefficient of the energy storage system, P in,t is the charging power of the energy storage system at the tth moment, P out,t is the discharge power of the energy storage system at the tth moment;

[0044] The operation and maintenance costs of wind power systems and photovoltaic systems are: Where FE is the operation and maintenance cost of wind power system and photovoltaic system, c wt,F is the operation and maintenance cost coefficient of the wind power system, c pv,F is the operation and maintenance cost coefficient of the photovoltaic system.

[0045] As a preferred solution of the present invention, the source-grid-load-storage integrated dispatching data includes P A,t , P B,t , P wt,t , P pv,t , P in,t , P out,t , where P A,t P is the power purchased by the distribution network from the upper grid at the tth moment, B,t is the power loss of the distribution network at the tth moment, P wt,t is the power generation of the wind system at the tth moment, P pv,t is the power generation of the photovoltaic system at the tth moment, P in,t is the charging power of the energy storage system at the tth moment, P out,t is the discharge power of the energy storage system at the tth moment.

[0046] As a preferred embodiment of the present invention, the present invention provides a source-grid-load-storage integrated optimization analysis method based on machine learning, which is applied to a source-grid-load-storage integrated optimization analysis method based on machine learning. The system includes:

[0047] The data acquisition unit is used to collect multiple operating scenarios of the distribution network and multiple optimization objectives of the integrated dispatching of the distribution network's source, grid, load and storage;

[0048] A data processing unit is used to measure the optimization effect with each optimization target in each operation scenario in turn, obtain the effect index achieved under each optimization target in each operation scenario, and determine the optimal optimization target for the integrated dispatch of the distribution network source, grid, load and storage in each operation scenario based on the effect index;

[0049] The deep learning unit is used to use a convolutional neural network to perform mapping training between the operating scenario and the optimal optimization target, and obtain an optimization target screening model that determines the optimal optimization target based on the operating scenario.

[0050] As a preferred embodiment of the present invention, the method for constructing the optimization target screening model by the deep learning unit includes:

[0051] The running scenario is used as the input of the convolutional neural network, and the optimal target set is used as the output of the convolutional neural network. The convolutional neural network is trained to obtain the optimization target screening model;

[0052] The optimization target screening model is:

[0053] Z best =CNN(S);

[0054] Where Z best is the optimal target set, S is the operating scenario, and CNN is the convolutional neural network.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] The present invention determines the optimal optimization target for the integrated dispatch of the distribution network source, grid, load and storage under various operating scenarios through indicator calculation. The dispatch data obtained from the optimal optimization target realizes the optimal dispatch of the distribution network source, grid, load and storage, thereby realizing the objective selection of the optimization target that achieves the effect of the optimal dispatch of the distribution network source, grid, load and storage. The convolutional neural network is then used to perform mapping training on the operating scenario and the optimal optimization target, and an optimization target screening model is obtained that determines the optimal optimization target according to the operating scenario. The model is used to objectively screen the optimization target, thereby improving the efficiency of the optimal dispatch of the distribution network source, grid, load and storage. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other implementation drawings based on the provided drawings without inventive effort.

[0058] Figure 1 A flow chart of the source-grid-load-storage integrated optimization analysis method based on machine learning provided in an embodiment of the present invention;

[0059] Figure 2 Block diagram of the source-grid-load-storage integrated optimization and analysis system based on machine learning provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0060] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0061] like Figure 1 As shown, the present invention provides a source-grid-load-storage integrated optimization analysis method based on machine learning, comprising the following steps:

[0062] Collect multiple operating scenarios of the distribution network and multiple optimization objectives for the integrated dispatch of distribution network sources, grids, loads and storage;

[0063] Calculate the optimization effect with each optimization target in each operation scenario in turn, and obtain the effect index achieved under each optimization target in each operation scenario;

[0064] Based on the performance indicators, determine the optimal optimization target for the integrated dispatch of distribution network sources, grids, loads and storage in each operation scenario;

[0065] A convolutional neural network is used to map the operating scenarios and the optimal optimization targets, and an optimization target screening model is obtained that determines the optimal optimization target according to the operating scenarios.

[0066] In order to achieve the best scheduling of the integrated source, grid, load and storage scheduling of the distribution network in any operating scenario, that is, to enable the distribution network to have the best operation after the integrated source, grid, load and storage scheduling in any operating scenario, and to achieve the highest operating effects of voltage stability, wind and solar reliability, and power supply quality, the present invention uses voltage stability, wind and solar reliability, and power supply quality as evaluation indicators to guide the determination of the most suitable or best distribution network integrated source, grid, load and storage scheduling optimization target (that is, the optimal optimization target) in the operating scenario, and uses the scheduling data obtained by the most suitable or best optimization target to optimize the integrated source, grid, load and storage scheduling of the distribution network to achieve the best operating effect of the distribution network.

[0067] When determining the optimal optimization target, the present invention uses a BP neural network to train an optimization effect measurement model due to the diverse combinations of operating scenarios and optimization targets. This can automatically calculate the operating effect (voltage stability, wind and solar reliability, and power supply quality) that the distribution network can obtain under the operating scenario based on the combination of the optimization targets, avoiding manual solution of all combinations of operating scenarios and optimization targets one by one, and quickly obtaining the optimal optimization target under the operating scenario, thereby improving efficiency.

[0068] Furthermore, the present invention also uses a convolutional neural network to directly train the mapping relationship between the operating scenario and the optimal optimization target, forming an optimization target screening model, and realizing the direct calculation of the optimal optimization target of the distribution network for the distribution network source, grid, load and storage integrated scheduling under the operating scenario based on the operating scenario, which is used for the distribution network to perform the distribution network source, grid, load and storage integrated scheduling optimization calculation under the operating scenario, obtain the most suitable distribution network source, grid, load and storage integrated scheduling data, and finally realize the optimal operation of the distribution network source, grid, load and storage integration, and achieve the highest operating effect of voltage stability, wind and solar reliability, and power supply quality.

[0069] Therefore, the present invention can use the model to automatically calculate the optimal optimization target of the distribution network's source-grid-load-storage integrated scheduling under any operating scenario, realize the distribution network's source-grid-load-storage integrated scheduling planning under any operating scenario, and achieve the highest distribution network operation effect in terms of voltage stability, wind and solar reliability, and power supply quality. In other words, the present invention objectively realizes the distribution network's source-grid-load-storage integrated scheduling planning under any operating scenario, provides a foundation for the distribution network's source-grid-load-storage integrated scheduling under any operating scenario, and helps to realize the optimization of the distribution network's source-grid-load-storage integrated scheduling under any operating scenario.

[0070] The performance indicators include voltage stability, wind and solar reliability, and power quality. The quantitative formula for voltage stability is:

[0071]

[0072] Where GU is the voltage stability of the distribution network, U i,t is the voltage value of the ith node in the distribution network at the tth moment, avg(U i ) is the mean voltage of the i-th node in the distribution network participating in the source-grid-load-storage integrated dispatch, N is the total number of nodes in the distribution network, and T is the total operation time of the source-grid-load-storage integrated dispatch of the distribution network;

[0073] The quantitative formula for wind and solar reliability is:

[0074]

[0075] Where GP is the wind and solar reliability rate of the distribution network, P i,t is the power consumption of the i-th node in the distribution network at the t-th moment, P wt,t is the power generation of the wind system at the tth moment, P pv,t is the power generation of the photovoltaic system at the tth moment, N is the total number of nodes in the distribution network, and T is the total operation time of the integrated source-grid-load-storage scheduling of the distribution network;

[0076] The quantitative formula for power supply quality is:

[0077]

[0078] Where GC is the power quality of the distribution network, P j,t is the power consumption of the jth load in the distribution network at the tth moment, {P j,t |t∈[1,T]} is the power consumption of the jth load in the distribution network from time 1 to time T, max({P j,t |t∈[1,T]}) is {P j,t |The maximum value in t∈[1,T]}, min({P j,t |t∈[1,T]}) is {P j,t |The minimum value in t∈[1,T]}, avg({P j,t |t∈[1,T]}) is {P j,t |t∈[1,T]}, M is the total number of loads in the distribution network, and T is the total operation time of the integrated dispatching of source, grid, load and storage in the distribution network.

[0079] The methods for measuring the optimization effect with each optimization objective in each operating scenario include:

[0080] Randomly select multiple running scenarios as sample scenarios;

[0081] In each sample scenario, multiple optimization objectives are randomly selected and combined as the target set for source-grid-load-storage integrated scheduling in that sample scenario;

[0082] Minimize each target set of each sample scenario one by one to obtain the source-grid-load-storage integrated scheduling data under the sample scenario;

[0083] Based on the source-grid-load-storage integrated dispatch data, the distribution network in the sample scenario is dispatched in an integrated manner. The voltage stability, wind-solar reliability, and power quality of the distribution network in the sample scenario are quantified between time 1 and time T.

[0084] The sample scenario and target set are used as the input of the BP neural network, and the voltage stability of the distribution network, the wind and solar reliability rate of the distribution network, and the power quality of the distribution network are used as the output of the BP neural network. The BP neural network is trained to obtain an optimization effect measurement model;

[0085] The optimization effect calculation model includes:

[0086] (GU,GP,GC)=BP(S,Z);

[0087] Where GU is the voltage stability of the distribution network, GP is the wind and solar reliability rate of the distribution network, GC is the power quality of the distribution network, S is the operation scenario, and Z is the target set;

[0088] Based on the optimization effect measurement model, the optimization effect of each target set formed by all optimization targets in each operation scenario is calculated to obtain the voltage stability of the distribution network, the wind and solar reliability rate of the distribution network, and the power supply quality of the distribution network corresponding to each target set formed by all optimization targets in each operation scenario.

[0089] When determining the optimal optimization target, the present invention uses a BP neural network to train an optimization effect measurement model due to the diverse combinations of operating scenarios and optimization targets. This can automatically calculate the operating effect (voltage stability, wind and solar reliability, and power supply quality) that the distribution network can obtain under the operating scenario based on the combination of the optimization targets, avoiding manual solution of all combinations of operating scenarios and optimization targets one by one, and quickly obtaining the optimal optimization target under the operating scenario, thereby improving efficiency.

[0090] The present invention provides a measurement example. First, the scenario {A, B, C, D, E, F, G...} is run, and the target {FA, FB, FC, FD, FE, FF} is optimized. The running scenario {A, B, C, D} is used as a sample scenario. The target set composed of the combination of the optimization targets includes {FA, FB}, {FA, FC}, {FA, FD}, {FA, FE}, {FA, FF}, {FB, FC}, {FB, FD}, {FB, FE}, {FB, FF}, {FC, FD}, {FC, FE}, {FC, FF}, {F D, FE}, {FD, FF}, {FE, FF}, {FA, FB, FC}, {FA, FB, FD}, {FA, FB, FE}, …, {FA, FB, FC, FD, FE, FF}. In sample scenario A, multiple targets are randomly selected from the target set for solution, and the voltage stability of the distribution network, the wind and solar reliability rate of the distribution network, and the power quality of the distribution network corresponding to each target set are calculated. Similarly, in sample scenarios B, C, and D, multiple targets are randomly selected from the target set for solution, and the voltage of the distribution network corresponding to each target set is calculated. Stability, wind and solar reliability of the distribution network and power quality of the distribution network constitute a data set to train the BP neural network, so that the BP neural network can output the voltage stability of the distribution network, wind and solar reliability of the distribution network and power quality of the distribution network according to the sample scenario and the target set. For example, the sample scenario B and {FA, FB} are input into the BP neural network to obtain the voltage stability of the distribution network, wind and solar reliability of the distribution network and power quality of the distribution network corresponding to the sample scenario B and {FA, FB}. In this way, the sample scenarios A, B, C, D and the target set can be solved one by one. The corresponding voltage stability of the distribution network, the wind and solar reliability rate of the distribution network, and the power supply quality of the distribution network are obtained. For example, the voltage stability of the distribution network, the wind and solar reliability rate, and the power supply quality of the distribution network corresponding to sample scenario A and {FA, FB, FE} are the highest. Then the optimal optimization target of sample scenario A is FA, FB, FE. The voltage stability of the distribution network, the wind and solar reliability rate, and the power supply quality of the distribution network corresponding to sample scenario B and {FA, FB, FE, FF} are the highest. Then the optimal optimization target of sample scenario B is FA, FB, FE, FF.

[0091] After obtaining the optimal optimization targets of sample scenario A as FA, FB, FE, the optimal optimization targets of sample scenario B as FA, FB, FE, FF, and the optimal optimization targets of sample scenarios C and D, they are used to train the convolutional neural network and establish a mapping relationship between the operating scenarios and the optimal optimization targets. This allows the convolutional neural network to input sample scenario A and output the optimal optimization targets as FA, FB, FE. This can be extended to more operating scenarios, and sample scenarios E, F, G... can be input to obtain the optimal optimization targets of E, F, G...

[0092] The methods for determining the optimal optimization target include:

[0093] In each operation scenario, the target set corresponding to the highest stability, the highest wind-solar reliability rate, and the highest power supply quality is selected as the optimal target set for that operation scenario;

[0094] All optimization targets in the optimal target set are taken as the optimal optimization targets in this operating scenario.

[0095] Methods for constructing optimized target screening models include:

[0096] The running scenario is used as the input of the convolutional neural network, and the optimal target set is used as the output of the convolutional neural network. The convolutional neural network is trained to obtain the optimization target screening model;

[0097] The optimization target screening model is:

[0098] Z best =CNN(S);

[0099] Where Z best is the optimal target set, S is the operating scenario, and CNN is the convolutional neural network.

[0100] The present invention also uses a convolutional neural network to directly train the mapping relationship between the operating scenario and the optimal optimization target, forming an optimization target screening model, and realizing the direct calculation of the optimal optimization target of the distribution network for the distribution network source, grid, load and storage integrated scheduling under the operating scenario based on the operating scenario. The distribution network is used to perform the distribution network source, grid, load and storage integrated scheduling optimization calculation under the operating scenario, obtain the most suitable distribution network source, grid, load and storage integrated scheduling data, and ultimately realize the optimal operation of the distribution network source, grid, load and storage integration, achieving the highest operating effect of voltage stability, wind and solar reliability, and power supply quality.

[0101] The operating scenarios are formed by the distribution network being connected to the wind system, photovoltaic system, and energy storage system. Among them, the distribution network being connected to the wind system and photovoltaic system forms operating scenario A, the distribution network being connected to the wind system and energy storage system forms operating scenario B, the distribution network being connected to the photovoltaic system and energy storage system forms operating scenario C, and the distribution network being connected to the wind system, photovoltaic system, and energy storage system forms operating scenario D.

[0102] Later, it can also be connected to thermal power generation systems, hydropower systems, gas turbine systems, ground source heat pump systems, etc., to form more operating scenarios {E, F, G…}.

[0103] Optimization objectives include: electricity purchase costs from the upper-level power grid, network loss costs, electricity purchase costs from wind and photovoltaic systems, operation and maintenance costs of wind and photovoltaic systems, and operation and maintenance costs of energy storage systems;

[0104] Among them, the cost of purchasing electricity from the superior power grid is: Where FA is the cost of purchasing electricity from the upper grid, c A,t is the electricity purchase price at the distribution network location at the tth moment, P A,t is the power purchased by the distribution network from the upper power grid at the tth moment, and T is the total operation time of the distribution network's source-grid-load-storage integrated dispatching;

[0105] The network loss cost is: Where FB is the network loss cost, c B,t is the electricity price at the distribution network location at the tth moment, P B,t is the power loss of the distribution network at the tth moment;

[0106] The cost of purchasing electricity from wind power systems and photovoltaic systems is: Where FC is the electricity purchase cost from wind power system and photovoltaic system, c wt,C,t is the electricity price of the wind power system at the tth moment, c pv,C,t is the electricity price of the photovoltaic system at the tth moment, P wt,t is the power generation of the wind system at the tth moment, P pv,t is the power generation of the photovoltaic system at the tth moment;

[0107] The operation and maintenance cost of the energy storage system is: Where FD is the operation and maintenance cost of the energy storage system, c D is the operation and maintenance cost coefficient of the energy storage system, P in,t is the charging power of the energy storage system at the tth moment, P out,t is the discharge power of the energy storage system at the tth moment;

[0108] The operation and maintenance costs of wind power systems and photovoltaic systems are: Where FE is the operation and maintenance cost of wind power system and photovoltaic system, c wt,F is the operation and maintenance cost coefficient of the wind power system, c pv,F is the operation and maintenance cost coefficient of the photovoltaic system.

[0109] In actual use, additional optimization targets can be added, such as the comprehensive daily operating cost of the distribution network, carbon trading costs, etc.

[0110] Source, grid, load and storage integrated dispatch data includes P A,t , P B,t , P wt,t , P pv,t , P in,t , P out,t , where P A,tP is the power purchased by the distribution network from the upper grid at the tth moment, B,t is the power loss of the distribution network at the tth moment, P wt,t is the power generation of the wind system at the tth moment, P pv,t is the power generation of the photovoltaic system at the tth moment, P in,t is the charging power of the energy storage system at the tth moment, P out,t is the discharge power of the energy storage system at the tth moment.

[0111] like Figure 2 As shown, the present invention provides a source-grid-load-storage integrated optimization analysis method based on machine learning, which is applied to a source-grid-load-storage integrated optimization analysis method based on machine learning. The system includes:

[0112] The data acquisition unit is used to collect multiple operating scenarios of the distribution network and multiple optimization objectives of the integrated dispatching of the distribution network's source, grid, load and storage;

[0113] A data processing unit is used to measure the optimization effect with each optimization target in each operation scenario in turn, obtain the effect index achieved under each optimization target in each operation scenario, and determine the optimal optimization target for the integrated dispatch of the distribution network source, grid, load and storage in each operation scenario based on the effect index;

[0114] The deep learning unit is used to use a convolutional neural network to perform mapping training between the operating scenario and the optimal optimization target, and obtain an optimization target screening model that determines the optimal optimization target based on the operating scenario.

[0115] The deep learning unit's method for constructing an optimization target screening model includes:

[0116] The running scenario is used as the input of the convolutional neural network, and the optimal target set is used as the output of the convolutional neural network. The convolutional neural network is trained to obtain the optimization target screening model;

[0117] The optimization target screening model is:

[0118] Z best =CNN(S);

[0119] Where Z best is the optimal target set, S is the operating scenario, and CNN is the convolutional neural network.

[0120] The present invention determines the optimal optimization target for the integrated dispatch of the distribution network source, grid, load and storage under various operating scenarios through indicator calculation. The dispatch data obtained from the optimal optimization target realizes the optimal dispatch of the distribution network source, grid, load and storage, thereby realizing the objective selection of the optimization target that achieves the effect of the optimal dispatch of the distribution network source, grid, load and storage. The convolutional neural network is then used to perform mapping training on the operating scenario and the optimal optimization target, and an optimization target screening model is obtained that determines the optimal optimization target according to the operating scenario. The model is used to objectively screen the optimization target, thereby improving the efficiency of the optimal dispatch of the distribution network source, grid, load and storage.

[0121] The above embodiments are merely exemplary embodiments of the present application and are not intended to limit the scope of the present application. The scope of protection of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and scope of protection of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the scope of protection of the present application.

Claims

1. A source-grid-load-storage integrated optimization analysis method based on machine learning, characterized in that: The following steps are involved: Collect multiple operating scenarios of the distribution network and multiple optimization objectives for the integrated dispatch of distribution network sources, grids, loads and storage; Calculate the optimization effect with each optimization target in each operation scenario in turn, and obtain the effect index achieved under each optimization target in each operation scenario; Based on the effect indicators, determine the optimal optimization target for the integrated dispatch of distribution network sources, grids, loads and storage under various operation scenarios; A convolutional neural network is used to map the operating scenarios and the optimal optimization targets, and an optimization target screening model is obtained that determines the optimal optimization target according to the operating scenarios.

2. The machine learning-based source-grid-load-storage integrated optimization analysis method according to claim 1, characterized in that: The effect indicators include voltage stability, wind and solar reliability, and power quality. The quantitative formula for voltage stability is: Where GU is the voltage stability of the distribution network, U i,t is the voltage value of the ith node in the distribution network at the tth moment, avg(U i ) is the voltage mean of the i-th node in the distribution network participating in the source-grid-load-storage integrated dispatch, N is the total number of nodes in the distribution network, and T is the total operation time of the source-grid-load-storage integrated dispatch of the distribution network; The quantitative formula of wind-solar reliability is: Where GP is the wind and solar reliability rate of the distribution network, P i,t is the power consumption of the i-th node in the distribution network at the t-th moment, P wt,t is the power generation of the wind system at the tth moment, P pv,t is the power generation of the photovoltaic system at the tth moment, N is the total number of nodes in the distribution network, and T is the total operation time of the integrated source-grid-load-storage scheduling of the distribution network; The quantitative formula for the power supply quality is: Where GC is the power quality of the distribution network, P j,t is the power consumption of the jth load in the distribution network at the tth moment, {P j,t |t∈[1,T]} is the power consumption of the jth load in the distribution network from time 1 to time T, max({P j,t |t∈[1,T]}) is {P j,t |The maximum value in t∈[1,T]}, min({P j,t |t∈[1,T]}) is {P j,t |The minimum value in t∈[1,T]}, avg({P j,t |t∈[1,T]}) is {P j,t |t∈[1,T]}, M is the total number of loads in the distribution network, and T is the total operation time of the integrated dispatching of source, grid, load and storage in the distribution network.

3. The machine learning-based optimization analysis method for source-grid-load-storage integration according to claim 2, characterized in that: The methods for measuring the optimization effect with each optimization objective in each operating scenario include: Randomly select multiple running scenarios as sample scenarios; In each sample scenario, multiple optimization objectives are randomly selected and combined as the target set for source-grid-load-storage integrated scheduling in that sample scenario; Minimize each target set of each sample scenario one by one to obtain the source-grid-load-storage integrated scheduling data under the sample scenario; Based on the source-grid-load-storage integrated scheduling data, the distribution network in the sample scenario is dispatched in an integrated manner, and the voltage stability, wind-solar reliability and power quality of the distribution network in the sample scenario are quantified between time 1 and time T; The sample scenario and target set are used as the input of the BP neural network, and the voltage stability of the distribution network, the wind and solar reliability rate of the distribution network, and the power quality of the distribution network are used as the output of the BP neural network. The BP neural network is trained to obtain an optimization effect measurement model; The optimization effect calculation model includes: (GU,GP,GC)=BP(S,Z); Where GU is the voltage stability of the distribution network, GP is the wind and solar reliability rate of the distribution network, GC is the power quality of the distribution network, S is the operation scenario, and Z is the target set; Based on the optimization effect measurement model, the optimization effect of each target set formed by all optimization targets in each operation scenario is calculated to obtain the voltage stability of the distribution network, the wind and solar reliability rate of the distribution network, and the power supply quality of the distribution network corresponding to each target set formed by all optimization targets in each operation scenario.

4. The machine learning-based optimization analysis method for source-grid-load-storage integration according to claim 3, characterized in that: The method for determining the optimal optimization target includes: In each operation scenario, the target set corresponding to the highest stability, the highest wind-solar reliability rate, and the highest power supply quality is selected as the optimal target set for that operation scenario; All optimization targets in the optimal target set are taken as the optimal optimization targets in the operation scenario.

5. The machine learning-based optimization analysis method for source-grid-load-storage integration according to claim 4, characterized in that: The method for constructing the optimization target screening model includes: The running scenario is used as the input of the convolutional neural network, and the optimal target set is used as the output of the convolutional neural network. The convolutional neural network is trained to obtain the optimization target screening model; The optimization target screening model is: Z best =CNN(S); Where Z best is the optimal target set, S is the operating scenario, and CNN is the convolutional neural network.

6. The machine learning-based optimization analysis method for source-grid-load-storage integration according to claim 5, characterized in that: The operation scenario is formed by the distribution network being connected to the wind power system, photovoltaic system, and energy storage system.

7. The machine learning-based optimization analysis method for source-grid-load-storage integration according to claim 6, characterized in that: The optimization objectives include: the cost of purchasing electricity from the upper power grid, network loss costs, the cost of purchasing electricity from the wind system and photovoltaic system, the operation and maintenance costs of the wind system and photovoltaic system, and the operation and maintenance costs of the energy storage system; Among them, the cost of purchasing electricity from the superior power grid is: Where FA is the cost of purchasing electricity from the upper grid, c A,t is the electricity purchase price at the distribution network location at the tth moment, P A,t is the power purchased by the distribution network from the upper power grid at the tth moment, and T is the total operation time of the distribution network's source-grid-load-storage integrated dispatching; The network loss cost is: Where FB is the network loss cost, c B,t is the electricity price at the distribution network location at the tth moment, P B,t is the power loss of the distribution network at the tth moment; The cost of purchasing electricity from wind power systems and photovoltaic systems is: Where FC is the electricity purchase cost from wind power system and photovoltaic system, c wt,C,t is the electricity price of the wind power system at the tth moment, c pv,C,t is the electricity price of the photovoltaic system at the tth moment, P wt,t is the power generation of the wind system at the tth moment, P pv,t is the power generation of the photovoltaic system at the tth moment; The operation and maintenance cost of the energy storage system is: Where FD is the operation and maintenance cost of the energy storage system, c D is the operation and maintenance cost coefficient of the energy storage system, P in,t is the charging power of the energy storage system at the tth moment, P out,t is the discharge power of the energy storage system at the tth moment; The operation and maintenance costs of wind power systems and photovoltaic systems are: Where FE is the operation and maintenance cost of wind power system and photovoltaic system, c wt,F is the operation and maintenance cost coefficient of the wind power system, c pv,F is the operation and maintenance cost coefficient of the photovoltaic system.

8. The machine learning-based optimization analysis method for source-grid-load-storage integration according to claim 7, characterized in that: The source-grid-load-storage integrated dispatching data includes P A,t , P B,t , P wt,t , P pv,t , P in,t , P out,t , where P A,t P is the power purchased by the distribution network from the upper grid at the tth moment, B,t is the power loss of the distribution network at the tth moment, P wt,t is the power generation of the wind system at the tth moment, P pv,t is the power generation of the photovoltaic system at the tth moment, P in,t is the charging power of the energy storage system at the tth moment, P out,t is the discharge power of the energy storage system at the tth moment.

9. A source-grid-load-storage integrated optimization analysis method based on machine learning, characterized in that: A method for optimizing and analyzing source-grid-load-storage integration based on machine learning, as described in any one of claims 1 to 8, wherein the system comprises: The data acquisition unit is used to collect multiple operating scenarios of the distribution network and multiple optimization objectives of the integrated dispatch of the distribution network's source, grid, load and storage; A data processing unit is used to measure the optimization effect with each optimization target in each operation scenario in turn, obtain the effect index achieved under each optimization target in each operation scenario, and determine the optimal optimization target for the integrated dispatch of the distribution network source, grid, load and storage in each operation scenario based on the effect index; The deep learning unit is used to use a convolutional neural network to perform mapping training between the operating scenario and the optimal optimization target, and obtain an optimization target screening model that determines the optimal optimization target based on the operating scenario.

10. The machine learning-based source-grid-load-storage integrated optimization and analysis system according to claim 9, characterized in that: The deep learning unit's method for constructing an optimization target screening model includes: The running scenario is used as the input of the convolutional neural network, and the optimal target set is used as the output of the convolutional neural network. The convolutional neural network is trained to obtain the optimization target screening model; The optimization target screening model is: Z best =CNN(S); Where Z best is the optimal target set, S is the operating scenario, and CNN is the convolutional neural network.

Images