Dispersed multi-element energy storage model aggregation method considering dynamic boundary adjustment

Abstract

The invention belongs to the technical field of model aggregation methods in the field of power system dispatching, and particularly provides a dynamic boundary adjustment-considered dispersed multivariate energy storage model aggregation method, which comprises the following steps of: realizing typical scene generation among three variables of wind power, photovoltaic and region on the basis of an R vine-Copula structure of a Frank-Copula function and a K-means algorithm on the basis of historical output; establishing a cluster division index based on energy storage multi-time scale output characteristics, and using an improved community discovery algorithm to realize space cluster division of dispersed energy storage; determining an adjustment direction of an energy storage model aggregation process according to cluster energy storage internal characteristics; a mapping function is established to realize coupling of cluster internal parameters and an equivalent energy storage state, so that a model boundary dynamically changes to adapt to internal cluster characteristics, and a cluster energy storage aggregation model is applied to multi-time scale scheduling to realize full utilization of various types of scenes by the aggregation model. According to the method, effective aggregation and utilization of dispersed multi-element energy storage resources in a provincial level scheduling link can be achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of model aggregation methods in the field of power system dispatching, and specifically relates to a decentralized multivariate energy storage model aggregation method considering dynamic boundary adjustment. Background Art

[0002] The large-scale and diversified development of energy storage resources provides more technical means for the flexible operation of power systems and the smoothing of wind and solar fluctuations. However, the characteristics of dispersed resources and different models at the provincial level have brought difficulties to the system optimization and scheduling at the provincial level.

[0003] At present, there are two main types of model aggregation problems for decentralized resources: mathematical aggregation and optimization algorithms. The application of data aggregation methods in power systems tends to ignore the impact of electrical distance, and the optimization algorithm mainly converts the aggregation of models into optimization problems or uses algorithms for processing, but it is easy to increase the processing pressure of the system. Due to the diversified development and decentralized layout of energy storage, it will be difficult to aggregate, accurately characterize, and fully apply energy storage models to system dispatch at the provincial level.

[0004] Therefore, how to reasonably aggregate and finely describe aggregated energy storage is of great significance to the rapid scheduling and resource utilization of the system. It is urgent to establish a complete aggregation method to adapt to the multi-time scale optimal scheduling of the system. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a distributed multi-element energy storage model aggregation method taking into account dynamic boundary adjustment, and to establish a model scheduling basis for the optimal scheduling of the system by solving problems such as the division of distributed energy storage resources, model aggregation and precise description.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: a decentralized multi-element energy storage model aggregation method considering dynamic boundary adjustment, comprising the following steps: S1. Based on the historical output data of wind power and photovoltaic power in various regions, the wind power and photovoltaic output scenarios of various regions are generated using the R-vine-Copula structure based on the Frank-Copula function, and the typical wind and photovoltaic output scenarios are obtained by K-means reduction; S2. Based on the multi-time scale output characteristics of energy storage, considering the electrical distance of energy storage, establishing cluster division indicators, and using the improved community discovery method to perform cluster space division on distributed energy storage, mainly improving the modularity formula of each node aggregation process; S3, taking into account the effect of energy storage on smoothing net load fluctuations and the output characteristics of cluster energy storage, determine the adjustment direction of the energy storage model in the aggregation process, and then perform model aggregation; S4. Establish a mapping function to achieve the coupling between the internal state of cluster energy storage and external model parameters, make the model boundary change dynamically, and apply the cluster energy storage aggregation model to multi-time scale scheduling.

[0007] In the preferred solution, in step S1, the Frank-Copula function-based R-Copula structure is used to realize the generation of the joint typical wind-solar scenario between the three variables of wind power, photovoltaic power and region, and the joint probability density function of each variable is: (1); In the formula, Represents wind power and photovoltaic random variables, with a total of n indivual; Representation variables Through the probability density function after fitting; Indicates the edge of R vine e Corresponding variables j and k Copula function; represents the edge set of the vine, where Indicates the nodes corresponding to the two ends of the edge, Represents a set of conditional variables; Representation variables x On the side e of The probability distribution of the end, Representation variables x On the side e of The probability distribution of the end.

[0008] In the preferred solution, in step S2, the electrical distance between each energy storage node is used as an indicator Q 1. Then, we established evaluation indexes at three levels: response rate, duration and power support, which are response coefficient Q 2. Energy efficiency coefficient Q 3 and support factor Q 4. The expressions are: (2); (3); (4); In the formula, Represents the net load curve of each node in the system, It indicates the fluctuation of net load per unit time. Indicates k The ramp rate of the energy storage in MW / h; , and Respectively represent k The initial energy, minimum and maximum capacity limits of each energy storage unit, in MWh; , Respectively represent the maximum values ​​of the corresponding energy storage discharge and charging power; , Indicates energy storage k The charging and discharging efficiency; Indicates the corresponding electricity cost; Indicates the corresponding k The maximum load fluctuation of the node indicates the contribution of energy storage to the smoothing of net load fluctuations in the system supply and demand balance after considering the impact of wind power and photovoltaic output of the system; Indicates the maximum charging or discharging power of energy storage device k, in MW; Q The weight of 1 indicator is 1, which is determined by the entropy weight method. Q 2. Q 3 and Q 4 The weights of the three indicators under the corresponding data are calculated, and the improved community discovery method is used to iteratively optimize the four cluster division indicators including the electrical distance indicator, and the modularity indicator used for the iteration of the community discovery method is obtained by comprehensive calculation Q : (5); In the formula, , and They represent the relative weight values ​​of the three indicators calculated by the entropy weight method; express i and j The state function of the partition process in the two nodes of the system through the community discovery method, when they are divided into the same cluster, And by setting the reward coefficient Increase the value of this modularity; when Indicates that the two nodes are not divided into the same cluster; in the process of calculating modularity, each node calculates the modularity increment by moving closer to the adjacent nodes ,If the modularity continues to increase during the movement process, the process will continue and each node will carry out this process until the modularity of each node cannot be further increased.

[0009] In a preferred solution, in step S3, based on the divided energy storage clusters, the energy storage capacity with larger capacity is transferred to keep the energy storage in a balanced state while charging and discharging. The expression is: (6); , (7); In the formula, and They represent the amount of energy storage for power output adjustment and capacity adjustment within a cluster energy storage. Indicates k The maximum discharge time of a stored energy, It represents the same discharge duration of cluster energy storage in the energy balance state. Indicates energy storage k The charging or discharging efficiency, , , Respectively represent k The maximum charge and discharge power of the energy storage i The maximum charge and discharge power of the energy storage k The maximum storable energy of an energy storage.

[0010] In the preferred solution, in step S4, a mapping function is established that can effectively map the capacity short board of the cluster energy storage, as shown in formula (8). Through this mapping function, the influence of the cost of electricity within the cluster energy storage on the power redistribution within the cluster can be simultaneously referred to, and the boundary dynamic change adjustment direction of the aggregated energy storage model can be determined; the energy constraint of the aggregated energy storage is improved as shown in formula (9), and the energy boundary of the aggregated energy storage is dynamically adjusted so that the aggregated energy storage can be adaptively adjusted according to the capacity boundary change during the time sequence scheduling process: (8); (9); In the formula, Indicates that the capacity boundary of the aggregation model is i The amount of adjustment at the moment; n Represents the amount of energy storage inside the aggregation model; Indicates energy storage k The unit cost of electricity is RMB / MW; Indicates that aggregate storage i The power value obtained by scheduling optimization at the moment; Indicates that the cluster energy storage model is i The capacity boundary deviation adjustment coefficient at the time; , , Respectively represent the storage i The energy stored at the moment, and the corresponding minimum and maximum energy boundaries.

[0011] The present invention provides a decentralized multi-element energy storage model aggregation method considering dynamic boundary adjustment, which has the following beneficial effects: 1. Compared with the traditional analysis of typical wind and solar scenes based on the Frank-Copula function, step S1 of the present invention realizes the wide-area regional division of decentralized energy storage resources, takes into account the geographical correlation factors of wind power and photovoltaics, realizes time coupling of wind power and photovoltaic output, and enhances the accuracy of the description of typical wind and solar scenes.

[0012] 2. Step S2 of the present invention considers the differences in ramp rate, power and capacity caused by the heterogeneous characteristics of multiple energy storage, and spatially divides the cluster energy storage by comprehensively considering various evaluation indicators. The improved community discovery method is used to effectively reduce the computing power requirements of the division process, while compensating for the output differences between the energy storages, thereby minimizing the differences in quantity and operating characteristics between the multiple heterogeneous energy storages.

[0013] 3. In step S3 of the present invention, considering the influence of the barrel effect of each multi-heterogeneous energy storage in the aggregation process, a "capacity transfer" strategy is used to adjust the capacity of the energy storage with larger capacity, and the energy storage equipment with smaller power reduces the power output, so as to ensure the operation of the aggregation model as much as possible so that the capacity of each energy storage is filled or the energy is exhausted at the same time, thereby reducing the restriction of the energy state on the power in the aggregation model, and at the same time determining the adjustment amount based on the division index, so that the external representation conversion of the aggregation model is targeted.

[0014] 4. Step S4 of the present invention is to ensure that the instructions assigned to the aggregation model during the scheduling process can be smoothly assigned to the cluster energy storage. The externally scheduled power instructions are coupled with the actual internal total operating characteristics through a mapping function, thereby achieving dynamic adjustment and fine characterization of the aggregation model, and further reducing the impact of the "barrel effect" within the cluster. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 It is a flow chart of the aggregation of the distributed energy storage model of the present invention; Figure 2 Typical photovoltaic output scenarios generated and reduced for the present invention; Figure 3 Typical wind power output scenarios generated and curtailed for the present invention; Figure 4 Divide processes for cluster energy storage; Figure 5 It is an explanation of the balanced aggregation of multi-heterogeneous energy storage SOC; Figure 6 It explains the operating mechanism of “capacity adjustment” for multi-heterogeneous energy storage; Figure 7 It is a cluster energy storage division result diagram in a specific embodiment; Figure 8It is a diagram showing the variation of sequential charging simulation parameters under different polymerization methods in a specific embodiment. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0017] In response to the problems mentioned in the above background technology, this embodiment mainly focuses on the spatial division and model aggregation method of cluster energy storage under the decentralized layout of multi-heterogeneous energy storage at the provincial level, and mainly develops from four steps: generation of typical cross-temporal and spatial scenarios of wind and solar power, cluster division of multi-heterogeneous energy storage, model aggregation method considering the "barrel effect", and internal mapping function of the aggregation model.

[0018] Among them, the generation of typical wind and solar cross-temporal and spatial scenarios is mainly based on the vine-Copula structure of the Frank-Copula function for scenario generation, and the K-means algorithm is used to reduce and obtain typical wind and solar output scenarios in different regions; the clustering of multi-heterogeneous energy storage mainly uses the improved community discovery algorithm, by establishing multi-time output indicators of energy storage, considering the influence of the electrical distance of each energy storage in the power grid, and using the obtained comprehensive modularity to achieve spatial clustering of energy storage; since energy storage is a controllable resource, the aggregation process can consider the influence of the "barrel effect" to aggregate the model, and determine the "capacity adjustment" to power type or capacity type according to the evaluation indicators of the previous cluster. When the aggregation model is stable, the aggregation model based on energy balance is used; at the same time, in order to ensure that the power instructions obtained by the aggregation model can be effectively distributed to the internal cluster, a mapping function that can dynamically characterize the internal state of the cluster energy storage is established, and by fine-tuning the capacity boundary, it is ensured that the aggregation model can effectively and accurately describe the cluster energy storage.

[0019] like Figure 1 As shown, a decentralized multi-element energy storage model aggregation method considering dynamic boundary adjustment includes the following steps: S1. Based on the historical output of wind power and photovoltaic power in various regions, the R-vine-Copula structure based on the Frank-Copula function is used to generate scenarios between the three variables of wind power, photovoltaic power and space, and the typical wind and solar power output scenarios are obtained through K-means reduction to establish a data analysis basis for the spatial division of each node.

[0020] The specific joint probability density function of each variable is: (1); In the formula, , Parameters such as wind power and photovoltaic power represent random variables.n indivual; Representation variables Through the probability density function after fitting; Indicates the edge of R vine e Corresponding variables j and k Copula function; Represents vine e The edge set of Indicates the nodes corresponding to the two ends of the edge, Represents a set of conditional variables; Representation variables x On the side e of The probability distribution of the end, Representation variables x On the side e of The probability distribution of the end.

[0021] The R-vine-Copula structure is used to realize the generation of wind power and photovoltaic cross-temporal and spatial scenarios, and then the K-means algorithm is used to reduce the generated scenarios to obtain typical wind and solar output scenarios. The specific generated data is as follows: Figure 2 and Figure 3 shown.

[0022] S2. Considering the uncertainty of wind power and photovoltaic power, a clustering index based on the multi-time scale output characteristics of energy storage is established, and the improved community discovery algorithm is used to achieve spatial clustering of distributed energy storage. By optimizing the spatial division of energy storage resources, the objects to be aggregated are preliminarily determined. Q 1 is used as a reference, and evaluation indexes are established at the three levels of response rate, duration and power support, which are response coefficient , Energy Efficiency Coefficient Support factor : (2); (3); (4); In the formula, Represents the net load curve of each node in the system, It indicates the fluctuation of net load per unit time. Indicates k The ramp rate of the energy storage in MW / h; , and Respectively represent kThe initial energy, minimum and maximum capacity limits of each energy storage unit, in MWh, and They represent the maximum values ​​of the corresponding energy storage discharge and charging power, and Indicates the charging and discharging efficiency of energy storage, Indicates the corresponding electricity cost; Indicates the corresponding k The maximum load fluctuation of the node indicates the contribution of energy storage to the smoothing of net load fluctuations in the system supply and demand balance after considering the impact of wind power and photovoltaic output of the system.

[0023] Then, the weights of the three indicators corresponding to the data are determined according to the entropy weight method, and the improved community discovery method is used to iteratively optimize the four cluster division indicators including the electrical distance indicator, and the modularity indicator used for the iteration of the community discovery method is obtained by comprehensive calculation. Q : (5); In the formula, Represents the electrical distance between each energy storage node, , and They represent the relative weight values ​​of the three indicators calculated by the entropy weight method; express i and j The state function of the partition process in the two nodes of the system through the community discovery method, when they are divided into the same cluster, And by setting the reward coefficient Increase the value of this modularity; when Indicates that the two nodes are not classified into the same cluster. In the process of calculating modularity, each node calculates by moving closer to the adjacent nodes. Modularity increment, if the modularity continues to increase during the movement process, continue to do this process for each node until the modularity of each node cannot be further increased. The overall cluster division process can be referred to Figure 4 shown.

[0024] S3. Considering the adaptability of each cluster energy storage to smoothing the net load of the system, according to the internal characteristics of the cluster energy storage and referring to the previous division indicators, determine whether the adjustment of the multi-energy storage equivalent model in the aggregation process is biased towards power type or energy type.

[0025] For details, please refer to Figure 5 The process of allocating instructions to the internal energy storage of the aggregated energy storage: Figure 5In the figure, the left figure shows the power changes of two electrochemical energy storage devices and gravity energy storage devices during instruction distribution; the right figure shows the details of the adjustment of the internal output of the three energy storage devices due to the boundary influence of power and capacity during the first distribution. Due to the difference in capacity, the traditional aggregation method based on the SOC balance principle leads to unreasonable power distribution due to the multiple influence of gravity, compressed air and other energy storage devices, and different initial energy states will also affect the internal distribution of power. Therefore, Figure 6 As shown in the figure, this method proposes to "transfer" the energy storage capacity with larger capacity, such as gravity energy storage. The left figure shows the adjustment of the power and capacity boundaries of the three energy storage devices during the first instruction allocation; the right figure shows the change of the external characteristic boundary of the aggregation model after adjustment. By appropriately "capacity adjustment" of the energy storage with larger capacity according to the parameter characteristics within the cluster, the charging and discharging behavior of each energy storage is kept in a balanced state: (6); (7); In the formula, and They represent the amount of energy storage for power output adjustment and capacity adjustment within a cluster energy storage. Indicates k The maximum discharge time of a stored energy, It represents the same discharge duration of cluster energy storage in the energy balance state. Indicates energy storage k By properly adjusting the power and capacity of each energy storage, the differences between the energy storage can be minimized to ensure the consistency of cluster actions, while guiding the aggregation model to transform into a power type or capacity type.

[0026] The final internal adjustment can make the aggregation model as Figure 6 As shown, the power limit of the model is less constrained by its internal energy when it is at the boundary state, creating a pre-conditioning effect for the internal allocation of the aggregate model.

[0027] S4. In order to make the aggregated energy storage model accurately describe the output characteristics of cluster energy storage and reduce the impact of the "barrel effect" within the cluster energy storage, a mapping function is established to achieve the coupling of the internal energy state of the cluster and the external model parameters, so that the model boundary changes dynamically to adapt to the internal cluster characteristics. Referring to the impact of the cost of electricity within the cluster energy storage on the power redistribution within the cluster, the direction of the dynamic change of the boundary of the aggregated energy storage model can be determined: (8); (9); In the formula, Indicates that the capacity boundary of the aggregation model is i The amount of adjustment at the moment; n Represents the amount of energy storage inside the aggregation model; Indicates energy storage k The unit cost of electricity is RMB / MW; Indicates that aggregate storage i The power value obtained by scheduling optimization at the moment; Indicates that the cluster energy storage model is i The capacity boundary deviation adjustment coefficient at the moment. The internal power state is preliminarily determined through the mapping function, and appropriate adjustments can be made for the next change in the capacity of the cluster energy storage. Specific embodiments First, using the generated wind power and photovoltaic data, the system grid structure is considered to obtain the system division results, such as Figure 7 shown.

[0029] Next, the model is aggregated, and the parameter simulation under time-sequential charging is performed to compare the traditional aggregation method based on SOC balance, and the specific parameter change characteristics of the cluster energy storage under the two aggregation methods are obtained, such as Figure 8 shown.

[0030] Figure 8 In the figure, the first figure is a simulation diagram of the example under the method proposed in this paper; the second figure is a comparative example result, which is a timing simulation diagram of the aggregation model based on the energy balance method.

[0031] Compared with the control group, the rated capacity and power of the polymer energy storage under this method dropped sharply at the fourth charge. This shows that based on this method, the devices inside the polymer energy storage can be maintained in a common charging state; they can be withdrawn at the same time in the later stage, thereby weakening the impact of the "barrel effect".

[0032] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A decentralized multivariate energy storage model aggregation method considering dynamic boundary adjustment, characterized in that: The following steps are involved: S1. Based on the historical output data of wind power and photovoltaic power in various regions, the wind power and photovoltaic output scenarios of various regions are generated using the R-vine-Copula structure based on the Frank-Copula function, and the typical wind and photovoltaic output scenarios are obtained by K-means reduction; S2. Based on the multi-time scale output characteristics of energy storage, considering the electrical distance of energy storage, establishing cluster division indicators, and using the improved community discovery method to perform cluster spatial division of distributed energy storage; S3, taking into account the effect of energy storage on smoothing net load fluctuations and the output characteristics of cluster energy storage, determine the adjustment direction of the energy storage model in the aggregation process, and then perform model aggregation; S4. Establish a mapping function to achieve the coupling between the internal state of cluster energy storage and external model parameters, make the model boundary change dynamically, and apply the cluster energy storage aggregation model to multi-time scale scheduling.

2. A decentralized multivariate energy storage model aggregation method considering dynamic boundary adjustment according to claim 1, characterized in that: In step S1, the Frank-Copula function-based R-Copula structure is used to realize the generation of the joint typical wind-solar scenario between the three variables of wind power, photovoltaic power and region, and the joint probability density function of each variable is: (1)； In the formula, Represents wind power and photovoltaic random variables, with a total of n indivual; Representation variables Through the probability density function after fitting; Indicates the edge of R vine e Corresponding variables j and k Copula function; represents the edge set of the vine, where Indicates the nodes corresponding to the two ends of the edge, Represents a set of conditional variables; Representation variables x On the side e of The probability distribution of the end, Representation variables x On the side e of The probability distribution of the end.

3. A decentralized multivariate energy storage model aggregation method considering dynamic boundary adjustment according to claim 1, characterized in that: In step S2, the electrical distance between each energy storage node is used as an indicator Q 1. Then, we established evaluation indexes at three levels: response rate, duration and power support, which are response coefficient Q 2. Energy efficiency coefficient Q 3 and support factor Q 4. The expressions are: (2)； (3)； (4)； In the formula, Represents the net load curve of each node in the system, It indicates the fluctuation of net load per unit time. Indicates k The ramp rate of the energy storage in MW / h; , and Respectively represent k The initial energy, minimum and maximum capacity limits of each energy storage unit, in MWh; , Respectively represent the maximum values ​​of the corresponding energy storage discharge and charging power; , Indicates energy storage k The charging and discharging efficiency; Indicates the corresponding electricity cost; Indicates the corresponding k The maximum load fluctuation of the node indicates the contribution of energy storage to the smoothing of net load fluctuations in the system supply and demand balance after considering the impact of wind power and photovoltaic output of the system; Indicates the maximum charging or discharging power of energy storage device k, in MW; Q The weight of 1 indicator is 1, which is determined by the entropy weight method. Q 2. Q 3 and Q 4 The weights of the three indicators under the corresponding data are calculated, and the improved community discovery method is used to iteratively optimize the four cluster division indicators including the electrical distance indicator, and the modularity indicator used for the iteration of the community discovery method is obtained by comprehensive calculation Q : (5)； In the formula, , and They represent the relative weight values ​​of the three indicators calculated by the entropy weight method; express i and j The state function of the partition process in the two nodes of the system through the community discovery method, when they are divided into the same cluster, And by setting the reward coefficient Increase the value of this modularity; when Indicates that the two nodes are not divided into the same cluster; in the process of calculating modularity, each node calculates the modularity increment by moving closer to the adjacent nodes ,If the modularity continues to increase during the movement process, the process will continue and each node will carry out this process until the modularity of each node cannot be further increased.

4. A decentralized multi-element energy storage model aggregation method considering dynamic boundary adjustment according to claim 1, characterized in that: In step S3, based on the divided energy storage clusters, the energy storage capacity with larger capacity is transferred to keep each energy storage in a balanced state while charging and discharging. The expression is: (6)； ， (7)； In the formula, and They represent the amount of energy storage for power output adjustment and capacity adjustment within a cluster energy storage. Indicates k The maximum discharge time of a stored energy, It represents the same discharge duration of cluster energy storage in the energy balance state. Indicates energy storage k The charging or discharging efficiency, , , Respectively represent k The maximum charge and discharge power of the energy storage i The maximum charge and discharge power of the energy storage k The maximum storable energy of an energy storage.

5. A decentralized multi-element energy storage model aggregation method considering dynamic boundary adjustment according to claim 1, characterized in that: In step S4, a mapping function that can effectively map the capacity short board of the cluster energy storage is established, as shown in formula (8). Through this mapping function, the influence of the cost of electricity inside the cluster energy storage on the power redistribution inside the cluster can be simultaneously referred to, and the boundary dynamic change adjustment direction of the aggregated energy storage model can be determined; the energy constraint of the aggregated energy storage is improved as shown in formula (9). By dynamically adjusting the energy boundary of the aggregated energy storage, the aggregated energy storage can be adaptively adjusted according to the capacity boundary change during the time sequence scheduling process. The expressions of formula (8) and formula (9) are respectively: (8)； (9)； In the formula, Indicates that the capacity boundary of the aggregation model is i The amount of adjustment at the moment; n Represents the amount of energy storage inside the aggregation model; Indicates energy storage k The unit cost of electricity is RMB / MW; Indicates that aggregate storage i The power value obtained by scheduling optimization at the moment; Indicates that the cluster energy storage model is i The capacity boundary deviation adjustment coefficient at the time; , , They respectively represent the energy stored in the energy storage at time i, and the corresponding minimum and maximum energy boundaries of the energy storage.

Images