Multi-time scale power balance method and system based on multi-element energy storage complementary cooperation

Abstract

The application discloses a multi-time scale electric quantity balance method and system based on multi-element energy storage complementary cooperation, relates to the electric quantity balance technical field, and comprises the following steps: acquiring a dynamic response parameter set of multi-element energy storage media, and constructing a response vector through clustering analysis; adopting a DTW algorithm DTW path distance matrix, and extracting a phase difference sequence of each energy storage medium through Hilbert transformation to obtain a candidate path set; initializing a redundant capacity allocation vector, and randomly superimposing the redundant capacity allocation vector to the candidate path set to obtain a first evaluation matrix; constructing a multi-objective optimization function, evaluating the deviation of the first evaluation matrix and a preset electric quantity balance curve through an NSGA-III algorithm, and correcting the redundant capacity allocation vector to obtain a best capacity allocation value sequence of each energy storage medium on a target time scale set, so that the problem that a traditional energy storage scheduling method does not fully consider the complex relationship between an electric quantity balance target and an energy storage path dynamic response, and the scheduling efficiency and system stability are reduced is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power balance, more particularly, the present application relates to a multi-time scale power balance method and system based on multi-element energy storage complementary cooperation. BACKGROUND

[0002] With the large-scale access of renewable energy and the continuous improvement of the demand for flexible adjustment of power systems, energy storage technology, as a key means to improve the stability and operating efficiency of power grids, has been increasingly concerned. Due to the significant differences in response speed, capacity density, and life cycle of various types of energy storage media (such as lithium batteries, supercapacitors, compressed air energy storage, flywheel energy storage, etc.), they have natural complementarity. In practical applications, how to achieve efficient cooperation between different energy storage media has become a key technical problem in the scheduling and optimization of multi-element energy storage systems.

[0003] For example, the invention patent with the announcement number: CN117595254A announces a multi-time scale power balance method applied to a power system. The multi-time scale power balance method includes a power balance optimization model and a power balance optimization process. The present application has the beneficial effect of clearing technical obstacles and laying a technical foundation for the smooth planning and construction of new power systems. It analyzes the influence of the changes of different flexible power sources, adjustable loads, and demand side response characteristics and typical fluctuation scenarios on the calculation results of traditional power balance, and finds out the problems and loopholes in traditional methods.

[0004] The above disclosed technical solution has at least the following technical problems:

[0005] Currently, traditional energy storage scheduling methods mostly use a single time scale or ignore the dynamic response characteristics of different energy storage media, making it difficult to fully exploit the compensation capabilities of multi-element energy storage at different time scales (such as millisecond, second, minute, hour, etc.). In addition, existing methods generally use fixed strategies or simple heuristic algorithms for capacity allocation, without fully considering the complex relationship between power balance targets and dynamic responses of energy storage paths, resulting in difficulties in balancing scheduling efficiency and system stability. In view of the above problems, the present application provides a solution. SUMMARY

[0006] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a multi-time scale power balance method and system based on multi-element energy storage complementary cooperation, which obtains the optimal capacity allocation value sequence of the energy storage medium on the target time scale set by fusing the dynamic characteristics of multi-element energy storage media, to solve the problem that traditional energy storage scheduling methods do not fully consider the complex relationship between power balance targets and dynamic responses of energy storage paths, resulting in reduced scheduling efficiency and system stability.

[0007] To achieve the above object, the present application provides the following technical solutions:

[0008] The multi-time scale power balance method based on multi-element energy storage complementary cooperation comprises the following steps: acquiring a dynamic response parameter set of multi-element energy storage media, and constructing a response vector through cluster analysis; performing trajectory matching on the response vector by using a DTW algorithm to obtain a DTW path distance matrix, and extracting a phase difference sequence of each energy storage medium through Hilbert transformation to obtain a candidate path set; initializing a redundant capacity allocation vector, and randomly superimposing the redundant capacity allocation vector into the candidate path set to obtain a first evaluation matrix; constructing a multi-objective optimization function, and evaluating the deviation of the first evaluation matrix from a preset power balance curve by using an NSGA-III algorithm; and modifying the redundant capacity allocation vector according to the evaluation result to obtain a best capacity allocation value sequence of each energy storage medium on a target time scale set.

[0009] In a preferred embodiment, the response vector is constructed through cluster analysis, specifically: a principal component analysis method is used to perform feature analysis on the dynamic response parameter set to obtain a plurality of first response vectors; the plurality of first response vectors are grouped according to a preset energy storage medium type, and the Euclidean distance of the first response vectors in each group is calculated to obtain a plurality of response distance clusters; a density clustering algorithm is used to perform cluster analysis on the plurality of response distance clusters to obtain a cluster center vector of each energy storage medium; and the cluster center vectors of each energy storage medium are weighted and fused according to a preset weight to obtain the response vector.

[0010] In a preferred embodiment, the trajectory matching on the response vector by using the DTW algorithm to obtain the DTW path distance matrix, and the extraction of the phase difference sequence of each energy storage medium through Hilbert transformation to obtain the candidate path set, specifically: a historical standard response trajectory is acquired, and the DTW algorithm is used to calculate the matching path of the response vector and the historical standard response trajectory to obtain a plurality of path matching sequences; based on the plurality of path matching sequences, the DTW path distance matrix is constructed; the DTW path distance matrix is subjected to frequency domain decomposition to extract the instantaneous phase feature of each energy storage medium; the instantaneous phase feature is converted into a phase difference sequence through Hilbert transformation; and the plurality of path matching sequences are screened according to the phase difference sequence to generate the candidate path set.

[0011] In a preferred implementation, the initial redundancy capacity allocation vector is randomly generated in the feasible solution space, specifically: a clustering algorithm is used to partition the feasible solution space to obtain a plurality of capacity clustering centers; the plurality of capacity clustering centers are adjusted to perform biased random sampling on the preset probability density function to obtain the initial redundancy capacity allocation vector.

[0012] In a preferred implementation, the initial redundancy capacity allocation vector is randomly generated in the feasible solution space, specifically: a clustering algorithm is used to partition the feasible solution space to obtain a plurality of capacity clustering centers; the plurality of capacity clustering centers are adjusted to perform biased random sampling on the preset probability density function to obtain the initial redundancy capacity allocation vector.

[0013] In a preferred implementation, the first evaluation matrix is constructed by vector superposition operation, specifically: the redundancy capacity allocation vector is decomposed by wavelet to obtain a basic capacity component and a fluctuation compensation component; the basic capacity component is linearly superimposed on the reference capacity of the candidate path set; the fluctuation compensation component is fused with the phase difference sequence of the candidate path by convolution operation to generate an adjusted capacity; the reference capacity and the adjusted capacity are fused according to a preset weight to obtain the first evaluation matrix.

[0014] In a preferred implementation, the deviation of the first evaluation matrix from the preset power balance curve is evaluated by using the NSGA-III algorithm, specifically: the first evaluation matrix is input into the NSGA-III algorithm, a reference point set and a population are initialized, and non-dominated sorting is performed to obtain a first candidate scheme set; the power output sequence corresponding to each candidate path in the first evaluation matrix is extracted and compared with the preset power balance curve to construct a first deviation vector; the first candidate scheme set is screened based on the first deviation vector to obtain a second candidate scheme set.

[0015] In a preferred implementation, the initial redundancy capacity allocation vector is randomly generated in the feasible solution space, specifically: a clustering algorithm is used to partition the feasible solution space to obtain a plurality of capacity clustering centers; the plurality of capacity clustering centers are adjusted to perform biased random sampling on the preset probability density function to obtain the initial redundancy capacity allocation vector.

[0016] extracting a redundant vector component corresponding to a pareto solution from the second candidate solution set; calculating a gradient direction deviation of each redundant vector component from a preset initial component to obtain a correction step; correcting the redundant capacity allocation vector based on the correction step to obtain a corrected vector; analyzing the corrected vector on the target time scale set to obtain a plurality of time scale components; and converting the plurality of time scale components into actual capacity instruction sequences respectively, and injecting the actual capacity instruction sequences into a storage control system interface protocol.

[0017] The multi-time scale power balance method and system based on multi-element energy storage complementary cooperation have the following technical effects and advantages:

[0018] 1. The representative response vector is extracted by clustering and principal component analysis on the dynamic response parameters of the multi-element energy storage medium, which effectively reflects the response capability difference of various energy storage media at different time scales; the phase difference feature is extracted by combining the DTW dynamic time warping algorithm and Hilbert transform, so as to establish a candidate path set and realize accurate modeling of the dynamic characteristics of the energy storage path. On this basis, wavelet decomposition and convolution fusion means are introduced, the redundant capacity allocation vector is superimposed with the dynamic characteristics of the energy storage path, an evaluation matrix of multiple time scales is formed, and fine data support is provided for subsequent multi-objective optimization. The NSGA-III algorithm is used for non-dominated sorting and deviation analysis of the capacity configuration scheme, so as to further improve the rationality of the redundant allocation and the coordination of the system output. Finally, the redundant allocation vector is iteratively corrected in combination with the multi-objective optimization result to form a capacity instruction sequence of different time scales, and fine capacity allocation under dynamic load change is realized. The method significantly improves the response efficiency and operation stability of the multi-element energy storage system in the complex power grid environment, solves the problems of rough scheduling strategy and lack of dynamic coupling analysis in the traditional method, and provides technical support and implementation path for constructing an energy storage optimization scheduling system in a new smart grid. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 A flowchart of the multi-time scale power balance method based on multi-element energy storage complementary cooperation of the present application is shown.

[0020] Figure 2 A decomposition result diagram of the initial redundant capacity allocation vector and the corrected vector in the time scale provided by the embodiment of the present application is shown.

[0021] Figure 3 A structure diagram of the multi-time scale power balance system based on multi-element energy storage complementary cooperation of the present application is shown. DETAILED DESCRIPTION

[0022] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative effort belong to the protection scope of the present application.

[0023] Embodiment 1, Figure 1 The multi-time scale power balance method based on multi-element energy storage complementary cooperation of the present application is given, including the following steps:

[0024] S1, obtaining a dynamic response parameter set of a multi-element energy storage medium, and constructing a response vector through cluster analysis;

[0025] In this example, the dynamic response parameter set of the multi-element energy storage medium is obtained, specifically:

[0026] First, modeling experiments are performed on three types of typical energy storage media connected to the power grid, namely lithium batteries, supercapacitors and compressed air energy storage devices, to obtain their dynamic response data under standard test conditions (ambient temperature 25℃, rated load fluctuation ±10%, cycle 30s). The experiment includes unit power step input test, load disturbance response test and power frequency scanning test.

[0027] During the experiment, the dynamic parameters are extracted by using system identification method, specifically: the step response test is used to calculate the response delay (the time point at which the output power first exceeds 5% of the target value) and the dynamic adjustment rate (the maximum slope of the power rise curve divided by the rated capacity, that is, the change rate in C / s); the load disturbance test is used to calculate the power tracking error, and the root mean square error formula is used, as follows:

[0028]

[0029] Wherein, N is the total number of sampling points; is the actual output power measured at the kth sampling point; is the power reference instruction value at the kth sampling point; is the rated power of the energy storage medium.

[0030] The experimental results are as follows:

[0031] For lithium batteries: the response delay is 0.5s, the dynamic adjustment rate is 0.2C / s, the maximum charge and discharge power is 100kW, and the root mean square value of power tracking error is 3.1%;

[0032] For super capacitor: response delay is 0.1s, dynamic adjustment rate is 1.0C / s, maximum output power is 60kW, tracking error is 1.2%, and inertial response time constant is 0.03s;

[0033] For compressed air energy storage system: response delay is 2.5s, adjustment rate is 0.05C / s, maximum output power is 300kW, tracking error is 5.6%, and heat exchange delay time is 4.1s;

[0034] The above parameters constitute a dynamic response parameter set after normalization, wherein the normalization method adopts the Min-Max method in a fixed interval, and the related formula is as follows:

[0035]

[0036] Wherein, x is the original parameter value to be normalized; is the engineering lower limit value set for the parameter; is the engineering upper limit value set for the parameter; is the dimensionless result after normalization, and the value range is 0 to 1.

[0037] In order to ensure reproducibility, the embodiment takes unified engineering upper and lower limits, for example, the upper limit of response delay is set to 10s, the upper limit of adjustment rate is set to 2C / s, the upper limit of power error is set to 10%, and the upper limit of time constant is set to 100s; the values exceeding the range are truncated at the boundary.

[0038] In the present example, the response vector is constructed by cluster analysis, specifically:

[0039] The principal component analysis method is used to analyze the dynamic response parameter set, and a plurality of first response vectors are obtained;

[0040] The plurality of first response vectors are grouped according to the pre-set energy storage medium type, and the Euclidean distance of the first response vectors in each group is calculated to obtain a plurality of response distance clusters;

[0041] The density clustering algorithm is used to analyze the plurality of response distance clusters, and the cluster center vector of each energy storage medium is obtained;

[0042] The cluster center vector of each energy storage medium is weighted and fused according to the pre-set weight to obtain the response vector.

[0043] The above steps are specifically implemented as follows:

[0044] Firstly, the principal component analysis (PCA) method is used to reduce the dimension and extract the features of the obtained dynamic response parameter set, specifically, the parameter matrix After standardization, the covariance matrix is calculated and its eigenvalues and eigenvectors are obtained, sorted in descending order of eigenvalues, and the first several principal components with cumulative contribution rate ≥95% are selected to form the first response vector set .

[0045] Subsequently, the first response vectors are grouped and classified according to the energy storage medium type (such as lithium battery, super capacitor, compressed air, etc.), and the Euclidean distance between each response vector in each group is calculated, and the formula is:

[0046]

[0047] wherein, , represent two first response vectors, , represent the kth component of , , and m is the dimension of the principal component. The response distance cluster reflecting the internal response difference of the energy storage medium is formed.

[0048] On this basis, the density clustering algorithm (such as DBSCAN) is used for clustering analysis of the response distance cluster, wherein the neighborhood radius parameter ε can be adaptively set according to the average value and standard deviation of the response distance (for example, ε = average value + 0.5 × standard deviation), and the minimum sample number is 3 to ensure the robustness of clustering and identify the representative response feature points in each type of energy storage medium, i.e. the cluster center vector.

[0049] Finally, each cluster center vector is weighted and fused according to the preset weight, and the formula is:

[0050]

[0051] wherein, is the cluster center vector of the i th energy storage medium, is the corresponding weight, satisfying , and q is the number of types of energy storage media. The weight can be set according to the capacity proportion, control priority or empirical parameter. The final response vector is generated as the basis data for subsequent path matching and capacity optimization analysis. Through the above method, the dynamic response difference of different types of energy storage devices can be fully considered, and the overall modeling accuracy and the practicality of dispatching control can be improved.

[0052] S2, using DTW algorithm to match the response vectors, obtaining DTW path distance matrix, and through Hilbert transformation to extract the phase difference sequence of each energy storage medium to obtain the candidate path set;

[0053] In the present example, the DTW algorithm is used to perform trajectory matching on the response vector, a DTW path distance matrix is obtained, and the Hilbert transform is used to extract the phase difference sequence of each energy storage medium to obtain a candidate path set, specifically as follows:

[0054] The historical standard response trajectory is obtained, and the DTW algorithm is used to calculate the matching path of the response vector and the historical standard response trajectory to obtain a plurality of path matching sequences;

[0055] Based on the plurality of path matching sequences, a DTW path distance matrix is constructed;

[0056] The DTW path distance matrix is subjected to frequency domain decomposition to extract the instantaneous phase feature of each energy storage medium;

[0057] The Hilbert transform is used to convert the instantaneous phase feature into a phase difference sequence;

[0058] The plurality of path matching sequences are screened according to the phase difference sequence to generate a candidate path set.

[0059] The above steps are specifically implemented as follows:

[0060] The dynamic time warping (DTW) algorithm is used to calculate the matching path between the response vector and the historical standard response trajectory The distance measurement formula is:

[0061]

[0062] And through the recursive relationship:

[0063]

[0064] The optimal alignment path is obtained, wherein, is the local distance of the i-th and j-th sampling points.

[0065] After obtaining a plurality of path matching sequences, a DTW path distance matrix is constructed, wherein the element represents the average DTW distance between the p-th path and the q-th path, which is used to quantify the matching degree of different paths.

[0066] For each path sequence in the matrix , the fast Fourier transform (FFT) is used for frequency domain decomposition to obtain the amplitude sequence under different frequency components. Then the Hilbert transform is used to calculate the instantaneous phase, and the calculation formula is as follows:

[0067]

[0068] wherein,​ denotes the Hilbert transform operator, is the time-domain signal amplitude.

[0069] According to the difference of the instantaneous phase of different energy storage media, the phase difference sequence is defined as:

[0070]

[0071] wherein, and are the instantaneous phases of energy storage media i and j, respectively.

[0072] Finally, according to the stability index of the phase difference sequence, the path is screened, and the stability index is defined as the standard deviation of the phase difference sequence When rad, it is determined that the trend is stable; meanwhile, the matching degree of the path and the standard trajectory is calculated When , it is determined that the degree of cooperation is high. The path that meets the above conditions constitutes a candidate path set. Wherein, The unit of is radian (rad), which represents the standard deviation of the phase difference sequence; represents the cumulative distance of dynamic time warping of a certain candidate path and the standard response trajectory.

[0073] S3, initialize the redundant capacity allocation vector and randomly superimpose it in the candidate path set to obtain a first evaluation matrix;

[0074] In this example, the redundant capacity allocation vector is initialized and randomly superimposed in the candidate path set to obtain a first evaluation matrix, specifically:

[0075] According to the preset power grid load fluctuation range and the characteristics of the energy storage medium, a feasible solution space of the redundant capacity allocation vector is constructed;

[0076] An initial redundant capacity allocation vector is randomly generated in the feasible solution space;

[0077] The initial redundant capacity allocation vector is mapped to each candidate path in the candidate path set according to the type of energy storage medium, to obtain a redundant capacity allocation vector for each candidate path;

[0078] Based on the redundant capacity allocation vector of each candidate path, a first evaluation matrix is constructed through vector superposition operation.

[0079] The above steps are specifically implemented as follows:

[0080] First, according to the load fluctuation range of the power grid system and the response characteristics of various types of energy storage media, a feasible solution space of redundant capacity allocation is established. The feasible solution space is defined by constraint conditions, including: the SOC range of the energy storage medium power range and charging and discharging efficiency to ensure that the randomly generated vector meets the energy conservation and device safety boundary. The space covers all possible, technically feasible and safe operation boundary compliant redundant capacity configuration modes.

[0081] On this basis, the method randomly generates an initial redundant capacity allocation vector in the solution space, which represents the redundant capacity shared by different energy storage media under the current load fluctuation background.

[0082] In this example, an initial redundant capacity allocation vector is randomly generated in the feasible solution space, specifically:

[0083] A clustering algorithm (such as K-means or DBSCAN) is used to partition the feasible solution space, obtaining a number of capacity clustering centers;

[0084] Adjust the number of capacity clustering centers to perform biased random sampling on the preset probability density function, obtaining an initial redundant capacity allocation vector, such as a Gaussian distribution:

[0085]

[0086] wherein, is the clustering center, is the standard deviation, x represents the capacity allocation vector, which is a value sampled from the probability distribution, representing the redundant capacity allocation of each energy storage medium; or a Beta distribution:

[0087]

[0088] wherein, , is the shape parameter, is the Beta function. Through this mechanism, the method effectively avoids the problem of uneven distribution or falling into an infeasible region that may be caused by pure random generation, providing a stable and accurate initial input basis for subsequent path superposition and optimization.

[0089] In this example, a first evaluation matrix is constructed through vector superposition operation, specifically:

[0090] The redundant capacity allocation vector is decomposed by wavelet to obtain a basic capacity component and a fluctuation compensation component;

[0091] The basic capacity component is linearly superimposed on the reference capacity of the candidate path set;

[0092] The fluctuation compensation component is fused with the phase difference sequence of the candidate path through convolution operation to generate an adjusted capacity;

[0093] The reference capacity and the adjusted capacity are fused according to preset weights to obtain a first evaluation matrix.

[0094] It should be noted that:

[0095] Wavelet decomposition adopts Daubechies (db4) wavelet basis for 3-layer decomposition, and the low-frequency part is used as the basic capacity component, and the high-frequency part is used as the fluctuation compensation component. The convolution operation adopts the formula:

[0096]

[0097] wherein, is the adjusted capacity, is the fluctuation compensation component, is the phase difference sequence, is a time delay parameter, indicating the time offset of the fluctuation compensation component. The weighted fusion adopts the formula:

[0098]

[0099] wherein, is the capacity sequence in the first evaluation matrix, is the reference capacity, is the fusion weight, and the value range is between 0 and 1, which can be set according to the system scheduling strategy.

[0100] The evaluation matrix not only contains the time sequence output characteristics of various energy storage media, but also comprehensively considers factors such as power fluctuation, phase response and path adaptation, providing a structured and quantifiable input basis for subsequent capacity configuration and scheduling path selection using multi-objective optimization algorithms (such as NSGA-III), thereby significantly improving the balancing performance and operation efficiency of the energy storage system under multiple time scales.

[0101] S4, constructing a multi-objective optimization function, and using the NSGA-III algorithm to evaluate the deviation of the first evaluation matrix and the preset power balance curve;

[0102] In this example, the NSGA-III algorithm is used to evaluate the deviation of the first evaluation matrix and the preset power balance curve, specifically:

[0103] The first evaluation matrix is input into the NSGA-III algorithm, the reference point set and the population are initialized, and the non-dominated sorting is performed to obtain the first candidate solution set;

[0104] The power output sequence corresponding to each candidate path in the first evaluation matrix is extracted and compared with the preset power balance curve to construct the first deviation vector;

[0105] Based on the first deviation vector, the first candidate solution set is screened to obtain the second candidate solution set.

[0106] It should be noted that the first evaluation matrix is first imported into the NSGA-III optimization framework as input data, and the reference point set and population individuals are initialized, wherein the population size is set to 100, the maximum number of iterations is set to 200 generations, the reference point set is generated by uniformly distributing in the target dimension, and then all evaluation solutions are classified by the non-dominated sorting method to form a preliminary candidate scheme set.

[0107] In terms of optimization objectives, the multi-objective function is defined as:

[0108]

[0109] wherein,

[0110]

[0111] represents the average absolute error between the path power output and the power balance curve, is the actual output power of the candidate path at time t; is the preset power reference instruction value of the power balance curve at time t, and T is the total duration of the optimization period;

[0112]

[0113] represents the capacity utilization rate, wherein is the actual energy used in the optimization period, is the total available energy;

[0114]

[0115] represents the standard deviation of the path output power sequence, which is used to measure the stability of the path.

[0116] Next, the corresponding power output time sequence is extracted from each candidate path, and it is compared with the pre-set power balance curve point by point, the difference between the two is calculated, and the "first deviation vector" reflecting the degree of power deviation is constructed. The deviation vector is defined as:

[0117]

[0118] The deviation vector is used to measure the performance of each candidate path in meeting the power balance target, and is used as an evaluation index of the optimization algorithm. Based on the index, the algorithm further selects the path combination with better performance, and selects the second candidate solution set through the Pareto frontier, that is, the optimal solution set that achieves a good trade-off between multiple targets (such as capacity utilization rate, path stability, power deviation minimization, etc.). Through the optimization evaluation process, the method can effectively identify the path combination with the best response ability and power allocation performance under different time scales, thereby improving the scheduling accuracy and global coordination of the energy storage system.

[0119] S5, according to the evaluation result, the redundant capacity allocation vector is modified to obtain the best capacity allocation value sequence of each energy storage medium on the target time scale set.

[0120] In this example, according to the evaluation result, the redundant capacity allocation vector is modified to obtain the best capacity allocation value sequence of each energy storage medium on the target time scale set, specifically:

[0121] The redundant vector component corresponding to the Pareto solution is extracted from the second candidate solution set;

[0122] The gradient direction deviation of each redundant vector component and the preset initial component is calculated to obtain a correction step;

[0123] The redundant capacity allocation vector is modified based on the correction step to obtain a modified vector;

[0124] The modified vector is analyzed on the target time scale set to obtain a plurality of time scale components;

[0125] The plurality of time scale components are respectively converted into actual capacity instruction sequences and injected into the energy storage control system interface protocol.

[0126] Exemplarily, the second candidate solution set is:

[0127]

[0128] wherein, is the kth candidate solution, corresponding to the redundant capacity allocation vector of different energy storage media; N is the number of candidate solutions; m is the total number of energy storage media; represents the redundant capacity allocation value of the jth energy storage medium in the kth solution.

[0129] The gradient direction deviation of the current component and the original component in each dimension is defined as:

[0130]

[0131] wherein, represents the capacity gradient direction deviation of the jth energy storage medium. is the jth energy storage medium corresponding to the redundant capacity allocation value in the Pareto optimal solution; represents the initial redundant capacity allocation value of the jth energy storage medium; j is an index variable, representing the energy storage medium number.

[0132] Based on the gradient direction deviation, the correction step is defined as:

[0133]

[0134] wherein, is the correction step of the jth energy storage medium, is a learning rate parameter (value range 0~1, used to control the correction amplitude).

[0135] The corrected capacity vector is:

[0136]

[0137] wherein, is the corrected redundant capacity allocation value of the jth energy storage medium.

[0138] It should be noted that first, the redundant capacity vector corresponding to the Pareto optimal solution is extracted from the second candidate solution set, and compared with the initially set initial component, the gradient direction deviation in the multi-dimensional parameter space is calculated, and then the correction step of each energy storage medium in the capacity adjustment process is determined. Subsequently, the original redundant capacity allocation vector is adjusted using the correction step to generate a new correction vector. The vector is then mapped to the target time scale set, and a number of capacity components with different frequencies and durations are extracted through time scale decomposition, as shown in Figure 2 , which shows the decomposition results of the initial redundant capacity allocation vector and the corrected vector on the time scale, wherein the low-frequency component reflects the long-term capacity configuration, and the high-frequency component corresponds to the short-term fluctuation compensation. Each time scale component is further converted into a capacity control instruction sequence with actual engineering significance, and finally injected into the dispatching system through the control interface of the energy storage system, realizing the closed-loop control from theoretical optimization to actual execution. This process not only improves the accuracy and timeliness of capacity allocation, but also ensures the dynamic response ability of the system to power fluctuations under multiple time scales, effectively solving the problem of rigid capacity configuration and difficulty in adapting to dynamic changes of the power system in traditional methods.

[0139] Embodiment 2, Figure 3 The multi-time scale power balance system based on multi-element energy storage complementary cooperation of the application is given, which includes a vector construction module, a phase matching module, a redundancy evaluation module, an optimization decision module, and a capacity allocation module:

[0140] A vector construction module is configured to acquire a dynamic response parameter set of the multi-element energy storage medium and construct a response vector through cluster analysis;

[0141] A phase matching module is configured to perform trajectory matching on the response vector by using a DTW algorithm, obtain a DTW path distance matrix, and extract a phase difference sequence of each energy storage medium through Hilbert transform to obtain a candidate path set;

[0142] A redundancy evaluation module is configured to initialize a redundancy capacity allocation vector and randomly superimpose the redundancy capacity allocation vector into the candidate path set to obtain a first evaluation matrix;

[0143] An optimization decision module is configured to construct a multi-objective optimization function and evaluate the deviation of the first evaluation matrix from a preset power balance curve by using an NSGA-III algorithm.

[0144] A capacity allocation module is configured to correct the redundancy capacity allocation vector according to the evaluation result to obtain a best capacity allocation value sequence of each energy storage medium in the target time scale set.

[0145] The above formulas are all dimensionless numerical calculations, the formulas are obtained by software simulation of a large amount of data to obtain a formula of the most recent real situation, and the preset parameters in the formulas are set by a person skilled in the art according to the actual situation.

[0146] The above embodiments can be realized wholly or partially by software, hardware, firmware or any other combination. When realized by software, the above embodiments can be realized wholly or partially in the form of a computer program product.

[0147] Those skilled in the art can realize that the modules and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized by hardware or software depends on the specific application and design constraints of the technical solutions. A person skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0148] In addition, the functional modules in each embodiment of the present application can be integrated in one processing module, or each module can exist physically alone, or two or more modules can be integrated in one module.

[0149] The above is merely specific embodiments of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0150] Finally: the above only for the preferred embodiments of the present application, and not for limiting the present application, any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application, should be included in the scope of protection of the present application.

Claims

1. A multi-timescale power balance method based on complementary and synergistic multi-element energy storage, characterized in that, Includes the following steps: The dynamic response parameter set of multi-element energy storage media is obtained, and the response vector is constructed through cluster analysis; The DTW algorithm is used to perform trajectory matching on the response vector to obtain the DTW path distance matrix, and the phase difference sequence of each energy storage medium is extracted by Hilbert transform to obtain the candidate path set; Initialize the redundancy capacity allocation vector and randomly superimpose it onto the candidate path set to obtain the first evaluation matrix. Specifically, map the initial redundancy capacity allocation vector to each candidate path in the candidate path set according to the energy storage medium type to obtain the redundancy capacity allocation vector of each candidate path; construct the first evaluation matrix based on the redundancy capacity allocation vector of each candidate path through vector superposition operation. A multi-objective optimization function is constructed, and the NSGA-III algorithm is used to evaluate the deviation between the first evaluation matrix and the preset power balance curve. Based on the evaluation results, the redundant capacity allocation vector is corrected to obtain the optimal capacity allocation value sequence for each energy storage medium on the target time scale set. The construction of the first evaluation matrix through vector superposition operation is specifically as follows: Wavelet decomposition is performed on the redundant capacity allocation vector to obtain the basic capacity component and the fluctuation compensation component. The baseline capacity components are linearly superimposed onto the baseline capacity of the candidate path set; The fluctuation compensation component is fused with the phase difference sequence of the candidate path through convolution operation to generate the adjustment capacity; The baseline capacity and the adjusted capacity are merged according to preset weights to obtain the first evaluation matrix.

2. The multi-timescale power balance method based on complementary and synergistic multi-element energy storage as described in claim 1, characterized in that, The construction of the response vector through cluster analysis is specifically as follows: Principal component analysis was used to perform feature analysis on the dynamic response parameter set, resulting in several first response vectors. Several first response vectors are grouped according to a preset energy storage medium type, and the Euclidean distance of the first response vectors in each group is calculated to obtain several response distance clusters. Density clustering algorithm is used to perform cluster analysis on several response distance clusters to obtain the cluster center vector of each energy storage medium; The cluster center vectors of each energy storage medium are weighted and fused according to preset weights to obtain the response vector.

3. The multi-timescale power balance method based on complementary and synergistic multi-element energy storage as described in claim 2, characterized in that, The process involves using the DTW algorithm to perform trajectory matching on the response vector, obtaining the DTW path distance matrix, and then extracting the phase difference sequence of each energy storage medium using Hilbert transform to obtain a candidate path set. Specifically: Historical standard response trajectories are obtained, and the DTW algorithm is used to calculate the matching path between the response vector and the historical standard response trajectory, resulting in several path matching sequences. Construct the DTW path distance matrix based on several path matching sequences; Frequency domain decomposition is performed on the DTW path distance matrix to extract the instantaneous phase characteristics of each energy storage medium; The instantaneous phase features are converted into a phase difference sequence using the Hilbert transform. A candidate path set is generated by filtering several path matching sequences based on the phase difference sequence.

4. The multi-timescale power balance method based on complementary and synergistic multi-element energy storage as described in claim 3, characterized in that, The initialization of the redundant capacity allocation vector is specifically as follows: Based on the preset power grid load fluctuation range and energy storage medium characteristics, a feasible solution space for the redundant capacity allocation vector is constructed. An initial redundancy capacity allocation vector is randomly generated within the feasible solution space.

5. The multi-timescale power balance method based on complementary and synergistic multi-element energy storage as described in claim 4, characterized in that, The step of randomly generating an initial redundant capacity allocation vector within the feasible solution space specifically involves: A clustering algorithm is used to spatially partition the feasible solution space, resulting in several capacity cluster centers; By adjusting the probability density function corresponding to several capacity cluster centers and performing biased random sampling, an initial redundant capacity allocation vector is obtained.

6. The multi-timescale power balance method based on complementary and synergistic multi-element energy storage as described in claim 5, characterized in that, The NSGA-III algorithm is used to evaluate the deviation between the first evaluation matrix and the preset power balance curve, specifically as follows: The first evaluation matrix is ​​input into the NSGA-III algorithm to initialize the reference point set and population, and a non-dominated sort is performed to obtain the first candidate scheme set; Extract the power output sequence corresponding to each candidate path in the first evaluation matrix and compare it with the preset power balance curve to construct the first deviation vector; The first candidate scheme set is filtered based on the first deviation vector to obtain the second candidate scheme set as the evaluation result.

7. The multi-timescale power balance method based on complementary and synergistic multi-element energy storage as described in claim 6, characterized in that, The process of correcting the redundant capacity allocation vector based on the evaluation results to obtain the optimal capacity allocation value sequence for each energy storage medium on the target time scale set is as follows: Extract the redundant vector components corresponding to the Pareto solution from the second candidate solution set; Calculate the gradient direction deviation between each redundant vector component and the preset initial component to obtain the correction step size; The redundant capacity allocation vector is corrected based on the correction step size to obtain the correction vector; The correction vector is analyzed on the target time scale set to obtain several time scale components; Several time-scale components are converted into actual capacity command sequences and injected into the energy storage control system interface protocol.

8. A multi-timescale power balance system based on complementary and synergistic multi-element energy storage, applied to the multi-timescale power balance method based on complementary and synergistic multi-element energy storage as described in any one of claims 1-7, characterized in that, It includes a vector construction module, a phase matching module, a redundancy assessment module, an optimization decision module, and a capacity allocation module: The vector construction module is used to obtain the dynamic response parameter set of multi-element energy storage media and construct response vectors through cluster analysis; The phase matching module is used to perform trajectory matching on the response vector using the DTW algorithm to obtain the DTW path distance matrix, and to extract the phase difference sequence of each energy storage medium through Hilbert transform to obtain the candidate path set; The redundancy assessment module is used to initialize the redundancy capacity allocation vector and randomly superimpose it onto the candidate path set to obtain the first assessment matrix; The optimization decision module is used to construct a multi-objective optimization function and uses the NSGA-III algorithm to evaluate the deviation between the first evaluation matrix and the preset power balance curve. The capacity allocation module is used to correct the redundant capacity allocation vector based on the evaluation results, so as to obtain the optimal capacity allocation value sequence for each energy storage medium on the target time scale set.

Images