Power grid energy storage method considering wind and light uncertainty and time sequence characteristics
Through technologies such as SOM clustering and generation of adversarial networks, typical seasonal scenarios of wind and light loads are generated, and a multi-time-scale energy storage model is established, which solves the long-term imbalance between wind and light output and load demand, and realizes the economic and low-carbon nature of the power grid. The goal is to ensure the stable operation of the power grid.
Patent Information
- Application Number
- CN202411166787.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-23
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2044-08-23
Smart Images

Figure SMS_3 
Figure SMS_7 
Figure SMS_10
Abstract
Description
Technical Field
[0001] The present invention relates to the field of power grid energy storage, and more specifically, to a power grid energy storage method that takes into account the uncertainty and timing characteristics of wind and solar power. Background Art
[0002] At present, the proportion of new energy installed capacity, mainly wind and solar, is increasing year by year, and the output of wind and solar is greatly affected by the external environment, showing obvious seasonal characteristics, which is mainly reflected in the abundant light resources in summer, large photovoltaic output and small wind turbine output, and abundant wind energy resources in winter, large wind turbine output and small photovoltaic output. Load demand also has obvious seasonal characteristics. In summer, the demand for cooling loads such as air conditioners increases, and in winter, the demand for heating loads such as electric heaters increases. There is a long-term power imbalance problem between the wind and solar supply on the source side and the user demand on the load side. At present, the general energy storage uses energy storage with short-term energy regulation characteristics, such as electrochemical energy storage, which cannot solve the problem of long-term power imbalance between source and load. Considering long-term energy storage, long-term scale transfer of electricity can be realized, which can improve the utilization rate of wind and solar while meeting load demand and reducing the power supply of thermal power units. How to use long-term energy storage to solve the problem of long-term power imbalance between source and load and ensure the stable operation of the power grid has become one of the research hotspots in recent years.
[0003] However, since long-term energy storage is based on weeks, months, or even seasons, its installation and storage costs are much higher than those of short-term energy storage. How to achieve economic operation of the system after considering long-term energy storage is a difficult point. At the same time, in a multi-time scale hybrid energy storage system composed of long-term and short-term energy storage, short-term energy storage achieves power balance during the day. Although long-term energy storage can also be charged and discharged normally during the day, how to achieve power balance during the day is still a major technical problem. It is necessary to establish a long-short-term multi-time scale hybrid energy storage model suitable for realistic conditions.
[0004] To this end, we proposed a grid energy storage method that takes into account the uncertainty and timing characteristics of wind and solar power. Summary of the invention
[0005] The purpose of the present invention is to provide a power grid energy storage method that takes into account the uncertainty and timing characteristics of wind and solar power, in order to solve the above problems in view of the deficiencies in the prior art.
[0006] The above technical objectives of the present invention are achieved through the following technical solutions: a power grid energy storage method considering the uncertainty and timing characteristics of wind and solar, including reducing the scene by SOM clustering method to obtain a typical seasonal scene of wind and solar load, establishing a multi-time scale energy storage model considering long-term energy storage, and building a power grid wind, solar and storage stochastic programming model considering the economy and low carbon of power grid investment;
[0007] The specific steps include:
[0008] S1. A method for generating seasonal scenarios of wind and solar load considering uncertainty;
[0009] This includes proposing an improved generative adversarial network to generate wind and solar power output scenarios that take uncertainty into account. On this basis, the wind and solar load data are subjected to scenario reduction using a combination of Latin hypercube sampling and SOM clustering algorithms to obtain typical wind and solar load scenarios with seasonal characteristics.
[0010] S2. Grid planning method considering multi-time scale energy storage coordination;
[0011] Based on the typical seasonal scenario of wind and solar load considering uncertainty, the power grid is planned, and a multi-time scale hybrid energy storage model with the coordination of short-term energy storage and long-term energy storage is proposed. On this basis, a power grid planning method considering the coordination of multi-time scale energy storage is proposed.
[0012] Furthermore, in step S1, the improved generative adversarial network includes a traditional generative adversarial network including a generator and a discriminator: the generator receives input noise generated data, the discriminator determines whether the data generated by the generator is true based on historical data, and after the discriminator determines that the synthetic data is obtained, the generator and the discriminator will be updated by fixing each other until the discriminator can no longer determine whether the data given by the generator is synthetic data, and the generator and the discriminator stop updating;
[0013] Assume that the historical wind and solar power output data is x, and its distribution is P data (x), random noise z adopts Gaussian distribution P z To describe, assume that the generator is G and the discriminator is D. During the training process, the noise z~P z Sent to the generator, outputting the generated wind and solar power output G(z)~P g At the same time, the historical wind and solar power output is sent to the discriminator, which judges the generated data.
[0014] The loss function of the generator is:
[0015] L G =-E z~p(z) [logD(G(z))] (1)
[0016] In the formula, E is the mathematical expectation, D(G(z)) is the result of the discriminator on the generated data,
[0017] The loss function of the discriminator is:
[0018]
[0019] In the formula, D(x) is the result of the discriminator on the real data, 1-D(G(z)) is the result of the discriminator on the synthetic data,
[0020] The goal of the generator is to generate synthetic data that is more like real data, and the goal of the discriminator is to identify as much synthetic data as possible. Therefore, the objective function of the generative adversarial network is:
[0021]
[0022] The objective function solution process is equivalent to solving P data (x) and P g JS divergence problem:
[0023]
[0024] But P data (x) and P g It is difficult to have an ideal overlap, so the JS divergence easily becomes a constant, causing the discriminator to be unable to recognize the distance between the two distributions, resulting in gradient disappearance and poor training results. To solve this problem, the traditional generative adversarial network is improved: the Wasserstein distance is introduced to optimize the neural network. The Wasserstein distance is:
[0025]
[0026] Where P r ,P w are two different distributions, Π(P r ,P w ) is the joint distribution of two distributions (P r ,P w ), γ is a possible distribution in the set, x is the real sample, y is the generated sample, (x, y) is generated from γ, ||xy|| is the distance between the two samples, E (x,y)~γ [||xy||] is the energy consumption,
[0027] The improved generative adversarial network objective function is as follows:
[0028]
[0029] In the formula, D∈1-Lipsschitz makes the discriminator conform to the 1-Lipsschitz function, limits the range of the discriminator, and makes P data (x) and P g Not much difference,
[0030] The loss function of the generator is:
[0031]
[0032] The loss function of the discriminator is:
[0033]
[0034] Furthermore, in step S1, the scene reduction is performed on the wind-solar-load data by combining the Latin hypercube sampling method and the SOM clustering algorithm to obtain the typical wind-solar-load scene with seasonal characteristics. The specific steps are as follows:
[0035] Scenery and seasonal scene reduction
[0036] The uncertain wind and solar data generated by the improved generative adversarial network and the historical load data at the same time throughout the year are combined into 8760 data pairs, which are sorted from small to large according to the time series. The Latin Hypercube Sampling (LHS) method is used to sample them to construct an initial data set with time series characteristics.
[0037] The specific steps of LHS are as follows:
[0038] Assume that the LHS sampling object size is N, there are k random variables, and the random variable x k The probability density function of is as follows:
[0039] y k =f k (x k ) (9)
[0040]
[0041] 4) For each random variable x k The probability space of is divided into N equal parts to form N subintervals, namely:
[0042] {[0,1 / N],[1 / N,2 / N],…,[(N-1) / N,1]},
[0043] 5) Randomly select a sampling point y in each subinterval ki (i=1,2,…,N)
[0044] 6) Use the inverse transformation of the function shown in formula (2) to find the sampling points x in each interval ki , and arrange them in time series from small to large, and get the initial sampling matrix S with sampling value of k×N order, that is, the initial data set of wind and solar load throughout the year,
[0045] The obtained initial data set of wind-solar load for the whole year is normalized, and the self-organizing maps (SOM) clustering algorithm is used to reduce the scenes, and four groups of clustering results are obtained, namely, wind-solar load data sets with seasonal characteristics. The four groups of data sets correspond to the four seasons of spring, summer, autumn and winter. In each data set, the wind-solar load data pairs have corresponding seasonal characteristics.
[0046] The SOM clustering algorithm automatically classifies data according to its characteristics. The basic principle is to classify two data with small Euclidean distance into one category, and use this method to group data. The specific steps of the SOM clustering algorithm are as follows:
[0047] First, suppose there is a data set D, the samples in set D are x1,x2,…,x n , that is: D = {x1, x2, ..., x n}, each sample contains n elements, such as sample x i The element is x i1 ,x i2 ,…,x in , that is: x i =(x i1 ,x i2 ,…,x in ), similarly, sample x j The element is x j1 ,x j2 ,…,x jn , that is, x j =(x j1 ,x j2 ,…,x jn ),
[0048] Calculate sample x i and sample x j The Euclidean distance between i ,x j ) is expressed as follows:
[0049]
[0050] 1) Initialization: Initialize the SOM network connection weights to random values, determine the initial learning value, the initial value of the input vector, and the number of output neurons, i.e., the number of clusters.
[0051] 2) Confirm the best matching unit (BMU), calculate and compare the distances between domain nodes using the Euclidean distance formula, find the minimum distance, and confirm the winning neuron.
[0052] 3) Weight update adjustment: update the node neuron weight within the BMU radius, and use the Gaussian function to shrink the range of each iteration. The neighborhood radius is expressed as:
[0053] r(n)=r0exp(-n / a) (12)
[0054] Where r(n) is the radius of the domain updated for the nth time; r0 is the initial value; a is a constant,
[0055] 4) Repeat steps 2)-3) until the maximum number of iterations, and output the results.
[0056] 5) Finally, the number of clusters is obtained by calculating the Davidson-Bouldin index (DBI), and the cluster with the smallest DBI is selected as the final result of clustering. The calculation method of DBI is as follows:
[0057]
[0058] Where: N is the total number of clusters; L i,j Describes the relationship between the i-th category and the j-th category in the clustering results. The smaller its value, the better the clustering situation. i,j is the distance between clusters i and j; S i is the Euclidean distance of cluster i, T i is the size of the i-th clustering matrix; Y k is the kth eigenvector in the cluster; C i is the centroid of cluster i; C j is the centroid of cluster j; p is the dimension of cluster vector.
[0059] Furthermore, the specific model establishment method of the multi-time scale hybrid energy storage model in step S2 is as follows:
[0060] Short-term energy storage charges and discharges in a short time scale to maintain power balance in a short period of time. The time scale is hours. Its model is as follows:
[0061] SOC constraints:
[0062] SOC0=0.5c (16)
[0063] SOC1=SOC 24 (17)
[0064] SOC min ≤SOC≤SOC max (18)
[0065] In the formula, SOC1 is the SOC state of the short-term energy storage at the initial moment, c is the capacity of the short-term energy storage, and SOC 24Indicates the SOC state of short-term energy storage at the end of the day, SOC min , SOC max are the maximum and minimum values allowed for short-term energy storage SOC,
[0066] Formula 16 indicates that the initial moment of short-term energy storage is half of the capacity, Formula 17 indicates that the SOC of the initial moment and the final moment of a day of short-term energy storage are the same, and Formula 18 indicates that the short-term energy storage operates within the range allowed by SOC.
[0067] Power Constraints:
[0068] P ESS,min ≤P ESS ≤P ESS,max (19)
[0069] Where P ESS,min , P ESS,max are the minimum and maximum values of short-term energy storage charging and discharging, respectively, P ESS is the charging and discharging power of short-term energy storage,
[0070] Long-term energy storage charges and discharges over a long time scale to maintain long-term power balance. The time scale is seasonal. Its model is as follows:
[0071] SOC constraints:
[0072]
[0073] In the formula, is the SOC state of long-term energy storage at the initial time of the first typical day, c ESS is the capacity of long-term energy storage, is the SOC state of long-term energy storage at the end of the kth typical day, is the SOC state of the long-term energy storage at the initial time of the k+1th typical day, is the SOC state of long-term energy storage at the end of the fourth typical day, is the SOC state of long-term energy storage at time t on the kth typical day, is the SOC state of the long-term energy storage at time t-1 on the kth typical day, is the self-discharge rate of long-term energy storage, α ESS For the efficiency of long-term energy storage, and are the charge and discharge of long-term energy storage at time t on the kth typical day, Δt is the scheduling duration, and SOC ESS,min , SOC ESS,max are the maximum and minimum values allowed for long-term energy storage SOC,
[0074] Formula 20 indicates that the long-term energy storage is half of its capacity at the initial moment of the first typical day. Formula 21 indicates that the SOC of the long-term energy storage at the end of the kth typical day is the same as that at the initial moment of the k+1th typical day. Formula 22 indicates that the SOC of the long-term energy storage at the beginning of the first typical day is the same as that at the end of the fourth typical day. Formula 23 is the SOC calculation method of the long-term energy storage at time t on the kth typical day. Formula 24 indicates that the long-term energy storage operates within the allowed range of SOC.
[0075] Power Constraints:
[0076]
[0077] In the formula, P is the charge and discharge state of long-term energy storage, which takes 1 or 0. When it takes 1, it means that the long-term energy storage is being charged and discharged, and when it takes 0, it means that the long-term energy storage is not being charged and discharged; ESS,max is the maximum value of long-term energy storage charge and discharge,
[0078] A grid planning method considering multi-time scale energy storage coordination
[0079] Based on the multi-time scale hybrid energy storage model obtained above, problem B in the proposed power grid planning model is described as follows:
[0080] Objective Function
[0081] The objective function of the optimization problem is to minimize the total cost C of the power grid:
[0082]
[0083] Where C1 is the economic cost of the power grid, C2 is the green cost of the power grid; In,RE 、p In,Sto are the installation costs per unit capacity of wind and solar power and energy storage (including short-term and long-term energy storage); c RE 、c Sto are the installed capacity of wind and solar power, and energy storage (including short-term energy storage and long-term energy storage); p OM,RE 、p OM,Sto are the unit capacity operation and maintenance costs of wind and solar power and energy storage (including short-term energy storage and long-term energy storage); P RE , P Sto They are wind, solar and energy storage output (including short-term and long-term energy storage); p Buy,MG 、p Sel,MG are the electricity purchase and sales prices of the power grid, P Buy,MG , P Sel,MG are the electricity purchased and sold from the power grid respectively; μ is the penalty coefficient for wind and solar power abandonment; P Waste The amount of wind and solar power abandoned:
[0084] P Waste=P RE -P Load -P Sto -P Sel,MG (29)
[0085] Where P Load is the load power,
[0086] Constraints
[0087] Capacity constraints:
[0088]
[0089] Where: c RE,max is the upper limit of wind and solar installation capacity, c Sto The upper limit of energy storage installation capacity (including short-term energy storage and long-term energy storage) is
[0090] Power balance constraints:
[0091] P RE +P Sto -P Sel,MG +P Buy,MG =P Load (31)
[0092] Wind and solar power output constraints:
[0093] P RE,min ≤P RE ≤P RE,max (32)
[0094] Where: P RE,min , P RE,max are the minimum and maximum values of wind and solar power output,
[0095] Constraints on large-scale electricity purchase and sales:
[0096]
[0097] Where: P Buy,MG,max , P Buy,MG,min are the upper and lower limits of the amount of electricity purchased from the large power grid, P Sel,MG,max , P Sel,MG,min are the upper and lower limits of electricity sold to the large power grid,
[0098] In addition to the above constraints, the constraints of the power grid planning model also include equations (9) to (20) mentioned above.
[0099] According to the above scheme,
[0100] Compared with the prior art, the beneficial effects are as follows:
[0101] The present invention achieves daytime power balance by constructing a long-short-term multi-time scale hybrid energy storage model suitable for realistic conditions. It also takes into account the uncertainty of wind and solar power output and the high efficiency of calculation, uses the SOM clustering method to generate typical scenarios, and obtains a typical seasonal data set of wind and solar load, thereby improving calculation efficiency while ensuring calculation accuracy. DETAILED DESCRIPTION
[0102] The invention is further described in detail below through specific examples. The following examples are only illustrative and not limiting of the protection scope of the present invention.
[0103] The present invention includes: a method for generating seasonal scenarios of wind, solar and load considering uncertainty and a method for planning a power grid considering the coordination of multi-time scale energy storage. The SOM clustering method is used to reduce the scenarios to obtain typical seasonal scenarios of wind, solar and load, a multi-time scale energy storage model is established considering long-term energy storage, and a random planning model of wind, solar and storage for power grid is constructed by comprehensively considering the economy and low carbon of power grid investment.
[0104] 1. Method for generating seasonal scenarios of wind, solar and load considering uncertainty
[0105] Affected by environmental factors, wind and solar power output is characterized by randomness and volatility. In addition, wind and solar power output and load demand have obvious seasonal characteristics: in summer, wind farms have less output and photovoltaic power stations have more output, in winter, wind farms have more output and photovoltaic power stations have less output, and load demand increases in summer and winter. In power grid planning, the selection of typical seasonal scenarios of wind and solar load that take into account uncertainty is very important. Generating typical seasonal scenarios of wind and solar load that contain uncertainty includes two problems. Problem A: An improved generative adversarial network is proposed to generate wind and solar power output scenarios that take into account uncertainty. On this basis, Problem B is proposed: The Latin hypercube sampling method and the SOM clustering algorithm are combined to reduce the scenarios of wind and solar load data to obtain typical scenarios of wind and solar load with seasonal characteristics.
[0106] A. Improved Generative Adversarial Network
[0107] The traditional generative adversarial network consists of a generator and a discriminator: the generator receives input noise-generated data, and the discriminator determines whether the data generated by the generator is real based on historical data. After the discriminator determines that the data is synthetic, the generator and the discriminator will be updated by fixing each other until the discriminator can no longer determine whether the data given by the generator is synthetic data, and then the generator and the discriminator stop updating.
[0108] Assume that the historical wind and solar power output data is x, and its distribution is P data (x), random noise z adopts Gaussian distribution P z Assume that the generator is G and the discriminator is D. During the training process, the noise z~P zSent to the generator, outputting the generated wind and solar power output G(z)~P g ,At the same time, the historical wind and solar output is sent to the discriminator, which makes a judgment on the generated data.
[0109] The loss function of the generator is:
[0110] L G =-E z~p(z) [logD(G(z))] (1)
[0111] Where E is the mathematical expectation, and D(G(z)) is the judgment result of the discriminator on the generated data.
[0112] The loss function of the discriminator is:
[0113]
[0114] Where D(x) is the result of the discriminator on real data, and 1-D(G(z)) is the result of the discriminator on synthetic data.
[0115] The goal of the generator is to generate synthetic data that is more like real data, and the goal of the discriminator is to identify as much synthetic data as possible. Therefore, the objective function of the generative adversarial network is:
[0116]
[0117] The objective function solution process is equivalent to solving P data (x) and P g JS divergence problem:
[0118]
[0119] But P data (x) and P g It is difficult to have an ideal overlap, so the JS divergence easily becomes a constant, causing the discriminator to be unable to recognize the distance between the two distributions, resulting in gradient disappearance and poor training results. To solve this problem, the traditional generative adversarial network is improved: the Wasserstein distance is introduced to optimize the neural network. The Wasserstein distance is:
[0120]
[0121] Where P r ,P w are two different distributions, Π(P r ,P w ) is the joint distribution of two distributions (P r ,P w), γ is a possible distribution in the set, x is the real sample, y is the generated sample, (x, y) is generated from γ, and ||xy|| is the distance between the two samples. (x,y)~γ [||xy||] is the energy consumption.
[0122] The improved generative adversarial network objective function is as follows:
[0123]
[0124] In the formula, D∈1-Lipsschitz makes the discriminator conform to the 1-Lipsschitz function, limits the range of the discriminator, and makes P data (x) and P g Not much difference.
[0125] The loss function of the generator is:
[0126]
[0127] The loss function of the discriminator is:
[0128]
[0129] B. Seasonal reduction of wind and solar load
[0130] The uncertain wind and solar data generated by the improved generative adversarial network and the historical load data at the same time throughout the year are combined into 8760 data pairs, which are sorted from small to large according to the time series. The Latin Hypercube Sampling (LHS) method is used to sample them to construct an initial data set with time series characteristics.
[0131] The specific steps of LHS are as follows:
[0132] Assume that the LHS sampling object size is N, there are k random variables, and the random variable x k The probability density function of is as follows:
[0133] y k =f k (x k ) (9)
[0134]
[0135] 7) For each random variable x k The probability space of is divided into N equal parts to form N subintervals, namely:
[0136] {[0,1 / N],[1 / N,2 / N],…,[(N-1) / N,1]}.
[0137] 8) Randomly select a sampling point y in each subinterval ki (i=1,2,…,N)
[0138] 9) Use the inverse transformation of the function shown in formula (2) to find the sampling points x in each interval ki , and arrange them in time series from small to large to obtain the initial sampling matrix S with sampling values of kⅹN order, that is, the initial data set of wind and solar load throughout the year.
[0139] The obtained wind-solar load data set for the whole year is normalized, and the self-organizing maps (SOM) clustering algorithm is used to reduce the scenes, and 4 groups of clustering results are obtained, namely, wind-solar load data sets with seasonal characteristics. The 4 groups of data sets correspond to the four seasons of spring, summer, autumn and winter, and in each data set, the wind-solar load data pairs have corresponding seasonal characteristics.
[0140] The SOM clustering algorithm automatically classifies data according to its characteristics. The basic principle is to group data into one category by classifying two data with a small Euclidean distance. The specific steps of the SOM clustering algorithm are as follows:
[0141] First, suppose there is a data set D, the samples in set D are x1,x2,…,x n , that is: D = {x1, x2, ..., x n}, each sample contains n elements, such as sample x i The element is x i1 ,x i2 ,…,x in , that is: x i =(x i1 ,x i2 ,…,x in ), similarly, sample x j The element is x j1 ,x j2 ,…,x jn , that is, x j =(x j1 ,x j2 ,…,x jn ).
[0142] Calculate sample x i and sample x j The Euclidean distance between i ,x j ) is expressed as follows:
[0143]
[0144] 1) Initialization: Initialize the SOM network connection weights to random values, and determine the initial learning value, the initial input vector value, and the number of output neurons, i.e., the number of clusters.
[0145] 2) Confirm the best matching unit (BMU). Use the Euclidean distance formula to calculate and compare the distances between domain nodes, find the minimum distance, and confirm the winning neuron.
[0146] 3) Weight update and adjustment. Update the node neuron weights within the BMU radius, and use the Gaussian function to shrink the range of each iteration. The neighborhood radius is expressed as:
[0147] r(n)=r0exp(-n / a) (12)
[0148] Where r(n) is the domain radius of the nth update; r0 is the initial value; and a is a constant.
[0149] 4) Repeat steps 2)-3) until the maximum number of iterations is reached and output the result.
[0150] 5) Finally, the number of clusters is obtained by calculating the Davidson-Bouldin index (DBI), and the cluster with the smallest DBI is selected as the final result of clustering. The calculation method of DBI is as follows:
[0151]
[0152] Where: N is the total number of clusters; Li,j describes the relationship between the i-th and j-th clusters in the clustering results. The smaller its value, the better the clustering situation; Ni,j is the distance between clusters i and j; Si is the Euclidean distance of cluster i. Ti is the size of the i-th cluster matrix; Yk is the k-th eigenvector in the cluster; Ci is the centroid of cluster i; Cj is the centroid of cluster j; p is the dimension of the cluster vector.
[0153] 2. Grid planning method considering multi-time scale energy storage coordination
[0154] Based on the typical seasonal scenario of wind and solar load considering uncertainty generated in the first part, the proposed grid planning model contains two problems. Problem A: Establish a multi-time scale hybrid energy storage model with short-term energy storage and long-term energy storage coordination. On this basis, Problem B: A grid planning method considering multi-time scale energy storage coordination is proposed.
[0155] A. Multi-time scale hybrid energy storage model
[0156] Short-term energy storage charges and discharges in a short time scale to maintain power balance in a short period of time. The time scale is hours. Its model is as follows:
[0157] 1) SOC constraint:
[0158] SOC0=0.5c (16)
[0159] SOC1=SOC 24 (17)
[0160] SOC min ≤SOC≤SOC max (18)
[0161] In the formula, SOC1 is the SOC state of the short-term energy storage at the initial moment, c is the capacity of the short-term energy storage, and SOC 24 Indicates the SOC state of short-term energy storage at the end of the day, SOC min , SOC max They are the maximum and minimum values allowed for short-term energy storage SOC respectively.
[0162] Formula 16 indicates that the initial moment of the short-term energy storage is half of the capacity, Formula 17 indicates that the SOC at the initial moment and the final moment of the short-term energy storage within a day is the same, and Formula 18 indicates that the short-term energy storage operates within the range allowed by the SOC.
[0163] 2) Power Constraints:
[0164] P ESS,min ≤P ESS ≤P ESS,max (19)
[0165] Where P ESS,min , P ESS,max are the minimum and maximum values of short-term energy storage charging and discharging, respectively, P ESS It is the charging and discharging power of short-term energy storage.
[0166] Long-term energy storage charges and discharges over a long time scale to maintain long-term power balance. The time scale is seasonal. Its model is as follows:
[0167] 1) SOC constraint:
[0168]
[0169] In the formula, is the SOC state of long-term energy storage at the initial time of the first typical day, c ESS is the capacity of long-term energy storage, is the SOC state of long-term energy storage at the end of the kth typical day, is the SOC state of the long-term energy storage at the initial time of the k+1th typical day, is the SOC state of long-term energy storage at the end of the fourth typical day, is the SOC state of long-term energy storage at time t on the kth typical day, is the SOC state of the long-term energy storage at time t-1 on the kth typical day, is the self-discharge rate of long-term energy storage, α ESS For the efficiency of long-term energy storage, and are the charge and discharge of long-term energy storage at time t on the kth typical day, Δt is the scheduling duration, and SOC ESS,min , SOC ESS,max They are respectively the maximum and minimum values allowed for long-term energy storage SOC.
[0170] Formula 20 indicates that the long-term energy storage is half of its capacity at the initial moment of the first typical day, Formula 21 indicates that the SOC of the long-term energy storage at the end of the k-th typical day and the initial moment of the k+1-th typical day is the same, Formula 22 indicates that the SOC of the long-term energy storage at the beginning of the first typical day and the end of the fourth typical day is the same, Formula 23 is the SOC calculation method of the long-term energy storage at time t on the k-th typical day, and Formula 24 indicates that the long-term energy storage operates within the allowable range of SOC.
[0171] 2) Power Constraints:
[0172]
[0173] In the formula, P is the charging and discharging state of long-term energy storage, which takes 1 or 0. When it takes 1, it means that the long-term energy storage is being charged and discharged, and when it takes 0, it means that the long-term energy storage is not being charged and discharged; ESS,max It is the maximum value of long-term energy storage charging and discharging.
[0174] B. Grid planning method considering multi-time scale energy storage coordination
[0175] Based on the multi-time-scale hybrid energy storage model obtained above, problem B in the proposed power grid planning model is described as follows.
[0176] 1. Objective Function
[0177] The objective function of the optimization problem is to minimize the total cost C of the power grid:
[0178]
[0179] Where C1 is the economic cost of the power grid, C2 is the green cost of the power grid; In,RE 、p In,Sto are the installation costs per unit capacity of wind and solar power and energy storage (including short-term and long-term energy storage); c RE 、c Sto are the installed capacity of wind and solar power, and energy storage (including short-term energy storage and long-term energy storage); p OM,RE 、p OM,Stoare the unit capacity operation and maintenance costs of wind and solar power and energy storage (including short-term energy storage and long-term energy storage); P RE , P Sto They are wind, solar and energy storage output (including short-term and long-term energy storage); p Buy,MG 、p Sel,MG are the electricity purchase and sales prices of the power grid, P Buy,MG , P Sel,MG are the electricity purchased and sold from the power grid respectively; μ is the penalty coefficient for wind and solar power abandonment; P Waste The amount of wind and solar power abandoned:
[0180] P Waste =P RE -P Load -P Sto -P Sel,MG (29)
[0181] Where P Load is the load power.
[0182] 2. Constraints
[0183] 1) Capacity constraints:
[0184]
[0185] Where: c RE,max is the upper limit of wind and solar installation capacity, c Sto Set an upper limit on the installed capacity of energy storage (including short-term and long-term storage).
[0186] 2) Power balance constraints:
[0187] P RE +P Sto -P Sel,MG +P Buy,MG =P Load (31)
[0188] 3) Wind and solar power output constraints:
[0189] P RE,min ≤P RE ≤P RE,max (32)
[0190] Where: P RE,min , P RE,max They are the minimum and maximum values of wind and solar power output respectively.
[0191] 4) Constraints on large-scale electricity purchase and sales:
[0192]
[0193] Where: P Buy,MG,max , P Buy,MG,minare the upper and lower limits of the amount of electricity purchased from the large power grid, P Sel,MG,max , P Sel,MG,min They are the upper and lower limits of electricity sold to the large power grid.
[0194] In addition to the above constraints, the constraints of the power grid planning model also include equations (9) to (20) mentioned above.
[0195] This specific embodiment is merely an explanation of the present invention and is not a limitation of the present invention. After reading this specification, those skilled in the art may make non-creative modifications to the present embodiment as needed. However, as long as they are within the scope of the claims of the present invention, they are protected by the patent law.
Claims
1. A power grid energy storage method considering the uncertainty and timing characteristics of wind and solar power, characterized by: This includes reducing the scenarios through the SOM clustering method to obtain typical seasonal scenarios of wind and solar loads, establishing a multi-time scale energy storage model considering long-term energy storage, and building a grid wind, solar and storage stochastic programming model taking into account the economic and low-carbon nature of grid investment; The specific steps include: S1. A method for generating seasonal scenarios of wind and solar load considering uncertainty; This includes proposing an improved generative adversarial network to generate wind and solar power output scenarios that take uncertainty into account. On this basis, the wind and solar load data are subjected to scenario reduction using a combination of Latin hypercube sampling and SOM clustering algorithms to obtain typical wind and solar load scenarios with seasonal characteristics. S2. Grid planning method considering multi-time scale energy storage coordination; Based on the typical seasonal scenario of wind and solar load considering uncertainty, the power grid is planned, and a multi-time scale hybrid energy storage model with the coordination of short-term energy storage and long-term energy storage is proposed. On this basis, a power grid planning method considering the coordination of multi-time scale energy storage is proposed.
2. A grid energy storage method considering wind and solar uncertainty and timing characteristics according to claim 1, characterized in that: In step S1, the improved generative adversarial network includes a traditional generative adversarial network including a generator and a discriminator: the generator receives input noise generated data, and the discriminator determines whether the data generated by the generator is true based on historical data. After the discriminator determines that the synthetic data is obtained, the generator and the discriminator will be updated by fixing each other until the discriminator can no longer determine whether the data given by the generator is synthetic data, and the generator and the discriminator stop updating; Assume that the historical wind and solar power output data is x, and its distribution is P data (x), random noise z adopts Gaussian distribution P z To describe, assume that the generator is G and the discriminator is D. During the training process, the noise z~P z Sent to the generator, outputting the generated wind and solar power output G(z)~P g At the same time, the historical wind and solar power output is sent to the discriminator, which judges the generated data. The loss function of the generator is: L G =-E z~p(z) [logD(G(z))] (1) In the formula, E is the mathematical expectation, D(G(z)) is the result of the discriminator on the generated data, The loss function of the discriminator is: In the formula, D(x) is the result of the discriminator on the real data, 1-D(G(z)) is the result of the discriminator on the synthetic data, The goal of the generator is to generate synthetic data that is more like real data, and the goal of the discriminator is to identify as much synthetic data as possible. Therefore, the objective function of the generative adversarial network is: The objective function solution process is equivalent to solving P data (x) and P g JS divergence problem: But P data (x) and P g It is difficult to have an ideal overlap, so the JS divergence easily becomes a constant, causing the discriminator to be unable to recognize the distance between the two distributions, resulting in gradient disappearance and poor training results. To solve this problem, the traditional generative adversarial network is improved: the Wasserstein distance is introduced to optimize the neural network. The Wasserstein distance is: Where P r ,P w are two different distributions, ∏(P r ,P w ) is the joint distribution of two distributions (P r ,P w ), γ is a possible distribution in the set, x is the real sample, y is the generated sample, (x, y) is generated from γ, ||xy|| is the distance between the two samples, E (x,y)~γ [||xy||] is the energy consumption, The improved generative adversarial network objective function is as follows: In the formula, D∈1-Lipsschitz makes the discriminator conform to the 1-Lipsschitz function, limits the range of the discriminator, and makes P data (x) and P g Not much difference, The loss function of the generator is: The loss function of the discriminator is: 。 3. A grid energy storage method considering wind and solar uncertainty and timing characteristics according to claim 1, characterized in that: In step S1, the wind-solar load data is subjected to scene reduction by combining Latin hypercube sampling method and SOM clustering algorithm to obtain typical wind-solar load scenes with seasonal characteristics. The specific steps are as follows: Scenery and seasonal scene reduction The uncertain wind and solar data generated by the improved generative adversarial network and the historical load data at the same time throughout the year are combined into 8760 data pairs, which are sorted from small to large according to the time series. The Latin Hypercube Sampling (LHS) method is used to sample them to construct an initial data set with time series characteristics. The specific steps of LHS are as follows: Assuming that the LHS sampling object size is N, there are k random variables, the probability density function of the random variable xk is as follows: y k =f k (x k ) (9) 1) For each random variable x k The probability space of is divided into N equal parts to form N subintervals, namely: {[0,1 / N],[1 / N,2 / N],…,[(N-1) / N,1]}, 2) Randomly select a sampling point y in each subinterval ki (i=1,2,…,N) 3) Use the inverse transformation of the function shown in formula (2) to find the sampling points x in each interval ki , and arrange them in time series from small to large, and get the initial sampling matrix S with sampling value of k×N order, that is, the initial data set of wind and solar load throughout the year, The obtained initial data set of wind-solar load for the whole year is normalized, and the self-organizing maps (SOM) clustering algorithm is used to reduce the scenes, and four groups of clustering results are obtained, namely, wind-solar load data sets with seasonal characteristics. The four groups of data sets correspond to the four seasons of spring, summer, autumn and winter. In each data set, the wind-solar load data pairs have corresponding seasonal characteristics. The SOM clustering algorithm automatically classifies data according to its characteristics. The basic principle is to classify two data with small Euclidean distance into one category, and use this method to group data. The specific steps of the SOM clustering algorithm are as follows: First, suppose there is a data set D, the samples in set D are x1,x2,…,x n , that is: D = {x1, x2, ..., x n }, each sample contains n elements, such as sample x i The element is x i1 ,x i2 ,…,x in , that is: x i =(x i1 ,x i2 ,…,x in ), similarly, sample x j The element is x j1 ,x j2 ,…,x jn , that is, x j =(x j1 ,x j2 ,…,x jn ), Calculate sample x i and sample x j The Euclidean distance between i ,x j ) is expressed as follows: 1) Initialization: Initialize the SOM network connection weights to random values, determine the initial learning value, the initial value of the input vector, and the number of output neurons, i.e., the number of clusters. 2) Confirm the best matching unit (BMU), calculate and compare the distances between domain nodes using the Euclidean distance formula, find the minimum distance, and confirm the winning neuron. 3) Weight update adjustment: update the node neuron weight within the BMU radius, and use the Gaussian function to shrink the range of each iteration. The neighborhood radius is expressed as: r(n)=r0exp(-n / a) (12) Where r(n) is the radius of the domain updated for the nth time; r0 is the initial value; a is a constant, 4) Repeat steps 2)-3) until the maximum number of iterations, and output the results. 5) Finally, the number of clusters is obtained by calculating the Davidson-Bouldin index (DBI), and the cluster with the smallest DBI is selected as the final result of clustering. The calculation method of DBI is as follows: N i,j =||C i -C j || p (15) Where: N is the total number of clusters; L i,j Describes the relationship between the i-th category and the j-th category in the clustering results. The smaller its value, the better the clustering situation. i,j is the distance between clusters i and j; S i is the Euclidean distance of cluster i, T i is the size of the i-th clustering matrix; Y k is the kth eigenvector in the cluster; C i is the centroid of cluster i; C j is the centroid of cluster j; p is the dimension of cluster vector.
4. A grid energy storage method considering wind and solar uncertainty and timing characteristics according to claim 1, characterized in that: The specific model establishment method of the multi-time scale hybrid energy storage model in step S2 is as follows: Short-term energy storage charges and discharges in a short time scale to maintain power balance in a short period of time. The time scale is hours. Its model is as follows: SOC constraints: SOC0=0.5c (16) SOC1=SOC 24 (17) SOC min ≤SOC≤SOC max (18) In the formula, SOC1 is the SOC state of the short-term energy storage at the initial moment, c is the capacity of the short-term energy storage, and SOC 24 Indicates the SOC state of short-term energy storage at the end of the day, SOC min , SOC max are the maximum and minimum values allowed for short-term energy storage SOC, Formula 16 indicates that the initial moment of short-term energy storage is half of the capacity, Formula 17 indicates that the SOC of the initial moment and the final moment of a day of short-term energy storage are the same, and Formula 18 indicates that the short-term energy storage operates within the range allowed by SOC. Power Constraints: P ESS,min ≤P ESS ≤P ESS,max (19) Where P ESS,min , P ESS,max are the minimum and maximum values of short-term energy storage charging and discharging, respectively, P ESS is the charging and discharging power of short-term energy storage, Long-term energy storage charges and discharges over a long time scale to maintain long-term power balance. The time scale is seasonal. Its model is as follows: SOC constraints: In the formula, is the SOC state of long-term energy storage at the initial time of the first typical day, c ESS is the capacity of long-term energy storage, is the SOC state of long-term energy storage at the end of the kth typical day, is the SOC state of the long-term energy storage at the initial time of the k+1th typical day, is the SOC state of long-term energy storage at the end of the fourth typical day, is the SOC state of long-term energy storage at time t on the kth typical day, is the SOC state of the long-term energy storage at time t-1 on the kth typical day, is the self-discharge rate of long-term energy storage, α ESS For the efficiency of long-term energy storage, and are the charge and discharge of long-term energy storage at time t on the kth typical day, Δt is the scheduling duration, and SOC ESS,min , SOC ESS,max are the maximum and minimum values allowed for long-term energy storage SOC, Formula 20 indicates that the long-term energy storage is half of its capacity at the initial moment of the first typical day. Formula 21 indicates that the SOC of the long-term energy storage at the end of the kth typical day is the same as that at the initial moment of the k+1th typical day. Formula 22 indicates that the SOC of the long-term energy storage at the beginning of the first typical day is the same as that at the end of the fourth typical day. Formula 23 is the SOC calculation method of the long-term energy storage at time t on the kth typical day. Formula 24 indicates that the long-term energy storage operates within the allowed range of SOC. Power Constraints: In the formula, P is the charge and discharge state of long-term energy storage, which takes 1 or 0. When it takes 1, it means that the long-term energy storage is being charged and discharged, and when it takes 0, it means that the long-term energy storage is not being charged and discharged; ESS,max is the maximum value of long-term energy storage charge and discharge, A grid planning method considering multi-time scale energy storage coordination Based on the multi-time scale hybrid energy storage model obtained above, problem B in the proposed power grid planning model is described as follows: Objective Function The objective function of the optimization problem is to minimize the total cost C of the power grid: Where C1 is the economic cost of the power grid, C2 is the green cost of the power grid; In,RE 、p In,Sto are the installation costs per unit capacity of wind and solar power and energy storage (including short-term and long-term energy storage); c RE 、c Sto are the installed capacity of wind and solar power, and energy storage (including short-term energy storage and long-term energy storage); p OM,RE 、p OM,Sto are the unit capacity operation and maintenance costs of wind and solar power and energy storage (including short-term energy storage and long-term energy storage); P RE , P Sto They are wind, solar and energy storage output (including short-term and long-term energy storage); p Buy,MG 、p Sel,MG are the electricity purchase and sales prices of the power grid, P Buy,MG , P Sel,MG are the electricity purchased and sold from the power grid respectively; μ is the penalty coefficient for wind and solar power abandonment; P Waste The amount of wind and solar power abandoned: P Waste =P RE -P Load -P Sto -P Sel,MG (29) Where P Load is the load power, Constraints Capacity constraints: Where: c RE,max is the upper limit of wind and solar installation capacity, c Sto The upper limit of energy storage installation capacity (including short-term energy storage and long-term energy storage) is Power balance constraints: P RE +P Sto -P Sel,MG +P Buy,MG =P Load (31) Wind and solar power output constraints: P RE,min ≤P RE ≤P RE,max (32) Where: P RE,min , P RE,max are the minimum and maximum values of wind and solar power output, Constraints on large-scale electricity purchase and sales: Where: P Buy,MG,max , P Buy,MG,min are the upper and lower limits of the amount of electricity purchased from the large power grid, P Sel,MG,max , P Sel,MG,min are the upper and lower limits of electricity sold to the large power grid, In addition to the above constraints, the constraints of the power grid planning model also include equations (9) to (20) mentioned above.
Citation Information
Patent Citations
Regional power grid wind and light absorption-oriented energy storage optimization configuration method and system
CN114529100A
Power distribution network stochastic planning method and device considering wind and light seasonal correlation
CN117239830A
Typical scene-based wind-solar-storage collaborative planning method and system
CN117540986A
Energy storage stochastic programming method considering wind power dynamic scene generation
CN118036972A
Wind-solar-storage multi-energy complementary system capacity configuration method suitable for remote areas
CN118213975A
