Hybrid energy storage system grading and control method based on probability prediction

Abstract

The invention belongs to the technical field of energy storage system control, and particularly relates to a hybrid energy storage system grading and control method based on probability prediction, which comprises the following steps: obtaining a wind power scene with time correlation based on probability prediction, and quantifying the uncertainty of wind power system control; using a wind power distribution method based on an empirical mode decomposition method to obtain pre-configured power of the wind power-hybrid energy storage system; constructing a wind power-hybrid energy storage two-stage random control optimization operation model, determining a grading decision of the wind power-hybrid energy storage system in the first stage, and optimizing the control action of the wind power-hybrid energy storage system in the second stage; and determining the power and capacity of the wind power-hybrid energy storage system to realize smooth wind power fluctuation. In order to verify the rationality of the method, the model provided by the invention is verified through an example of actual wind power data, and under the condition of considering the uncertainty of wind power, the installation cost and the operation and maintenance cost of HESS can be remarkably reduced, and the wind power fluctuation can be remarkably stabilized.

Description

Technical Field

[0001] The present invention belongs to the technical field of energy storage system control, and in particular relates to a hybrid energy storage system classification and control method based on probability prediction. Background Art

[0002] In the context of coping with the energy crisis and building a new power system, wind power has been rapidly developed worldwide due to its cleanliness and sustainability. The inherent volatility and intermittency of wind power pose a major challenge to the safety and economy of power system operation. How to smooth the fluctuation of wind power and improve the utilization rate of wind power has become a research hotspot; with the rapid development of energy storage technology, using energy storage systems to smooth wind power fluctuations has become a research frontier; different energy storage devices have their own unique advantages and disadvantages in terms of cost, life and efficiency. A single type of energy storage system cannot smooth wind power fluctuations well, so a hybrid energy storage system composed of multiple devices with different characteristics is proposed to alleviate the fluctuations of renewable energy; to consider the uncertainty of wind power, the existing technology uses classical robust optimization, deterministic optimization, random optimization and other methods to solve the problem of power system optimization scheduling under random conditions, but the classical optimization method relies on deterministic prediction and does not consider the inevitable prediction errors in practice; to reduce the conservatism of robust optimization, the selection of uncertainty sets is difficult; considering the increasing energy storage cost and operation and maintenance cost of energy storage systems, how to plan energy storage for wind power smoothing under the condition of considering economy is very important and must be considered. Therefore, there is an urgent need for a hybrid energy storage system classification and control method based on probability prediction, which effectively smooths wind power fluctuations with high economy while considering the uncertainty of wind power, and is more suitable for actual conditions. The wind power generation system can be applied to transmission system, distribution system or microgrid. Summary of the invention

[0003] The object of the present invention is to provide a hybrid energy storage system classification and control method based on probability prediction, which is characterized by comprising the following steps:

[0004] Step A: Obtain time-dependent wind power scenarios based on probability forecasting to quantify the uncertainty of wind power system control; use a wind power allocation method based on empirical mode decomposition to obtain the pre-configured power of the wind power-hybrid energy storage system;

[0005] Step B: Construct a two-stage stochastic control optimization operation model for wind power-hybrid energy storage, determine the hierarchical decision of the wind power-hybrid energy storage system in the first stage, and optimize the control action of the wind power-hybrid energy storage system in the second stage; determine the power and capacity of the wind power-hybrid energy storage system to achieve smooth wind power fluctuations.

[0006] The wind power scenario with time correlation obtained based on probability prediction includes:

[0007] Using prediction quantiles Approximate the probability distribution of wind power generation and obtain and predict quantiles Matching wind power cumulative probability distribution function F t , using direct quantile regression method to obtain wind power forecast quantile series Based on the quantile series of wind power forecast The linear interpolation method is used to approximate the entire wind power prediction CDF curve to complete the probability prediction. Based on the probability prediction results, a wind power scenario reflecting the randomness and time correlation of wind power is generated. The wind power scenario is a set of time series with random and probabilistic characteristics.

[0008] The predicted quantile for:

[0009]

[0010] Where: Pr is the probability operator; x t is the random wind power at time t; α is the quantile The nominal ratio of It is the inverse function of the cumulative probability distribution function of wind power at time t;

[0011] The predicted quantile series for:

[0012]

[0013] Where: is the actual quantile estimate; is the nominal proportion α that needs to be estimated r The predicted quantile sequence of (r=0,1,...R), where R is infinite.

[0014] The wind power allocation method based on the empirical mode decomposition method includes:

[0015] Based on the probability prediction, a wind power scenario with time correlation is obtained. A clustering method is used to construct a typical wind power scenario set, extract the frequency components, and use the empirical mode decomposition method to process the original wind power signal P W.O.n , obtain the IMF components of each order, reconstruct the IMF according to the preset wind power fluctuation regulations, allocate the power that energy storage needs to smooth according to the high-frequency fluctuations, and obtain the pre-configured power of the wind power-hybrid energy storage system;

[0016] The reconstructed IMF includes:

[0017] After obtaining the high-frequency reconstruction f2c and the low-frequency reconstruction c2f, for c2f(1), calculate the fluctuation of c2f(1). If the grid connection limit is met, there is no direct grid connection power, and the original wind power signal is all connected to the grid after being smoothed by the HESS. If it is not met, it is ignored. For c2f(k), calculate the fluctuation of c2f(k) respectively. If the grid connection limit is met, no processing is done and it is ignored. If the grid connection limit is not met, the direct grid connection power P is allocated. W.I.n =c2f(k-1), HESS allocated power P HESS =f2c(p+1-k), thereby obtaining the pre-configured power of the wind power-hybrid energy storage system.

[0018] The wind power original power signal P W.O.n for:

[0019] P W.O.n =P W.I.n +P HESS,n (1)

[0020] Where: P W.O.n is the original wind power signal at time n; P W,I,n The low frequency signal is not processed at time n, P HESS,n Allocate pre-capacity and power to HESS at time n;

[0021] The high frequency reconstruction is shown in formula (5):

[0022]

[0023] The low frequency reconstruction is shown in formula (6):

[0024]

[0025] Where f2c (p) is the p-order high-frequency reconstruction component, c2f (p) is the p-order low-frequency reconstruction component, res is the decomposition residual component, IMF p is the p-order intrinsic mode function.

[0026] The construction of the wind power-hybrid energy storage two-stage stochastic control optimization operation model includes:

[0027] 1) Phase I HESS classification: Phase I determines the HESS classification before any uncertainty occurs; in all scenarios, the rated capacity E, rated power P and initial energy level E of the HESS are determined. s ;

[0028] 2) Second stage HESS control: The second stage optimizes the control action of HESS to smooth the fluctuations after the uncertainty of wind power realization; respectively determine the optimized grid-connected wind power Battery energy storage system optimizes charging and discharging power P t BESS , the optimized charging and discharging power P of supercapacitor t SC , obtain the optimized charging and discharging power control action of HESS, as well as the deviation between the scheduled power and the optimized charging and discharging power of the battery energy storage system Deviation between the preset power and the optimized charge and discharge power of the supercapacitor

[0029] In the wind power-hybrid energy storage two-stage stochastic control optimization operation model, the inherent energy storage cost of the HESS system is:

[0030]

[0031] In the formula, The daily cost per unit capacity of energy storage is is the capacity cost of lithium battery and supercapacitor; W n It represents the average number of operating days per year for a certain energy storage device n; is the daily amortized cost per unit power, is the energy storage power cost; T n Management information system energy storage equipment project life; r is the risk discount rate, and the set M represents the energy storage equipment type, including BESS and SC types;

[0032] The control objective function of the wind power-hybrid energy storage two-stage stochastic control optimization operation model applied to the HESS system is as follows:

[0033]

[0034] In the formula, and are the daily amortized cost of energy storage capacity of BESS and SC respectively; and is the rated capacity and rated power of BESS; and is the rated capacity and rated power of SC; and are the daily amortized cost of energy storage power of BESS and SC respectively; and are the unit operating costs associated with the total energy storage of BESS and SC respectively; C pun P is the unit power penalty for the deviation between the optimal grid-connected wind power determined by TSSO and the target preset value; t BESS and P t SC is the charging / discharging power of BESS and SC at time t; is the target grid-connected wind power at time t, P t W is the wind power-HESS system optimized grid-connected wind power at time t; the time interval Δt is 5 minutes; x is the wind power random variable, E[f s (x)] is the expected term, which is approximately f s The probability weighted average of typical wind power scenarios, f s (x) is the energy storage operation and maintenance cost, f M is the energy storage investment cost; T is the life of the energy storage device;

[0035] Expected E[f s (x)] can be approximated as f s The probability weighted average of typical wind power scenarios is expressed as:

[0036]

[0037] Where N * is the number of samples required for the experiment; π n is the ratio of each sample n required for the experiment to the total number of samples, that is, π n =n n / N; The target power value set for a specific sample n at time t; is the actual power value of a specific sample n at time t; T is the life of the energy storage device.

[0038] The constraints of the wind power-hybrid energy storage two-stage random control optimization operation model include: power balance constraint, energy conversion constraint, HESS energy state constraint, HESS power constraint, optimization of HESS charging and discharging power constraint, and BESS life cycle constraint;

[0039] The power balancing constraint is:

[0040]

[0041] Where: and are the power deviations of the pre-planned and optimized charging and discharging powers for BESS and SC, respectively; is the pre-power of BESS of sample i at time t; is the pre-power of SC of sample i at time t; is the wind power of sample i at time t, is the optimized wind power grid-connected power;

[0042] The energy conversion constraint is:

[0043]

[0044] Where: is the energy level of device m at time t; m is a binary variable, λ m =1, the device is in the charging state, otherwise, the energy storage device is in the discharging state; η m is the charging power conversion rate of device m, if P t m >0, then P t m is the optimized energy storage charging power, otherwise, its value is the discharge power, P t *m is a continuous variable, used to replace With P t m The set M represents the type of energy storage equipment, including BESS and SC types;

[0045] The HESS energy state constraint is:

[0046]

[0047] Where: and Represent the minimum state of charge and maximum state of charge of energy storage respectively; and are the capacity and rated capacity of the equipment at time t, and They represent the equipment capacity at the beginning and end respectively; the set M represents the type of energy storage equipment, including BESS and SC types;

[0048] The HESS power constraint is:

[0049]

[0050] Where: P t m and are the power and rated power of equipment m at time t respectively;

[0051] The optimized HESS charging and discharging power constraint is:

[0052]

[0053] The optimized HESS power control action cannot exceed equation (16);

[0054] The BESS life cycle constraints are:

[0055]

[0056] Where: q tis the sign of the charge and discharge state change of BESS. If it is equal to 1, it changes from the discharge state to the charge state at time t; N0 is the cycle life of BESS; Q t The sum of is the number of BESS cycles in one day, λ t BESS is the BESS state quantity, W BESS is the average number of working days of BESS in a year, T BESS is the estimated life of BESS, are two 0-1 integer variables, and C is a large constant.

[0057] The beneficial effects of the present invention are:

[0058] The present invention proposes a hybrid energy storage system (HESS) classification and control method based on probability prediction, a two-stage random control optimization operation of HESS classification and control based on probability prediction, and an optimal control effect of wind power fluctuation smoothing is integrated into the model formula. Based on the probability prediction of wind power, the EMD method is used to determine the pre-power of the corresponding HESS, so as to propose a two-stage random control optimization operation model of wind power-hybrid energy storage in combination with typical wind power scenarios. The use of random optimization relies on the accurate probability density function of random variables. Probabilistic prediction can accurately estimate the probability density function of wind power, which brings favorable conditions for the application of random optimization, thereby solving the problems described in the background technology and being able to not rely on deterministic predictions.

[0059] A hybrid energy storage system classification and control method based on probability prediction disclosed in the present invention is applied to economically and effectively smooth out wind power fluctuations, with the goal of minimizing total cost and smoothing wind power fluctuations. A random wind power scenario is obtained based on probability prediction, and a typical wind power scenario set is constructed by a clustering method. For each typical wind power scenario, an empirical mode decomposition (EMD) method is used for analysis to obtain the pre-planned capacity and pre-dispatched power of the HESS. Considering the operating constraints of the equipment, the classification results, and the pre-planned capacity and power of the HESS, a two-stage random control optimization operation model of wind power-hybrid energy storage is constructed to optimize the control actions of the equipment, determine the capacity of the HESS and correct its predetermined power. The model proposed by the present invention is verified by a calculation example of actual wind power data. Taking into account the uncertainty of wind power, the installation cost and operation and maintenance cost of the HESS can be significantly reduced, and wind power fluctuations can be significantly smoothed. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 This is a flow chart of the wind power-hybrid storage two-stage random control optimization operation of the present invention;

[0061] Figure 2It is the flow chart of wind power allocation strategy based on EMD;

[0062] Figure 3 This is the structural diagram of the wind power-hybrid energy storage two-stage stochastic control optimization operation model;

[0063] Figure 4 This is the structural diagram of the wind power-hybrid energy storage system;

[0064] Figure 5 This is a statistical diagram of the order of low-frequency reconstructed components obtained by EMD decomposition for typical day 7;

[0065] Figure 6 Comparison diagram of wind power, low-fluctuation direct grid-connected power, and high-fluctuation HESS smoothing power, where (a) is a statistical diagram of wind power, (b) is a statistical diagram of low-fluctuation direct grid-connected power, and (c) is a statistical diagram of high-fluctuation HESS smoothing power;

[0066] Figure 7 It is the charging and discharging power curve of supercapacitor;

[0067] Figure 8 This is the expansion diagram of the high-frequency power fluctuation curve before and after smoothing;

[0068] Fig. 9 This is the expansion diagram of the wind power grid-connected power fluctuation curve after smoothing; DETAILED DESCRIPTION

[0069] The present invention provides a hybrid energy storage system classification and control method based on probability prediction, and the present invention is further described in detail below in conjunction with the accompanying drawings.

[0070] like Figure 1 The embodiment of the present invention shown discloses a hybrid energy storage system classification and control method based on probability prediction, including:

[0071] Step A: Obtain time-dependent wind power scenarios based on probability forecasting to quantify the uncertainty of wind power system control; use a wind power allocation method based on empirical mode decomposition to obtain the pre-configured power of the wind power-hybrid energy storage system;

[0072] Step B: Construct a two-stage stochastic control optimization operation model for wind power-hybrid energy storage, determine the hierarchical decision of the wind power-hybrid energy storage system in the first stage, and optimize the control action of the wind power-hybrid energy storage system in the second stage; determine the power and capacity of the wind power-hybrid energy storage system to achieve smooth wind power fluctuations.

[0073] In this embodiment, step A quantifies the uncertainty of wind power system control based on probabilistic wind power forecasting. Use the forecast quantile to approximate the probability distribution of wind power generation; use the clustering method to obtain the cumulative probability distribution function (CDF) of wind power for a typical wind power scenario set; use the empirical mode decomposititon (EMD) method based on the probability distribution function to process typical wind power scenario data, obtain the intrinsic mode function (IMF) of each order, and then reconstruct the IMF according to the relevant regulations on wind power fluctuations, and eliminate the fluctuations that need to be smoothed by energy storage according to the maximum fluctuation limit, so as to determine the pre-configured capacity and power of the wind power-hybrid energy storage system, and obtain the pre-configured power of the wind power-hybrid energy storage system.

[0074] The wind power scenario with time correlation obtained based on probability prediction includes:

[0075] Using prediction quantiles Approximate the probability distribution of wind power generation and obtain and predict quantiles Matching wind power cumulative probability distribution function F t , using direct quantile regression method to obtain wind power forecast quantile series Based on the quantile series of wind power forecast The linear interpolation method is used to approximate the entire wind power prediction CDF curve to complete the probability prediction, and a wind power scenario reflecting the randomness and time correlation of wind power is generated based on the probability prediction result. The wind power scenario is a set of time series with random and probabilistic characteristics;

[0076] In this embodiment, the prediction quantile is used The wind power probability forecast is estimated by approximating the distribution, and the cumulative probability distribution function of wind power is obtained, and then a wind power scenario with time correlation is generated. The CDF is expressed as F t , predicted quantile Match with CDF.

[0077] The predicted quantile for:

[0078]

[0079] Where: Pr is the probability operator; x t is the random wind power at time t; α is the quantile The nominal ratio of It is the inverse function of the cumulative probability distribution function of wind power at time t.

[0080] In this embodiment, the predicted quantile sequence is obtained according to the direct quantile regression method. There is a one-to-one correspondence with the values ​​on the CDF curve, so the entire wind power prediction CDF curve can be obtained using linear interpolation.

[0081] The predicted quantile series for:

[0082]

[0083] Where: is the actual quantile estimate; is the nominal proportion α that needs to be estimated r The predicted quantile sequence of (r=0,1,...R), where R is infinite.

[0084] In this embodiment, Corresponding to the discrete values ​​of the horizontal coordinates on the CDF curve, the entire wind power prediction CDF curve can be approximated by linear interpolation. The entire process of generating CDF is the probability prediction process. The wind power cumulative probability distribution function CDF is generated through the process of formula (1) to formula (3).

[0085] like Figure 2 As shown, the wind power allocation method based on the empirical mode decomposition method includes:

[0086] Based on the probability prediction, a wind power scenario with time correlation is obtained. A clustering method is used to construct a typical wind power scenario set, extract the frequency components, and use the empirical mode decomposition method to process the original wind power signal P W.O.n , obtain the IMF components of each order, reconstruct the IMF according to the preset wind power fluctuation regulations, allocate the power that energy storage needs to smooth according to the high-frequency fluctuations, and obtain the pre-configured power of the wind power-hybrid energy storage system;

[0087] The reconstructed IMF includes:

[0088] After obtaining the high-frequency reconstruction f2c and the low-frequency reconstruction c2f, for c2f(1), calculate the fluctuation of c2f(1). If the grid connection limit is met, there is no direct grid connection power, and the original wind power signal is all connected to the grid after being smoothed by the HESS. If it is not met, it is ignored. For c2f(k), calculate the fluctuation of c2f(k) respectively. If the grid connection limit is met, no processing is done and it is ignored. If the grid connection limit is not met, the direct grid connection power P is allocated. W.I.n =c2f(k-1), HESS allocated power P HESS =f2c(p+1-k), thereby obtaining the pre-configured power of the wind power-hybrid energy storage system.

[0089] In this embodiment, based on the wind power scenarios with time correlation obtained by probability prediction, a clustering method is used to construct a typical wind power scenario set, frequency components are extracted, and the frequency signal is processed using the EMD decomposition method to obtain IMFs of various orders. Then, according to the relevant regulations on wind power fluctuations, the IMF is reconstructed, and the power that needs to be smoothed by energy storage is allocated according to high-frequency fluctuations.

[0090] In this embodiment, based on the above-mentioned probability prediction and signal decomposition, combined with my country's wind power fluctuation regulations, the pre-configured capacity and power of the wind power-hybrid energy storage system are obtained, and the original wind power signal is decomposed using EMD to obtain the inherent mode functions IMF of each order, and then the IMF is reconstructed to obtain low-frequency and high-frequency components. The low-frequency grid-connected and high-frequency are used as energy storage power tasks. The reconstructed IMF response corresponds to the direct grid-connected power and the energy storage smoothing component as shown in the process Figure 2 As shown, the energy storage smoothing component determines the pre-configured capacity and power of the energy storage system.

[0091] The wind power original power signal P W.O.n for:

[0092] P W.O.n =P W.I.n +P HESS,n (2)

[0093] Where: P W.O.n is the original wind power signal at time n; P W,I,n The low frequency signal is not processed at time n, P HESS,n Allocate pre-capacity and power to the HESS at time n.

[0094] In this embodiment, the row vector (wind turbine fluctuation curve) is brought in, and then the reconstructed IMF is fed back. The IMF reconstruction is high-frequency and low-frequency components from top to bottom. Therefore, during reconstruction, the high-frequency accumulation curve is accumulated from top to bottom, or the low-frequency accumulation curve is accumulated from bottom to top.

[0095] The high frequency reconstruction is shown in formula (5):

[0096]

[0097] The low frequency reconstruction is shown in formula (6):

[0098]

[0099] Where f2c (p) is the p-order high-frequency reconstruction component, c2f (p) is the p-order low-frequency reconstruction component, res is the decomposition residual component, IMF p is the p-order intrinsic mode function;

[0100] Step B: Construct a two-stage stochastic control optimization operation model for wind power-hybrid energy storage, determine the hierarchical decision of the wind power-hybrid energy storage system in the first stage, and optimize the control action of the wind power-hybrid energy storage system in the second stage; determine the power and capacity of the wind power-hybrid energy storage system to achieve smooth wind power fluctuations.

[0101] In this embodiment, step B is combined with the pre-configured power of the wind power-hybrid energy storage system in step A, while taking into account the intermittent nature of wind power and the installation and operation costs of energy storage planning, taking into account both economic efficiency and leveling effects, with a smaller total cost of the wind storage system; constructing an objective function, and constructing a two-stage stochastic control optimization operation model of wind power-hybrid energy storage, so as to correct the pre-configured power and determine the power and capacity of the wind power-hybrid energy storage system.

[0102] In this embodiment, the objective function is constructed according to the energy storage cost and operation and maintenance cost of the HESS system; and the size of the HESS is determined in the first stage decision before the uncertainty is exposed. The size of the HESS is the rated power and rated capacity of each energy storage system in the hybrid energy storage determined in the first stage, and the capacity of each energy storage device of the HESS is completed to determine the hierarchical decision of the wind power-hybrid energy storage system. According to the predetermined power of the HESS, the action of the HESS is controlled; after the uncertainty occurs, the two-stage idea of ​​optimizing in the second stage to smooth the fluctuation of wind power is constructed to construct a two-stage random control optimization operation model of wind power-hybrid energy storage.

[0103] In this embodiment, the wind power-hybrid energy storage two-stage stochastic control optimization operation model for HESS classification and control is divided into two stages. The structure of the wind power-hybrid energy storage two-stage stochastic control optimization operation model is as follows: Figure 3 As shown; the structure of wind power-hybrid energy storage system is as follows Figure 4 As shown;

[0104] The construction of the wind power-hybrid energy storage two-stage stochastic control optimization operation model includes:

[0105] 1) Phase I HESS classification: Phase I determines the HESS classification before any uncertainty occurs; in all scenarios, the rated capacity E, rated power P and initial energy level E of the HESS are determined. s ;

[0106] In this embodiment, the capacity and power allocation in the first stage considers the frequency conditions under almost all wind power scenarios, and the variables in the first stage are related to the decision-making at this time and place;

[0107] 2) Second stage HESS control: The second stage optimizes the control action of HESS to smooth the fluctuations after the uncertainty of wind power realization; respectively determine the optimized grid-connected wind power Battery energy storage system optimizes charging and discharging power Pt BESS , the optimized charging and discharging power P of supercapacitor t SC , obtain the optimized charging and discharging power control action of HESS, as well as the deviation between the scheduled power and the optimized charging and discharging power of the battery energy storage system Deviation between the preset power and the optimized charge and discharge power of the supercapacitor

[0108] In this embodiment, the second-stage variables are specific to some scenarios, representing the decision after the realization of uncertain wind power generation, and can only make a final decision on the power of HESS. Therefore, the optimized HESS charging and discharging power P t BESS and P t SC Only time is used as an indicator. Based on the hierarchical decision of HESS in the first stage, the control action of HESS in the second stage is optimized. Then the HESS is controlled to achieve the purpose of smoothing wind power fluctuations.

[0109] In the wind power-hybrid energy storage two-stage stochastic control optimization operation model, the inherent energy storage cost of the HESS system is:

[0110]

[0111] In the formula, The daily cost per unit capacity of energy storage is is the capacity cost of lithium battery and supercapacitor; W n It represents the average number of operating days per year for a certain energy storage device n; is the daily amortized cost per unit power, is the energy storage power cost; T n Management information system energy storage equipment project life; r is the risk discount rate, and the set M represents the energy storage equipment type, including BESS and SC types;

[0112] The control objective function of the wind power-hybrid energy storage two-stage stochastic control optimization operation model applied to the HESS system is as follows:

[0113]

[0114] In the formula, and are the daily amortized cost of energy storage capacity of BESS and SC respectively; and is the rated capacity and rated power of BESS; and is the rated capacity and rated power of SC; and are the daily amortized cost of energy storage power of BESS and SC respectively; and are the unit operating costs associated with the total energy storage of BESS and SC respectively; C pun P is the unit power penalty for the deviation between the optimal grid-connected wind power determined by TSSO and the target preset value; t BESS and P t SC is the charging / discharging power of BESS and SC at time t; is the target grid-connected wind power at time t, P t W is the wind power-HESS system optimized grid-connected wind power at time t; the time interval Δt is 5 minutes; x is the wind power random variable, E[f s (x)] is the expected term, which is approximately f s The probability weighted average of typical wind power scenarios, f s (x) is the energy storage operation and maintenance cost, f M is the energy storage investment cost; T is the life of the energy storage device;

[0115] Expected E[f s (x)] can be approximated as f s The probability weighted average of typical wind power scenarios is expressed as:

[0116]

[0117] Where N * is the number of samples required for the experiment; π n is the ratio of each sample n required for the experiment to the total number of samples, that is, π n =n n / N; The target power value set for a specific sample n at time t; is the actual power value of a specific sample n at time t; T is the life of the energy storage device.

[0118] In this embodiment, the control objective function includes the investment cost of energy storage, the operation and maintenance cost of energy storage, and the penalty for wind power fluctuation.

[0119] The constraints of the wind power-hybrid energy storage two-stage random control optimization operation model include: power balance constraints, energy conversion constraints, HESS energy state constraints, HESS power constraints, optimized HESS charging and discharging power constraints, and BESS life cycle constraints.

[0120] The power balancing constraint is:

[0121]

[0122] Where: and are the power deviations of the pre-planned and optimized charging and discharging powers for BESS and SC, respectively; is the pre-power of BESS of sample i at time t; is the pre-power of SC of sample i at time t; is the wind power of sample i at time t, is the optimized wind power grid-connected power;

[0123] The energy conversion constraint is:

[0124]

[0125] Where: is the energy level of device m at time t; m is a binary variable, λ m =1, the device is in the charging state, otherwise, the energy storage device is in the discharging state; η m is the charging power conversion rate of device m, if P t m >0, then P t m is the optimized energy storage charging power, otherwise, its value is the discharge power, P t *m is a continuous variable, used to replace With P t m The set M represents the type of energy storage equipment, including BESS and SC types;

[0126] The HESS energy state constraint is:

[0127]

[0128] Where: and Represent the minimum state of charge and maximum state of charge of energy storage respectively; and are the capacity and rated capacity of the equipment at time t, and They represent the equipment capacity at the beginning and end respectively; the set M represents the type of energy storage equipment, including BESS and SC types;

[0129] The HESS power constraint is:

[0130]

[0131] Where: P t m and are the power and rated power of equipment m at time t respectively;

[0132] The optimized HESS charging and discharging power constraint is:

[0133]

[0134] The optimized HESS power control action cannot exceed equation (16);

[0135] The BESS life cycle constraints are:

[0136]

[0137] Where: q t is the sign of the charge and discharge state change of BESS. If it is equal to 1, it changes from the discharge state to the charge state at time t; N0 is the cycle life of BESS; Q t The sum of is the number of BESS cycles in one day, is the BESS state quantity, W BESS is the average number of working days of BESS in a year, T BESS is the estimated life of BESS, are two 0-1 integer variables, and C is a large constant;

[0138] The control objective function of the HESS system controlled by the two-stage stochastic control optimization operation model of wind power-hybrid energy storage is adopted. Power balance constraint, energy conversion constraint, HESS energy state constraint, HESS power constraint, optimized HESS charging and discharging power constraint and BESS life cycle constraint are used as the constraint conditions of the wind power-hybrid energy storage system. The cplex solver is used to optimize and solve the model to determine the system cost and the ability to smooth wind power fluctuations.

[0139] In order to enable those skilled in the art to better understand the present invention and appreciate the advantages of the present invention over the prior art, the applicant further explains the present invention in conjunction with specific embodiments.

[0140] This embodiment uses the actual wind power data of a wind farm in northern Shanxi Province in 2023 as an example for simulation analysis. First, facing the wind power scene, probability prediction is performed to obtain the wind power probability scenario. The 5-minute data of wind power in 2023 is clustered by the K-means clustering algorithm and divided into 8 clusters. The total number of days and the proportion of each typical scenario in the whole year are shown in Table 1.

[0141] Table 1 Statistics of days in different scenarios

[0142]

[0143] After clustering the 365-day 5-minute data of wind power into 8 clusters through K-means, the power and capacity of HESS are pre-determined in the first stage. The 7th day is selected as a typical day for EMD decomposition, and the corresponding wind power, low-fluctuation direct grid-connected power, and high-fluctuation energy storage smoothing power can be obtained, and then low-frequency reconstruction, maximum fluctuation calculation, and fluctuation limit selection are performed, such as Figure 6 As shown, (a) is the wind power statistics, (b) is the low-fluctuation direct grid-connected power statistics, and (c) is the high-fluctuation HESS smoothing power statistics. The 5-minute wind power fluctuation limit is selected as the critical value, and the low-frequency reconstruction component greater than the wind power fluctuation critical value is excluded, and the value less than is selected. Therefore, c2f is selected (2) ,c2f (3) ,c2f (4) ,c2f (5) and res (i.e. IMF2, IMF3, IMF4, IMF5, res) are not used as the capacity and power of the hybrid energy storage configuration, select f2c (1) (i.e. IMF1) is used as the basis for pre-configuration of capacity and power of HESS. Figure 5 As shown. Then, according to the capacity and power configuration tasks, combined with the supercapacitor charging and discharging power curve, the capacity and power of HESS are pre-configured, and the optimal configuration of supercapacitor capacity is 42.07MW. Figure 7 shown.

[0144] In the second stage, the charging and discharging power control actions of the HESS are determined after uncertainty is realized to smooth the fluctuation of wind power. Table 2 below shows the HESS capacity and power configuration proposed for the actual wind power data of a wind farm in northern Shanxi Province. Scheme 1 is the configuration obtained by the two-stage stochastic control optimization operation model of wind power-hybrid energy storage proposed in this invention; Scheme 2 is a random optimization, which takes the optimized grid-connected wind power as the preset value; Scheme 3 is a random optimization based on a wind power scenario without time correlation.

[0145] From the data in Table 1, we can see that:

[0146] In Scheme 2, the optimized grid-connected wind power is consistent with the predetermined value, eliminating the penalty for wind power fluctuations. However, a larger HESS needs to be installed in Scheme 2, resulting in a total cost that is 39.7% higher than that in Scheme 1. It can be concluded that the stochastic optimal scale and control method proposed in the present invention can achieve a better effect of smoothing wind power fluctuations at a lower total cost. In addition, since the grid-connected wind power in Scheme 2 is exactly set to the predetermined value, the strict power fluctuation smoothing requirement will increase the number of BESS cycles by 8 times compared to Case 1.

[0147] Scheme 3 adopts a typical wind power scenario without time correlation, and its installation cost and wind power fluctuation penalty are 14.3% and 162.3% higher than those of Scheme 1, respectively, and the total cost is 59% higher than that of Scheme 1. In fact, if the wind power scenario cannot reflect the time correlation well, the wind power fluctuation in continuous time periods cannot be well captured, resulting in inappropriate size and control effect of HESS. Therefore, it can be found that the stochastic optimization control model based on probability prediction proposed in the present invention can well balance the losses caused by wind power fluctuations and the installation cost of large HESS.

[0148] Table 2 Comparison of HESS optimization results under different models

[0149]

[0150] The high-frequency power fluctuations before and after the smoothing, such as Figure 8 The wind power grid-connected power fluctuation curve after smoothing is shown in Fig. 9 As shown in the figure, it can be seen that the proposed wind power-hybrid energy storage two-stage random control optimization operation model has a good effect on wind power peak shaving and valley filling. And after hybrid energy storage: the pre-power and capacity of HESS are configured in the first stage; the operation constraints of HESS are considered in the second stage, and the control action of wind turbine smoothing is optimized according to the classification results of wind turbines and the pre-planned power of HESS. After the smoothing, the wind power participating in the grid connection has fully met the grid connection fluctuation limit, with good cost savings and power smoothing effects.

[0151] In summary, by applying the hybrid energy storage system classification and control method based on probability prediction disclosed in the present invention, the penalty for wind power fluctuation is only 30% of the total cost, which verifies the effectiveness of HESS charging and discharging control actions in alleviating wind power fluctuation. In addition, the hybrid energy storage system classification and control method based on probability prediction proposed in the present invention takes into account the optimal control of HESS when selecting HESS to smooth out wind power fluctuations, and ensures the lowest installation cost of HESS on the basis of effectively reducing the penalty for wind power fluctuations.

Claims

1. A hybrid energy storage system classification and control method based on probability prediction, characterized in that: The steps include: Step A: Obtain wind power scenarios with time correlation based on probability forecasting to quantify the uncertainty of wind power system control; use wind power allocation method based on empirical mode decomposition method to obtain the pre-configured power of wind power-hybrid energy storage system; Step B: Construct a two-stage stochastic control optimization operation model for wind power-hybrid energy storage, determine the hierarchical decision of the wind power-hybrid energy storage system in the first stage, and optimize the control action of the wind power-hybrid energy storage system in the second stage; determine the power and capacity of the wind power-hybrid energy storage system to achieve smooth wind power fluctuations.

2. The hybrid energy storage system classification and control method based on probability prediction according to claim 1 is characterized in that: The wind power scenario with time correlation obtained based on probability prediction includes: Using prediction quantiles Approximate the probability distribution of wind power generation and obtain and predict quantiles Matching wind power cumulative probability distribution function F t , using direct quantile regression method to obtain wind power forecast quantile series Based on the quantile series of wind power forecast The linear interpolation method is used to approximate the entire wind power prediction CDF curve to complete the probability prediction. Based on the probability prediction results, a wind power scenario reflecting the randomness and time correlation of wind power is generated. The wind power scenario is a set of time series with random and probabilistic characteristics.

3. The hybrid energy storage system classification and control method based on probability prediction according to claim 2 is characterized in that: The predicted quantile for: Where: Pr is the probability operator; x t is the random wind power at time t; α is the quantile The nominal ratio of It is the inverse function of the cumulative probability distribution function of wind power at time t; The predicted quantile series for: Where: is the actual quantile estimate; is the nominal proportion α that needs to be estimated r The predicted quantile sequence of (r=0,1,...R), where R is infinite.

4. The hybrid energy storage system classification and control method based on probability prediction according to claim 1 is characterized in that: The wind power allocation method based on the empirical mode decomposition method includes: Based on the probability prediction, a wind power scenario with time correlation is obtained. A clustering method is used to construct a typical wind power scenario set, extract the frequency components, and use the empirical mode decomposition method to process the original wind power signal P W.O.n , obtain the IMF components of each order, reconstruct the IMF according to the preset wind power fluctuation regulations, allocate the power that energy storage needs to smooth according to the high-frequency fluctuations, and obtain the pre-configured power of the wind power-hybrid energy storage system; The reconstructed IMF includes: After obtaining the high-frequency reconstruction f2c and the low-frequency reconstruction c2f, for c2f(1), calculate the fluctuation of c2f(1). If the grid connection limit is met, there is no direct grid connection power, and the original wind power signal is all connected to the grid after being smoothed by the HESS. If it is not met, it is ignored. For c2f(k), calculate the fluctuation of c2f(k) respectively. If the grid connection limit is met, no processing is done and it is ignored. If the grid connection limit is not met, the direct grid connection power P is allocated. W.I.n =c2f(k-1), HESS allocated power P HESS =f2c(p+1-k), thereby obtaining the pre-configured power of the wind power-hybrid energy storage system.

5. The hybrid energy storage system classification and control method based on probability prediction according to claim 4 is characterized in that: The wind power original power signal P W.O.n for: P W.O.n =P W.I.n +P HESS,n (1) Where: P W.O.n is the original wind power signal at time n; P W,I,n The low frequency signal is not processed at time n, P HESS,n Allocate pre-capacity and power to HESS at time n; The high frequency reconstruction is shown in formula (5): The low frequency reconstruction is shown in formula (6): Where f2c (p) is the p-order high-frequency reconstruction component, c2f (p) is the p-order low-frequency reconstruction component, res is the decomposition residual component, IMF p is the p-order intrinsic mode function.

6. The hybrid energy storage system classification and control method based on probability prediction according to claim 1 is characterized in that: The construction of the wind power-hybrid energy storage two-stage stochastic control optimization operation model includes: 1) Phase I HESS classification: Phase I determines the HESS classification before any uncertainty occurs; in all scenarios, the rated capacity E, rated power P and initial energy level E of the HESS are determined. s ; 2) Second stage HESS control: The second stage optimizes the control action of HESS to smooth the fluctuations after the uncertainty of wind power realization; respectively determine the optimized grid-connected wind power Battery energy storage system optimizes charging and discharging power Optimized charging and discharging power of supercapacitors Obtain the optimized charging and discharging power control action of HESS, as well as the deviation between the scheduled power and the optimized charging and discharging power of the battery energy storage system Deviation between the preset power and the optimized charge and discharge power of the supercapacitor 7. The hybrid energy storage system classification and control method based on probability prediction according to claim 1 is characterized in that: In the wind power-hybrid energy storage two-stage stochastic control optimization operation model, the inherent energy storage cost of the HESS system is: In the formula, The daily cost per unit capacity of energy storage is is the capacity cost of lithium battery and supercapacitor; W n It represents the average number of operating days per year for a certain energy storage device n; is the daily amortized cost per unit power, is the energy storage power cost; T n Management information system energy storage equipment project life; r is the risk discount rate, and the set M represents the energy storage equipment type, including BESS and SC types; The control objective function of the wind power-hybrid energy storage two-stage stochastic control optimization operation model applied to the HESS system is as follows: In the formula, and are the daily amortized cost of energy storage capacity of BESS and SC respectively; and is the rated capacity and rated power of BESS; and is the rated capacity and rated power of SC; and are the daily amortized cost of energy storage power of BESS and SC respectively; and are the unit operating costs associated with the total energy storage of BESS and SC respectively; C pun The unit power penalty for the deviation between the optimal grid-connected wind power determined by TSSO and the target preset value; and is the charging / discharging power of BESS and SC at time t; is the target grid-connected wind power at time t, is the wind power-HESS system optimized grid-connected wind power at time t; the time interval Δt is 5 minutes; x is the wind power random variable, E[f s (x)] is the expected term, which is approximately f s The probability weighted average of typical wind power scenarios, f s (x) is the energy storage operation and maintenance cost, f M is the energy storage investment cost; T is the life of the energy storage device; Expected E[f s (x)] can be approximated as f s The probability weighted average of typical wind power scenarios is expressed as: Where N * is the number of samples required for the experiment; π n is the ratio of each sample n required for the experiment to the total number of samples, that is, π n =n n / N; The target power value set for a specific sample n at time t; is the actual power value of a specific sample n at time t; T is the life of the energy storage device.

8. The hybrid energy storage system classification and control method based on probability prediction according to claim 1 is characterized in that: The constraints of the wind power-hybrid energy storage two-stage random control optimization operation model include: power balance constraint, energy conversion constraint, HESS energy state constraint, HESS power constraint, optimization of HESS charging and discharging power constraint, and BESS life cycle constraint; The power balancing constraint is: Where: and are the power deviations of the pre-planned and optimized charging and discharging powers for BESS and SC, respectively; is the pre-power of BESS of sample i at time t; is the pre-power of SC of sample i at time t; is the wind power of sample i at time t, is the optimized wind power grid-connected power; The energy conversion constraint is: Where: is the energy level of device m at time t; m is a binary variable, λ m =1, the device is in charging state, otherwise, the energy storage device is in discharging state; η m is the charging power conversion rate of device m, if but is the optimized energy storage charging power, and vice versa, its value is the discharge power. is a continuous variable, used to replace and The set M represents the type of energy storage equipment, including BESS and SC types; The HESS energy state constraint is: Where: and Represent the minimum state of charge and maximum state of charge of energy storage respectively; and are the capacity and rated capacity of the equipment at time t, and They represent the equipment capacity at the beginning and end respectively; the set M represents the type of energy storage equipment, including BESS and SC types; The HESS power constraint is: Where: and are the power and rated power of equipment m at time t respectively; The optimized HESS charging and discharging power constraint is: The optimized HESS power control action cannot exceed equation (16); The BESS life cycle constraints are: Where: q t is the sign of the charge and discharge state change of BESS. If it is equal to 1, it changes from the discharge state to the charge state at time t; N0 is the cycle life of BESS; Q t The sum of is the number of BESS cycles in one day, is the BESS state quantity, W BESS is the average number of working days of BESS in a year, T BESS is the estimated life of BESS, are two 0-1 integer variables, and C is a large constant.

Images