A method for battery energy storage to smooth wind power fluctuations based on robust model predictive control

Abstract

This invention proposes a method for battery energy storage to participate in wind power fluctuation smoothing based on robust model predictive control. First, based on the maximum error that may exist in wind power output forecast, a wind power uncertainty set is constructed. With the goal of more comprehensively considering the uncertainty of wind power output, a wind power fluctuation smoothing model containing the wind power uncertainty set is established based on the robust optimization principle. According to the duality theory, the established min-max two-layer optimization problem is converted into a single-layer deterministic optimization problem. Secondly, in view of the problem that the charging and discharging margin may be insufficient when the BESS participates in smoothing wind power fluctuations, the state of charge of the energy storage battery is optimized to ensure that the BESS has sufficient margin to smooth wind power fluctuations. Finally, in order to address the problem that the accuracy of wind power forecast data decreases as the forecast time domain increases, based on the model predictive control principle, the BESS is subjected to rolling optimization control based on the ultra-short-term forecast data for the next 4 hours reported by the wind farm, and a wind power fluctuation smoothing strategy based on robust model predictive control is proposed.

Description

Technical Field

[0001] The present invention relates to the technical field of battery energy storage participating in wind power fluctuation smoothing, and in particular to a method for battery energy storage participating in wind power fluctuation smoothing based on robust model predictive control. Background Art

[0002] Power fluctuations associated with a high proportion of grid-connected wind power can easily cause active power imbalances in the power system, disrupt system frequency stability, and even trigger large-scale load shedding, seriously threatening the safe and economic operation of the power system. In recent years, the use of flexible charging / discharging battery energy storage systems (BESS) to smooth out wind power output fluctuations has attracted widespread attention. However, the uncertainty of wind power output and its prediction error will affect the effectiveness of BESS in smoothing wind power output fluctuations. Therefore, when formulating control strategies for BESS to smooth wind power output fluctuations, it is necessary to consider the possible prediction error of wind power output to improve the effectiveness of BESS in smoothing wind power output fluctuations. In addition, BESS needs to maintain a strong charging and discharging capacity at all times to be able to cope with and smooth out future wind power output fluctuations.

[0003] At present, the optimization control methods for battery energy storage systems to participate in smoothing wind power fluctuations mainly include control methods based on time-frequency analysis such as wavelet packet decomposition and empirical mode decomposition, and control methods based on model predictive control (MPC).

[0004] The control method based on time-frequency analysis mainly studies the historical output curve of wind power, lacks the effective use of future forecast information of wind farms, and the control instructions of its energy storage system are only calculated based on the wind power output information at the current moment. In actual projects, there is a certain degree of time lag.

[0005] MPC, a rolling optimization control algorithm that comprehensively considers system dynamics, control objectives, and constraints, can determine the current energy storage control strategy based on the predicted wind power output over a period of time. Traditional MPC uses a rolling time domain optimization approach, which provides a certain degree of robustness, but it is not designed to cope with system uncertainty. If there is a significant error between the predicted wind power and the actual output, relying solely on traditional MPC control methods is insufficient to cope with the strong uncertainty of wind power output.

[0006] Regarding the problem of smoothing wind power fluctuations considering prediction errors, the above two optimization control methods both have different drawbacks, which will cause traditional wind power fluctuation smoothing technologies to fail to achieve the expected results. Summary of the Invention

[0007] In view of the defects and shortcomings of the existing technology, the present invention proposes a method for battery energy storage to participate in wind power fluctuation smoothing based on robust model predictive control.

[0008] Its design points include:

[0009] (1) Aiming at the maximum possible error in wind power output prediction, a wind power uncertainty set is constructed. With the goal of more comprehensively considering the uncertainty of wind power output, a wind power fluctuation smoothing model containing the wind power uncertainty set is established based on the robust optimization principle. According to the duality theory, the established min-max two-level optimization problem is transformed into a single-level deterministic optimization problem, making the model easier to solve.

[0010] (2) In order to address the problem of insufficient charging and discharging margin when BESS participates in smoothing wind power fluctuations, the model constructed by this method includes the optimization of the charge state of the energy storage battery to ensure that the BESS has sufficient margin to smooth wind power fluctuations. In order to address the problem that the accuracy of wind power forecast data decreases as the forecast time domain increases, based on the principle of model predictive control, the BESS is controlled by rolling optimization based on the ultra-short-term forecast data of the next 4 hours reported by the wind farm, and a wind power fluctuation smoothing strategy based on robust model predictive control is proposed.

[0011] The present invention specifically adopts the following technical solutions:

[0012] A method for battery energy storage to participate in wind power fluctuation smoothing based on robust model predictive control, characterized by comprising the following steps:

[0013] Step S1: At the beginning of the optimization, based on the wind power forecast data reported by the wind farm, a wind power uncertainty set U is constructed. k+N-1|k ;

[0014] Step S2: Construct a wind power fluctuation smoothing model based on robust model predictive control, and transform the min-max two-level optimization problem of the model into a single-level deterministic optimization problem according to the dual transformation;

[0015] Step S3: Optimize and calculate at time k to obtain the optimal control sequence Z in the future prediction time domain N k ={P b (k+1),…,P b (k+i),…P b (k+N-1)};

[0016] Step S4: The optimal control sequence Z k The first value P b (k+1) acts on the wind-storage system to calculate and update the state variables SOC(k+1), P g (k+1) and fed back to the system to build the wind power fluctuation smoothing model at the next moment;

[0017] Step S5: If the optimization time has exceeded the control time, the optimization is terminated; otherwise, the time is rolled to the next time and the process returns to step S1.

[0018] Furthermore, the wind power fluctuation smoothing model based on robust model predictive control includes:

[0019] Wind power uncertainty set U k+N-1|k Expressed as:

[0020]

[0021] Where: P w (k) represents the actual output power of the wind farm at time k, represents the predicted value of wind power at time k, represents the maximum allowable error value of wind power prediction at time k, Г is the number of fluctuation periods in the prediction time domain, and takes an integer value in the range of [0, N]. Its value affects the robustness of the control strategy. N represents the length of the prediction time domain, and β(k) is the prediction error coefficient, which takes a value of {0, 1};

[0022] In each rolling optimization period, the objective function is:

[0023]

[0024] Where: k represents the starting time of the scroll, k = [1, 2, ... M], M is the total duration; P g (k+i) represents the grid-connected power of the wind-storage combined power generation system at time k+i, P b (k+i) represents the BESS output power at time k+i, SOC(k+i) represents the state of charge of the energy storage battery at time k+i, SOC b is the optimal state of charge of the energy storage battery, N is the prediction time domain length, a and b are weight coefficients;

[0025] Constraints include:

[0026] Grid-connected power constraints of wind-storage combined power generation system:

[0027] P g (k+1)=P b (k)+P w (k) (3)

[0028] |P g (k+i+1)-P g (k+i)|≤δ,i=0,1,...,N-1 (4)

[0029] Where: δ represents the grid-connected power fluctuation limit of the wind-storage system;

[0030] Battery energy storage system operation constraints:

[0031] The SOC of the battery energy storage system at the kth moment is related to the charge / discharge power in that period and the state of charge in the previous period:

[0032]

[0033] Where: T is the battery energy storage system control period, represents the charging power of the battery energy storage system at time k, represents the discharge power of the battery energy storage system at time k;

[0034] The operation of the battery energy storage system is constrained by the battery state of charge and the maximum charge and discharge power. Since the battery energy storage system can only be charged or discharged at the same time, a 0 / 1 variable is introduced to represent the charge and discharge status of the battery energy storage system:

[0035]

[0036] Where: SOC min , SOC max Respectively represent the upper and lower limits of the energy storage battery state of charge, They represent the upper and lower limits of the battery energy storage system output respectively. It is a 0 / 1 variable that represents the charge and discharge status of the battery energy storage system. Indicates charging. Indicates discharge.

[0037] Furthermore, the solution of the wind power fluctuation smoothing model based on robust model predictive control includes the following process:

[0038] The wind power fluctuation smoothing model based on robust model predictive control is described as follows:

[0039]

[0040] Where: X k+N-1|k Indicates the decision variables for the next k+N-1 control periods at time k; U k+N-1|k It represents the wind power uncertainty set constructed at time k for the next k+N-1 control periods; C represents a positive definite matrix, A, B, and I are constant coefficient matrices; the constraint AX k+N-1|k ≤d represents the inequality constraint of the deterministic variable; the constraint BX k+N-1|k =e represents the equality constraint of deterministic variables; constraint formula IX k+N-1|k =U k+N-1|k Represents equality constraints involving uncertain variables;

[0041] The above model is a min-max optimization model with uncertain variables, which makes it difficult to solve the optimization problem using a solver. To solve this model, the original min-max optimization problem is transformed into a deterministic optimization problem based on the strong duality theory. The dual problem of the original problem is expressed as follows:

[0042]

[0043] Where: λ1, λ2 and λ3 are the dual variables of the corresponding constraints;

[0044] The uncertainty of wind power is aggregated into U k+N-1|k Substituting into equation (8), and using the Big-M method (processing the bilinear terms of the multiplication of state variables and continuous variables), we obtain:

[0045]

[0046] Where: and is the predicted value and maximum prediction error of wind power; α + and α - is a continuous variable used to replace the bilinear term, where α + =β + λ3 T , α - =β + λ3 T ; M is a positive number greater than a certain threshold;

[0047] At this time, only the first component P of the optimal control sequence is taken b (k+1) is sent to the battery energy storage system. As time goes by, the prediction time domain also rolls forward. Based on the updated system information and prediction data, the above process is repeated to achieve rolling robust optimization.

[0048] The present invention and its preferred embodiment enhance the ability of a grid-connected wind-storage system to cope with wind power output uncertainty, effectively enabling the BESS to smooth wind power fluctuations within a range of wind power forecast error. Based on the grid-connected power fluctuation limits of the wind-storage system and the actual operating constraints of the BESS, this system not only smooths wind power fluctuations but also reduces the total charge and discharge energy of the BESS to protect the energy storage battery, taking into account the BESS output capacity at each moment. This gives the BESS a strong ability to cope with future wind power output uncertainty. Based on the ultra-short-term forecast data reported by wind farms on a rolling basis for the next four hours, a wind power uncertainty set is constructed, taking into account the maximum possible forecast error of the wind farm. Compared to traditional deterministic smoothing methods that do not consider wind power forecast error, the wind power fluctuation smoothing effect of this method becomes more significant as the wind power forecast error increases, thereby enhancing the ability of the grid-connected wind-storage system to adapt to uncertain environments. Furthermore, the accuracy of wind power forecast data increases as the forecast horizon shortens. This method performs rolling optimization based on the ultra-short-term forecast data reported by wind farms, thereby improving the effectiveness of the wind power smoothing strategy. Finally, the robust model predictive control strategy proposed in this method can enable the wind-storage system to achieve a stabilization effect under its worst conditions. Therefore, the power dispatching department can flexibly configure a BESS of appropriate capacity for the wind farm and reasonably control and dispatch it according to the requirements of different prediction error levels. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0050] Figure 1 This is a diagram of a wind power fluctuation smoothing strategy based on robust model predictive control according to an embodiment of the present invention.

[0051] Figure 2 Schematic diagram of the control flow of an embodiment of the present invention. DETAILED DESCRIPTION

[0052] Hereinafter, specific embodiments of the present application will be described in detail with reference to the accompanying drawings. Based on these detailed descriptions, those skilled in the art will be able to clearly understand the present application and implement the present application. Without violating the principles of the present application, the features of different embodiments may be combined to obtain new implementations, or certain features of certain embodiments may be substituted to obtain other preferred implementations.

[0053] To make the features and advantages of the present invention more clearly understood, the following embodiments are specifically described in detail with reference to the accompanying drawings.

[0054] This embodiment proposes a method for battery energy storage to participate in wind power fluctuation smoothing based on robust model predictive control. First, based on the maximum possible error in wind power output forecast, a wind power uncertainty set is constructed. With the goal of more comprehensively considering the uncertainty of wind power output, a wind power fluctuation smoothing model containing the wind power uncertainty set is established based on the robust optimization principle. According to the duality theory, the established min-max two-layer optimization problem is transformed into a single-layer deterministic optimization problem, making the model easier to solve. Secondly, in view of the problem that the charging and discharging margin may be insufficient when the BESS participates in smoothing wind power fluctuations, the state of charge of the energy storage battery is optimized to ensure that the BESS has sufficient margin to smooth wind power fluctuations. Finally, in order to address the problem that the accuracy of wind power forecast data decreases as the forecast time domain increases, based on the model predictive control principle, the BESS is subjected to rolling optimization control based on the ultra-short-term forecast data for the next 4 hours reported by the wind farm, and a wind power fluctuation smoothing strategy based on robust model predictive control is proposed.

[0055] It specifically includes the following designs:

[0056] 1. Wind power fluctuation smoothing model based on robust model predictive control

[0057] (1) Wind power uncertainty set

[0058] The traditional wind power output fluctuation smoothing strategy only relies on the point prediction data of wind power to optimize the control of the energy storage system. However, the strategy proposed in this embodiment takes into account the uncertainty of wind power output. According to the possible error range of the wind power prediction value, the wind power output is calculated as the uncertainty set U. k+N-1|k express:

[0059]

[0060] Where: P w (k) represents the actual output power of the wind farm at time k, represents the predicted value of wind power at time k, represents the maximum allowable error value of wind power prediction at time k, Г is the number of fluctuation periods in the prediction domain, and takes an integer value in the range of [0, N]. Its size affects the robustness of the control strategy. N represents the length of the prediction domain, and β(k) is the prediction error coefficient, which takes a value of {0, 1}.

[0061] (2) Objective function

[0062] Robust optimization minimizes the objective function under the worst-case scenario of uncertainty parameters. Therefore, this embodiment can still smooth wind power fluctuations even in the worst-case scenario of the wind-storage combined power generation system, and can also meet the grid's requirements for wind power grid-connected power in other operating scenarios.

[0063] This embodiment reduces the total energy output of energy storage from the perspective of protecting energy storage batteries, while taking into account the output capacity of each energy storage to smooth the wind power. In each rolling optimization period, the objective function of the wind power fluctuation smoothing model based on robust MPC is:

[0064]

[0065] Where: k represents the starting time of rolling, k=[1,2,…M], M is the total simulation time. g (k+i) represents the grid-connected power of the wind-storage combined power generation system at time k+i, P b (k+i) represents the BESS output power at time k+i, SOC(k+i) represents the state of charge of the energy storage battery at time k+i, SOC b is the optimal state of charge of the energy storage battery, which is 0.5 in this embodiment, N is the prediction time domain length, and a and b are weight coefficients.

[0066] (3) Constraints

[0067] ① Grid-connected power constraints of wind-storage combined power generation system:

[0068] P g (k+1)=P b (k)+P w (k) (3)

[0069] |P g (k+i+1)-P g (k+i)|≤δ,i=0,1,...,N-1 (4)where: δ represents the grid-connected power fluctuation limit of the wind-storage system.

[0070] ②Battery energy storage system operation constraints:

[0071] The SOC of the battery energy storage system at the kth moment is related to the charge / discharge power in that period and the state of charge in the previous period.

[0072]

[0073] Where: T is the battery energy storage system control period, represents the charging power of the battery energy storage system at time k, represents the discharge power of the battery energy storage system at time k.

[0074] The operation of a battery energy storage system is constrained by the battery's state of charge and maximum charge and discharge power. Since a battery energy storage system can only be charging or discharging at the same time, a 0 / 1 variable is introduced to represent the battery energy storage system's charge and discharge status.

[0075]

[0076] Where: SOC min , SOC max Respectively represent the upper and lower limits of the energy storage battery state of charge, They represent the upper and lower limits of the battery energy storage system output respectively. It is a 0 / 1 variable that represents the charge and discharge status of the battery energy storage system. Indicates charging. Indicates discharge.

[0077] 2. Model solution

[0078] The power fluctuation smoothing model of wind-storage power generation system based on robust model predictive control is described as follows:

[0079]

[0080] Where: X k+N-1|k Indicates the decision variables for the next k+N-1 control periods at time k; U k+N-1|k It represents the wind power uncertainty set constructed at time k for the next k+N-1 control periods; C represents a positive definite matrix, and A, B, and I are constant coefficient matrices. Constraint AX k+N-1|k ≤d represents the inequality constraint of the deterministic variable; the constraint BX k+N-1|k =e represents the equality constraint of deterministic variables; constraint formula IX k+N-1|k =U k+N-1|k Represents an equality constraint involving uncertain variables.

[0081] The above model is a min-max optimization model with uncertain variables, which makes it difficult to solve the optimization problem using a solver. To solve this model, this embodiment transforms the original min-max optimization problem into a deterministic optimization problem based on the strong duality theory. The dual problem of the original problem can be expressed as follows:

[0082]

[0083] Where: λ1, λ2 and λ3 are the dual variables of the corresponding constraints.

[0084] Furthermore, the wind power uncertainty set U k+N-1|k Substituting into equation (8) and using the Big-M method (i.e., the Big-M method is a method of finding the initial basis feasible solution after using the artificial variable method when the constraints of the linear programming problem are (=) equality or (≥) greater than type)) to process the bilinear term of the multiplication of the state variable and the continuous variable, we can get:

[0085]

[0086] Where: and is the predicted value and maximum prediction error of wind power; α + and α - is a continuous variable used to replace the bilinear term, where α + =β + λ3 T , α - =β + λ3 T ; M is a sufficiently large positive number.

[0087] At this time, only the first component P of the optimal control sequence is taken b (k+1) is sent to the battery energy storage system. As time goes by, the prediction time domain also rolls forward. Based on the updated system information and prediction data, the above process is repeated to achieve rolling robust optimization.

[0088] 3. Method flow

[0089] like Figure 1 、 Figure 2 As shown, based on the above design, the specific steps of the method for battery energy storage to smooth wind power fluctuations based on robust model predictive control proposed in this embodiment are as follows:

[0090] Step S1: At the beginning of the optimization, based on the wind power forecast data reported by the wind farm, a wind power uncertainty set U is constructed. k+N-1|k ;

[0091] Step S2: Construct a wind power fluctuation smoothing model based on robust model predictive control, and transform the min-max two-level optimization problem of the model into a single-level deterministic optimization problem according to the dual transformation;

[0092] Step S3: Optimize and calculate at time k to obtain the optimal control sequence Z in the future prediction time domain N k ={P b (k+1),…,P b (k+i),…P b (k+N-1)};

[0093] Step S4: The optimal control sequence Z k The first value P b (k+1) acts on the wind-storage system to calculate and update the state variables SOC(k+1), P g (k+1) and fed back to the system to build the wind power fluctuation smoothing model at the next moment;

[0094] Step S5: If the optimization time has exceeded the control time, the optimization is terminated; otherwise, the time is rolled to the next time and the process returns to step S1.

[0095] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

[0096] The present invention is not limited to the above-mentioned optimal implementation mode. Anyone can derive various other forms of battery energy storage participating in wind power fluctuation smoothing based on robust model predictive control under the inspiration of the present invention. All equal changes and modifications made according to the scope of the patent application of the present invention should fall within the scope of the present invention.

Claims

1. A method for battery energy storage to participate in wind power fluctuation smoothing based on robust model predictive control, characterized in that: The following steps are involved: Step S1: At the beginning of the optimization, based on the wind power forecast data reported by the wind farm, a wind power uncertainty set U is constructed. k+N-1|k ; Step S2: Construct a wind power fluctuation smoothing model based on robust model predictive control, and transform the min-max two-level optimization problem of the model into a single-level deterministic optimization problem according to the dual transformation; Step S3: Optimize and calculate at time k to obtain the optimal control sequence Z in the future prediction time domain N k ={P b (k+1),…,P b (k+i),…P b (k+N-1)}; Step S4: The optimal control sequence Z k The first value P b (k+1) acts on the wind-storage system to calculate and update the state variables SOC(k+1), P g (k+1) and fed back to the system to build the wind power fluctuation smoothing model at the next moment; Step S5: If the optimization time has exceeded the control time, the optimization is terminated; otherwise, the time is rolled to the next time and the process returns to step S1; The wind power fluctuation smoothing model based on robust model predictive control includes: Wind power uncertainty set U k+N-1|k Expressed as: Where: P w (k) represents the actual output power of the wind farm at time k, represents the predicted value of wind power at time k, represents the maximum allowable error value of wind power prediction at time k, Г is the number of fluctuation periods in the prediction time domain, and takes an integer value in the range of [0, N]. Its value affects the robustness of the control strategy. N represents the length of the prediction time domain, and β(k) is the prediction error coefficient, which takes a value of {0, 1}; In each rolling optimization period, the objective function is: Where: k represents the starting time of the scroll, k = [1, 2, ... M], M is the total duration; P g (k+i) represents the grid-connected power of the wind-storage combined power generation system at time k+i, P b (k+i) represents the BESS output power at time k+i, SOC(k+i) represents the state of charge of the energy storage battery at time k+i, SOC b is the optimal state of charge of the energy storage battery, N is the prediction time domain length, a and b are weight coefficients; Constraints include: Grid-connected power constraints of wind-storage combined power generation system: P g (k+1)=P b (k)+P w (k) (3) |P g (k+i+1)-P g (k+i)|≤δ,i=0,1,...,N-1 (4)where: δ represents the grid-connected power fluctuation limit of the wind-storage system; Battery energy storage system operation constraints: The SOC of the battery energy storage system at the kth moment is related to the charge / discharge power in that period and the state of charge in the previous period: Where: T is the battery energy storage system control period, represents the charging power of the battery energy storage system at time k, represents the discharge power of the battery energy storage system at time k; The operation of the battery energy storage system is constrained by the battery state of charge and the maximum charge and discharge power. 0 / 1 variables are introduced to represent the charge and discharge status of the battery energy storage system: Where: SOC min , SOC max Respectively represent the upper and lower limits of the energy storage battery state of charge, They represent the upper and lower limits of the battery energy storage system output respectively. It is a 0 / 1 variable that represents the charge and discharge status of the battery energy storage system. Indicates charging. Indicates discharge.

2. The method for battery energy storage participating in wind power fluctuation smoothing based on robust model predictive control according to claim 1 is characterized by: The solution of the wind power fluctuation smoothing model based on robust model predictive control includes the following process: The wind power fluctuation smoothing model based on robust model predictive control is described as follows: Where: X k+N-1|k Indicates the decision variables for the next k+N-1 control periods at time k; U k+N-1|k It represents the wind power uncertainty set constructed at time k for the next k+N-1 control periods; C represents a positive definite matrix, A, B, and I are constant coefficient matrices; the constraint AX k+N-1|k ≤d represents the inequality constraint of the deterministic variable; the constraint BX k+N-1|k =e represents the equality constraint of deterministic variables; constraint formula IX k+N-1|k =U k+N-1|k Represents equality constraints involving uncertain variables; The above model is a min-max optimization model with uncertain variables, which makes it difficult to solve the optimization problem using a solver. To solve this model, the original min-max optimization problem is transformed into a deterministic optimization problem based on the strong duality theory. The dual problem of the original problem is expressed as follows: Where: λ1, λ2 and λ3 are the dual variables of the corresponding constraints; The uncertainty of wind power is aggregated into U k+N-1|k Substituting into equation (8) and using the Big-M method to process the bilinear term of the multiplication of the state variable and the continuous variable, we obtain: Where: and is the predicted value and maximum prediction error of wind power; α + and α - is a continuous variable used to replace the bilinear term, where α + =β + λ3 T , α - =β + λ3 T ; M is a positive number greater than a certain threshold; At this time, only the first component P of the optimal control sequence is taken b (k+1) is sent to the battery energy storage system. As time goes by, the prediction time domain also rolls forward. Based on the updated system information and prediction data, the above process is repeated to achieve rolling robust optimization.

Images