Wind power plant active power distribution system with reliability evaluation and hierarchical MPC combined

By conducting reliability evaluation and stratified MPC control in the wind farm, dividing fan clusters and optimizing active power distribution, the calculation efficiency and stability of the MPC algorithm under the number of large-scale fans in the wind farm are solved, and efficient and effective wind farms are achieved Power distribution and stable output.

CN118920613APending Publication Date: 2024-11-08YUNNAN POWER INVESTMENT LVNENG TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411053524.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-01
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

With the expansion of wind farm size and the increase in the number of fans, the existing MPC algorithms face the problem of too long search time and local extreme value traps in the optimization process, and it is difficult to complete active power allocation within an acceptable prediction time.

Method used

The reliability evaluation joint layered MPC system is adopted to measure and evaluate fan operation and environmental data, divide the fan cluster, and the central controller and local controller respectively process the active power target and alternative value sequence, and use the MPC algorithm to perform stability Optimized to distribute active power.

Benefits of technology

It effectively reduces the computing burden of the central controller, breaks down large-scale constraint problems, avoids frequent fan replacement, and ensures stable output and overall reliability of active power.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118920613A_ABST
    Figure CN118920613A_ABST
Patent Text Reader

Abstract

The invention provides a reliability evaluation and hierarchical MPC combined wind power plant active power distribution system. The system comprises a measurement unit, an evaluation unit, a central controller and a plurality of local controllers. The measuring unit is used for measuring operation and environment data of each fan in the wind power plant; the evaluation unit evaluates the reliability of each fan based on the operation and environment data and divides each fan into a plurality of fan clusters according to an evaluation result; each local controller is generated by the central controller and controls each fan cluster in a one-to-one correspondence manner; and the central controller and each local controller distribute an active power output instruction to each fan in the wind power plant through a hierarchical MPC distribution strategy. According to the distribution system provided by the invention, the fans are divided into different fan clusters based on the reliability of the running state of each fan, and a layered MPC control architecture and stability optimization are adopted, so that the instruction optimization data volume can be remarkably reduced, and the stable running of the wind power plant is ensured.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application belongs to the technical field of wind power generation control, and in particular, relates to a wind farm active power distribution system with reliability assessment and hierarchical MPC. Background Art

[0002] With the increasing demand for renewable energy, wind power, as an important component of clean energy, has received widespread attention and promotion. In recent years, the development trend of wind farms has shown a rapid growth trend. With the continuous advancement of technology and the continuous development of wind power demand, the number of wind turbines that wind farms can accommodate is also showing a growing trend.

[0003] Since a wind farm may be composed of wind turbines of various specifications and types, the differences in their power factor, operating reliability and wind turbine location will affect their active power output, and ultimately affect the ability and quality of the wind farm to participate in grid peak regulation. Therefore, after receiving the grid peak regulation energy task (generally issued in the form of grid output instructions, used to specify the active power that the wind farm needs to output during a specified period of time), the wind farm must reasonably allocate it to each wind turbine according to the characteristics of each wind turbine in the wind farm. Among them, the Model Predictive Control (MPC) algorithm is a relatively effective multi-wind turbine output control strategy. Others combine dynamic system modeling, rolling optimization theory and feedback correction technology to optimize the control of the system within a limited prediction time domain. The main idea is to predict the future behavior of the system in each control cycle and calculate the optimal control input based on these prediction results to achieve system stability, performance and constraint conditions.

[0004] However, although the conventional MPC algorithm can well control a limited number of wind turbines in a small-scale wind farm, as the scale of wind farms expands and the number and types of wind turbines increase, the dimension of the objective function that the MPC algorithm uses to optimize will show an explosive growth, which will undoubtedly lead to excessively long search times and greatly increase the possibility of search results falling into local extreme value traps, and even lead to the problem of being unable to complete the optimization within an acceptable prediction time. Summary of the invention

[0005] In order to solve the problems existing in the above-mentioned prior art, an embodiment of the present application provides a wind farm active power distribution system of reliability evaluation combined with hierarchical MPC, the distribution system comprising: a measurement unit, used to measure the operation and environmental data of each wind turbine in the wind farm; an evaluation unit, based on the operation and environmental data, to evaluate the reliability of each wind turbine in the wind farm, and to divide each wind turbine into a plurality of wind turbine clusters according to the reliability evaluation results; a central controller and a plurality of local controllers, wherein the plurality of local controllers are generated by the central controller and control the plurality of wind turbine clusters one by one. group; the central controller determines the active power target value of each wind turbine cluster through the MPC algorithm based on the reliability evaluation result and the cost function, and performs stability optimization on the active power candidate value sequence of each wind turbine cluster to determine the active power allocation value of each wind turbine cluster; each of the local controllers receives the active power target value and the local objective function from the central controller, generates the active power candidate value sequence of the wind turbine cluster it controls through the MPC algorithm, and allocates active power output instructions to each wind turbine in the wind turbine cluster it controls based on the active power allocation value received from the central controller.

[0006] Furthermore, the operation and environmental data of each wind turbine in the wind farm include active power sampling values ​​and wind speed sampling values ​​of each wind turbine in the wind farm at each sampling moment; and the evaluation unit evaluates the reliability of each wind turbine in the wind farm by obtaining the average utilization vector and reliability matrix of each wind turbine in the wind farm, wherein the average utilization vector and reliability matrix are determined based on the relationship between the ideal output model of each wind turbine and the operation status probability model.

[0007] Specifically, the ideal output model is:

[0008]

[0009] Where v is the real-time wind speed, P(v) is the ideal active power value of the wind turbine when the wind speed is v, and v in 、v out are the cut-in wind speed and cut-out wind speed of the fan, respectively, v rate , P rate are the wind speed rating and active power rating of the wind turbine respectively;

[0010] The operating condition probability model is:

[0011]

[0012] Among them, P status,P(v) , P status,0 , P status,αP(v) are the probabilities of the fan operating normally, shutting down, and operating at a reduced rating according to the derating factor α, τ1 , τ 2 , τ 3 are the duration of normal operation, shutdown and derating operation of the fan, λ, λ d are the transfer rates of the fan from normal operation to shutdown and derating operation, μ, μ d are the recovery rates of the fan from shutdown and derating operation to normal operation, U 1 , U 2 , U 3 are uniformly distributed random numbers between [0, 1].

[0013] Furthermore, the evaluation unit obtains the average utilization vector and reliability matrix of each wind turbine in the wind farm through the following steps:

[0014] S21, obtaining operation and environmental data of each wind turbine in the wind farm;

[0015] S22, select any fan;

[0016] S23, pre-processing the operation and environmental data of the wind turbine, eliminating wind speeds below the cut-in wind speed v in Or greater than its cut-out wind speed v out The wind speed sampling value and the corresponding active power sampling value;

[0017] S24, determine the expected active power value P of the wind turbine at each sampling time k based on the following formula: E (k):

[0018]

[0019] Among them, v rate , P rate are the wind speed rating and active power rating of the wind turbine, respectively, v S (k) is the wind speed sampling value of the wind turbine at sampling time k, K is the total number of sampling times after preprocessing, and the cumulative time of K sampling times is the total time for reliability assessment of the wind turbine;

[0020] S25, determining the operating state n(k) of the fan at each sampling time k based on the following formula:

[0021]

[0022] Among them, P S (k) is the active power sampling value of the wind turbine at each sampling time k, round[] is used to find the nearest integer of the value in the brackets, and N+1 is the total number of possible operating states of the wind turbine;

[0023] S26, determining the average utilization vector AUR of the wind turbine based on the following formula:

[0024]

[0025] S27, determine the reliability matrix R of the wind turbine based on the following formula:

[0026]

[0027] Among them, the elements in R The operating state of the fan at two adjacent sampling moments is n 1 Transfer to n 2 The probability of

[0028] S28, replacing the wind turbine and returning to step S23 until all wind turbines in the wind farm are traversed.

[0029] Furthermore, the central controller determines the active power target value of each wind turbine cluster through the MPC algorithm based on the cost function minimization objective shown in the following formula:

[0030]

[0031] Where Φ is the cost function, I is the total number of wind turbine clusters, T is the total length of the forecast period, is the active power target value of the i-th wind turbine cluster at the prediction time t, P WT,i is the active power reference value of the i-th wind turbine cluster at the prediction time t, are the shaft torque variation and thrust variation of the i-th wind turbine cluster at the prediction time t, Q p , Q T , Q F is the weight of each square term in the formula.

[0032] Preferably, P WT,i It is determined by the central controller according to the reliability evaluation result.

[0033] Preferably, the constraint conditions when the central controller searches for the active power target value of each wind turbine cluster are specifically:

[0034]

[0035] Among them, P WG The peak load regulation command received by the wind farm. is the initial value of active power of the i-th wind turbine cluster in the forecast period, is the active power increment of the i-th wind turbine cluster during the forecast period, is the active power rating of the i-th wind turbine cluster.

[0036] Preferably, each local controller generates a sequence of candidate active power values ​​of the wind turbine cluster controlled by it through the following steps:

[0037] The central controller decomposes the cost function Φ into the local objective function Φ of each wind turbine cluster as shown in the following formula: i (u i ) is in the form of:

[0038]

[0039] Among them, u i is the candidate value of active power of the i-th wind turbine cluster;

[0040] The central controller sends the local objective function and active power target value of each wind turbine cluster to the corresponding local controller;

[0041] Each local controller receives active power target value based on its received minΦ i (u i ) as the target, the MPC algorithm is used to generate the active power candidate value sequence corresponding to three consecutive prediction times t-1, t, and t+1

[0042] Furthermore, the central controller performs stability optimization on the active power candidate value sequence of each wind turbine cluster based on the following formula to determine the active power allocation value of each wind turbine cluster:

[0043]

[0044] in, is the active power distribution value of the i-th wind turbine cluster at the prediction time t, A, B, and C are The stability optimization coefficient.

[0045] Preferably, the stability optimization coefficient is determined by searching for the minimum value of the stability optimization function f(A, B, C) using a gradient descent method, and the stability optimization function is specifically:

[0046]

[0047] The reliability assessment combined with hierarchical MPC wind farm active power distribution system provided in the embodiment of the present application evaluates the operational reliability characteristics of each motor in the wind farm based on its historical operational data and corresponding environmental data, and clusters different wind turbines into several wind turbine clusters according to their reliabilities. Since the operational reliability of a wind turbine can better reflect the degree of trustworthiness of the wind turbine in participating in the peak-shaving process of the power grid, this can be used as a basis for clustering, and in the process of allocating power grid output instructions, an objective function is adaptively constructed to tilt the output distribution toward clusters with higher overall reliability. Secondly, after all wind turbines in the wind farm are clustered into several wind turbine clusters, the active power target value of each wind turbine cluster is preliminarily determined by the central controller based on the overall objective function, and the local controllers corresponding to each wind turbine cluster obtain a sequence of active power alternative values ​​at multiple prediction moments through the disassembled local objective function. Part of the computing tasks are assigned to local controllers, which can decompose large-scale constraint problems. For the central controller, its computing task is mainly to solve the objective function with a constraint dimension equivalent to the number of wind turbine clusters. When the wind farm consists of hundreds of wind turbines, the computing burden of the central controller can be significantly reduced. At the same time, for the local controller, the computing task involves solving an optimization problem with smaller-scale constraints. Therefore, the amount of calculation for each wind turbine controller will not be large. The above-mentioned hierarchical MPC mechanism can effectively solve the challenge of a large state space model with large-scale constraints when there are a large number of wind turbines in a wind farm. In addition, the multiple active power alternative values ​​generated by each controller at adjacent moments are further optimized for stability by the central controller. The proportional coefficients are allocated by multiple iterations and the optimal proportional value is searched to avoid the problem of frequent changes in working modes of wind turbines due to frequent changes in instructions and inability to output stably. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A schematic diagram of the architecture of a wind farm active power distribution system of reliability assessment combined with hierarchical MPC provided according to an embodiment of the present application;

[0049] Figure 2 A schematic diagram of a process for evaluating the reliability of each wind turbine in a wind farm by an evaluation unit provided in an embodiment of the present application;

[0050] Figure 3 A schematic diagram of a process of allocating active power output instructions of each wind turbine by a central controller and a local controller through a hierarchical MPC algorithm according to an embodiment of the present application;

[0051] Figure 4 A schematic diagram of the principle of stability optimization based on a sequence of candidate active power values ​​provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0052] Hereinafter, the present application will be further described based on preferred embodiments with reference to the accompanying drawings.

[0053] In the description of the embodiments of the present application, it should be noted that if the terms "upper", "lower", "inside", "outside", etc. indicate an orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the products of the embodiments of the present application are usually placed when in use, it is only for the convenience of describing the present application and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present application. In addition, in the description of the present application, in order to distinguish different units, the words first, second, etc. are used in this specification, but these are not limited by the order of manufacture, nor can they be understood as indicating or implying relative importance, and their names may be different in the detailed description and claims of the present application.

[0054] The vocabulary in this specification is used to illustrate the embodiments of the present application, but is not intended to limit the present application. It should also be noted that, unless otherwise clearly specified and limited, the terms "disposed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, a direct connection, an indirect connection through an intermediate medium, or a connection between the two elements. For those skilled in the art, the specific meanings of the above terms in this application can be specifically understood.

[0055] An embodiment of the present application provides a reliability assessment combined with hierarchical MPC wind farm active power distribution system, which is used to reasonably distribute active power output instructions to multiple wind turbines included in the wind farm after receiving the output instruction issued by the power grid.

[0056] Figure 1 FIG. 1 is a schematic diagram of the architecture of the distribution system provided according to some embodiments, such as Figure 1 As shown, the distribution system includes a measuring unit, an evaluation unit, a central controller, and I local controllers generated by the central controller for one-to-one control of I wind turbine clusters described later.

[0057] Among them, the measurement unit is used to measure the operation and environmental data of each wind turbine in the wind farm. In some preferred embodiments, the operation and environmental data of each wind turbine include the active power sampling value and wind speed sampling value of each wind turbine at each sampling time. By setting a voltage and current measuring device at the output end of each wind turbine, and setting a wind force, wind direction measuring device and other sensor equipment at the location of the wind turbine, the active power and wind speed data of each wind turbine can be continuously obtained at each sampling time with a certain sampling interval.

[0058] The operation and environmental data measured by the above-mentioned measurement unit, as the data of the wind turbines in the wind farm, can reflect the reliability of each wind turbine during operation. Furthermore, the evaluation unit evaluates the reliability of each wind turbine in the wind farm based on the operation and environmental data, and divides each wind turbine into I wind turbine clusters according to the reliability evaluation results.

[0059] The number of wind turbine clusters I can be determined based on the total number of wind turbines in the wind farm and the reliability evaluation results described later. For example, in some specific embodiments, the value of I can be 2, 3, 4 or more. Accordingly, the scale of each wind turbine cluster can be set according to certain rules. For example, the number of wind turbines in each wind turbine cluster can be kept within a certain range. For example, the upper limit of the number of wind turbines in each wind turbine cluster is not more than 20, and the lower limit is not less than 5. For example, the scale of each wind turbine cluster can be controlled by limiting the sum of the rated active power of the wind turbines in each wind turbine cluster to be kept within a certain range.

[0060] As described above, after the evaluation unit divides the wind turbines in the wind farm into I wind turbine clusters, a central controller generates I local controllers according to the division results of the wind turbine clusters, so that it can control the I wind turbine clusters one by one.

[0061] In the embodiment of the present application, for the above-mentioned central controller + multiple local controllers architecture, a hierarchical MPC allocation mechanism is used to achieve reasonable allocation of active power output instructions of each wind turbine in the wind farm, and its specific operations are as follows:

[0062] 1) Based on the reliability evaluation results and cost function of the evaluation unit, the central controller determines the active power target value of each wind turbine cluster through the MPC algorithm, and performs stability optimization on the active power candidate value sequence of each wind turbine cluster to determine the active power allocation value of each wind turbine cluster;

[0063] 2) Each local controller receives the active power target value and the local objective function from the central controller, generates a sequence of active power candidate values ​​for the wind turbine cluster it controls through the MPC algorithm, and allocates active power output instructions to each wind turbine in the wind turbine cluster it controls based on the active power allocation value received from the central controller.

[0064] Compared with the existing wind farm control scheme based on model predictive control (MPC), the wind farm active power distribution system provided in this application has the following advantages:

[0065] First, the technical solution of the present application evaluates the operational reliability characteristics of each motor in the wind farm based on its historical operational data and corresponding environmental data, and clusters different wind turbines into several wind turbine clusters according to their reliability. Since the operational reliability of wind turbines can better reflect the degree of trustworthiness of wind turbines in participating in the peak load regulation of the power grid, this can be used as the basis for clustering, and in the process of allocating power grid output instructions, the objective function is adaptively constructed to tilt the output allocation toward clusters with higher overall reliability.

[0066] Secondly, after clustering all wind turbines in the wind farm into several wind turbine clusters, the active power target value of each wind turbine cluster is preliminarily determined by the central controller based on the overall objective function, and the local controllers corresponding to each wind turbine cluster obtain the active power candidate value sequence at multiple prediction moments through the disassembled local objective function. By allocating part of the computing tasks to the local controllers, the large-scale constraint problem can be decomposed. For the central controller, its computing task is mainly to solve the objective function whose constraint dimension is equivalent to the number of wind turbine clusters. When the wind farm consists of hundreds of wind turbines, the computing burden of the central controller can be significantly reduced. At the same time, for the local controller, the computing task involves solving an optimization problem with smaller-scale constraints, so the computing amount of each wind turbine controller will not be very large. Through the above-mentioned hierarchical MPC mechanism, the challenge of a large state space model with large-scale constraints when there are a large number of wind turbines in the wind farm can be effectively solved;

[0067] Finally, due to the frequent changes in instructions during actual operation of the wind turbine, which will cause the wind turbine to frequently change its working mode, this is not conducive to the stable output of active power. Therefore, in the technical solution of the present application, the multiple active power alternative values ​​​​generated by each controller at adjacent moments are further optimized for smoothness through the central controller, and the proportional coefficient is allocated by multiple iterations and the optimal proportional value is searched to ensure that the public power distribution value finally allocated to each wind turbine cluster has better consistency.

[0068] The following is a detailed description of the specific steps of the reliability evaluation of the evaluation unit and the hierarchical MPC control mechanism of the central controller*local controller in conjunction with the accompanying drawings.

[0069] [Reliability assessment and wind turbine clustering]

[0070] In an embodiment of the present application, the evaluation unit evaluates the reliability of each wind turbine in the wind farm by acquiring an average utilization vector and a reliability matrix of each wind turbine in the wind farm.

[0071] Among them, the average utilization vector and the reliability matrix are selected based on the analysis of the relationship between the ideal output model of each wind turbine and the operating condition probability model, and are used to characterize the reliability of each wind turbine from different perspectives.

[0072] The following is an explanation of the relationship between the ideal output model and the operating condition probability model:

[0073] <Ideal output model of fan>

[0074] Wind speed and the operating state of the wind turbine are the main factors affecting the output of a wind farm. The output of the wind turbine in an ideal working condition can be approximated according to the principle of aerodynamics. Specifically, in the embodiment of the present application, an ideal output model as shown in formula (1) can be established:

[0075]

[0076] Where v is the real-time wind speed, P(v) is the ideal active power value of the wind turbine when the wind speed is v, and v in 、v out are the cut-in wind speed and cut-out wind speed of the fan, respectively, v rate , P rate They are the wind speed rating and active power rating of the fan respectively.

[0077] (1) indicates that when the wind speed is lower than the cut-in wind speed v in The fan cannot start when the wind speed exceeds the cut-out wind speed v out For safety reasons, the fan output needs to be stopped; in , v rate ] interval, the output of the fan in the ideal operating state increases with the wind speed until the wind speed enters [v rate , v out ] interval, the fan always runs at P rate Output active power.

[0078] <Operational status probability model>

[0079] It should be noted that the wind turbine output of 0 in formula (1) does not indicate that it is in an abnormal operating state (such as a shutdown state due to a fault). In fact, the output of a wind turbine in a normal operating state (i.e., an ideal operating state) is independent of the operating state it may be in. In the embodiment of the present application, the operating state of the wind turbine is expressed by formula (2) in the operating condition probability model:

[0080]

[0081] Among them, P status,P(v) , P status,0, P status,αP(v) are the probabilities of the fan operating normally, shutting down, and operating at a reduced rating according to the derating factor α, τ 1 , τ 2 , τ 3 are the duration of normal operation, shutdown and derating operation of the fan, λ, λ d are the transfer rates of the fan from normal operation to shutdown and derating operation, μ, μ d are the recovery rates of the fan from shutdown and derating operation to normal operation, U 1 , U 2 , U 3 are uniformly distributed random numbers between [0, 1].

[0082] Specifically, as shown in formula (2), at any moment, a wind turbine may be in normal operation, that is, it will output the ideal active power P(v) according to formula (1); or it may output αP(v) with a derating factor α due to some abnormal reasons, that is, it is in a derating operation state; in addition, due to serious faults or maintenance, the wind turbine may also be in a shutdown state, at which time its active power output is 0 (it should be noted that the 0 output at this time is different from the active shutdown caused by excessive wind speed).

[0083] At the same time, for any two adjacent moments, the operating state of the fan will transfer, wherein it may change from the normal operating state to the shutdown or derated operating state with a certain probability, or it may recover from the shutdown or derated operating state to the normal operating state with a certain probability.

[0084] It can be seen that when the reliability of the fan needs to be evaluated, it is necessary to evaluate it from two aspects: the time or probability of it being in each operating state, and the ability to switch between different states. In addition, since the derating factor of the fan may be constantly changing when it is in a derating operating state, it is necessary to further refine the operating state of the fan.

[0085] In some preferred embodiments, Figure 2 As shown, the evaluation unit obtains the average utilization vector and reliability matrix of each wind turbine in the wind farm through the following steps:

[0086] S21, obtaining operation and environmental data of each wind turbine in the wind farm.

[0087] S22, select any fan.

[0088] S23, pre-processing the operation and environmental data of the wind turbine, eliminating wind speeds below the cut-in wind speed v in Or greater than its cut-out wind speed v out The wind speed sampling value and the corresponding active power sampling value.

[0089] Specifically, by eliminating the sampling data when the wind speed exceeds the limit, it is avoided that when the active power sampling value of the wind turbine is 0, it is impossible to determine whether the shutdown is caused by wind speed or by its own fault, resulting in uncertainty in subsequent statistics.

[0090] S24, determine the expected active power value P of the wind turbine at each sampling time k based on the following formula: E (k):

[0091]

[0092] Formula (3) is the expression of Formula (1) after removing the wind speed exceeding the limit, where v S (k) is the wind speed sampling value of the wind turbine at sampling time k, K is the total number of sampling times after preprocessing, and the cumulative time of K sampling times is the total time for reliability assessment of the wind turbine.

[0093] S25, determining the operating state n(k) of the fan at each sampling time k based on the following formula:

[0094]

[0095] Among them, P S (k) is the active power sampling value of the wind turbine at each sampling time k, round[] is used to obtain the nearest integer of the value in the brackets, and N+1 is the total number of possible operating states of the wind turbine.

[0096] (4) compares the sampled value P of the actual active power of the wind turbine S (k) and the expected value of active power P determined according to the ideal output model E (k) At the same time, the derating operation status of the fan is further refined, and the operation status of the fan at any historical moment is classified into state 1 (i.e. shutdown state), state 2 to state N (i.e. multiple derating operation states with different derating coefficients) and state N+1 (i.e. normal operation state), so as to provide a basis for subsequent statistics.

[0097] S26, determining the average utilization vector AUR of the wind turbine based on the following formula:

[0098]

[0099] S27, determine the reliability matrix R of the wind turbine based on the following formula:

[0100]

[0101] Among them, the elements in R The operating state of the fan at two adjacent sampling moments is n1 Transfer to n 2 probability.

[0102] Specifically, AUR and R can be obtained by counting the numbers of the operating states obtained at each sampling time k. For example, the duration of the same operating state (such as state 1, 2, ... N+1) of a wind turbine at each sampling time k can be accumulated to obtain the cumulative duration of its being in each operating state, and then divided by the total duration of reliability assessment, thereby obtaining the vector AUR; for another example, the transition process of the operating state of the wind turbine at any two sampling times (such as sampling time k and k+1) can be counted in sequence (such as the process of the wind turbine from sampling time k to k+1, the operating state changes from N+1 to 1, indicating that the wind turbine changes from normal operation to shutdown), the number of the same transition process is accumulated, and divided by the total number of state transitions, thereby obtaining the matrix R. Obviously, as analyzed above, AUR is used to characterize the probability of a wind turbine being in each operating state. The greater the probability of being in a higher-numbered state, the longer the normal or near-normal operation time is; R characterizes the ability of a wind turbine to recover from a worse state to a better state, and the probability of failure from a better state.

[0103] S28, replacing the wind turbine and returning to step S23 until all wind turbines in the wind farm are traversed.

[0104] After obtaining the above-mentioned average utilization vector and reliability matrix of each wind turbine respectively, the evaluation unit can use various clustering algorithms known to technical personnel in this field to classify wind turbines with similar reliability into the same wind turbine cluster according to pre-set clustering rules, number of categories, and the upper limit of the number of wind turbines that can be included in each category, and finally obtain I wind turbine clusters.

[0105] [Hierarchical MPC allocation mechanism]

[0106] After the evaluation unit determines I wind turbine clusters through reliability evaluation, the central controller generates corresponding local controllers according to the number of wind turbine clusters to achieve hierarchical control.

[0107] Specifically, in the present application, a hierarchical MPC allocation mechanism is used for the wind turbines in the wind farm to reasonably allocate the active power output of each wind turbine in the wind farm. Figure 3 The following is a schematic diagram of the implementation process of the hierarchical MPC allocation mechanism in some preferred embodiments. Figure 3 , and its specific implementation method is described in detail.

[0108] <Overall MPC algorithm>

[0109] like Figure 3As shown in the figure, the central controller obtains the power grid output command P issued by the power grid. WG (Generally, P WG After the active power that needs to be supplied to the grid at each specified time in the future is calculated, the active power target value of each wind turbine cluster is determined by the MPC algorithm based on the cost function minimization objective shown in the following formula:

[0110]

[0111] Where Φ is the cost function, I is the total number of wind turbine clusters, T is the total length of the forecast period, is the active power target value of the i-th wind turbine cluster at the prediction time t, P WT,i is the active power reference value of the i-th wind turbine cluster at the prediction time t, are the shaft torque variation and thrust variation of the i-th wind turbine cluster at the prediction time t, Q p , Q T , Q F is the weight of each square term in the formula.

[0112] Accordingly, the constraints of the central controller when searching for the active power target value of each wind turbine cluster are as follows:

[0113]

[0114] in, is the initial value of active power of the i-th wind turbine cluster in the forecast period, is the active power increment of the i-th wind turbine cluster during the forecast period, is the active power rating of the i-th wind turbine cluster.

[0115] In some preferred embodiments, the active power reference value P of each wind turbine cluster used to construct the cost function WT,i It can be determined by the central controller based on the reliability evaluation results of each wind turbine. For example, the probability of the wind turbine being in each operating state can be combined with the wind speed forecast to obtain the active power output values ​​corresponding to each operating state with different probabilities output by the wind turbine at the predicted time t, and then sum them up to obtain P as shown in formula (9): WT,i :

[0116]

[0117] Obviously, compared with directly using the rated active power or the ideal active power under the predicted wind speed to determine the active power reference value, using formula (9) to take the reliability of the wind turbine's previous operation into consideration can more reasonably set the reference value at the future prediction time t, thereby improving the accuracy of the MPC algorithm.

[0118] <Local MPC algorithm>

[0119] In some preferred embodiments, Figure 3 As shown, each local controller generates a sequence of candidate active power values ​​of the wind turbine cluster it controls through the following steps (taking the i-th local controller as an example):

[0120] First, the central controller decomposes the cost function Φ into the local objective function Φ of each wind turbine cluster as shown in the following formula: i (u i ) is in the form of:

[0121]

[0122] Among them, u i is the candidate value of active power of the i-th wind turbine cluster;

[0123] Then, the central controller transforms the local objective function Φ of each wind turbine cluster into i (u i ) and active power target value Send to the corresponding local controller;

[0124] Finally, each local controller receives active power target value based on its received minΦ i (u i ) as the target, the MPC algorithm is used to generate the active power candidate value sequence corresponding to three consecutive prediction times t-1, t, and t+1

[0125] Generate a sequence of candidate active power values The purpose of this is to solve the fluctuation problem of the output command directly generated by MPC. In the actual operation of the wind turbine, if the output command changes frequently, the wind turbine will frequently change its working mode, which is not conducive to the stable output of power. Therefore, after each local controller generates three output results at adjacent moments, it needs to be further processed through the stability optimization operation described below. The number of alternative values ​​contained in the alternative value sequence can be selected according to the wind turbine's ability to withstand fluctuations, except Figure 3 In addition to the three shown, there may be four, five or more.

[0126] <Stability optimization operation>

[0127] refer to Figure 3 In some preferred embodiments, the central controller performs stability optimization on the active power candidate value sequence of each wind turbine cluster based on the following formula to determine the active power allocation value of each wind turbine cluster:

[0128]

[0129] in, is the active power distribution value of the i-th wind turbine cluster at the prediction time t, A, B, and C are The stability optimization coefficient.

[0130] Preferably, the stability optimization coefficient can be determined by searching for the minimum value of the stability optimization function f(A, B, C) using a gradient descent method, wherein the stability optimization function is specifically:

[0131]

[0132] The steps of optimizing A, B, and C using the above optimization function are known to those skilled in the art. For example, the following gradient descent formula can be continuously iterated until f(A, B, C) reaches a minimum value:

[0133]

[0134]

[0135]

[0136] Among them, α l is the learning rate, usually set to 0.0001.

[0137] Figure 4 The following schematically shows the distribution of the active power candidate values ​​before stability optimization and the active power distribution values ​​after optimization in a specific embodiment. Figure 4 As shown in FIG. 1 , the active power candidate value obtained after optimization can significantly improve stability.

[0138] Continue to refer Figure 3 The central controller optimizes the active power distribution value obtained by stability optimization After being sent to each local controller, it is further divided into active power output instructions of each wind turbine and sent to each wind turbine, thereby completing the distribution process from the power grid output instruction to the active power output instruction of each wind turbine.

[0139] The above is a detailed introduction to the specific implementation methods of the present application. For those skilled in the art, several improvements and modifications may be made to the present application without departing from the principles of the present application. These improvements and modifications also fall within the scope of protection of the claims of the present application.

Claims

1. A wind farm active power distribution system with reliability assessment and hierarchical MPC, characterized in that: include: The measurement unit is used to measure the operation and environmental data of each wind turbine in the wind farm; An evaluation unit, which evaluates the reliability of each wind turbine in the wind farm based on the operation and environmental data, and divides each wind turbine into a plurality of wind turbine clusters according to the reliability evaluation result; A central controller and a plurality of local controllers, wherein the plurality of local controllers are generated by the central controller and control the plurality of wind turbine clusters in a one-to-one correspondence; The central controller determines the active power target value of each wind turbine cluster through the MPC algorithm based on the reliability evaluation result and the cost function, and performs stability optimization on the active power candidate value sequence of each wind turbine cluster to determine the active power allocation value of each wind turbine cluster; Each of the local controllers receives an active power target value and a local objective function from the central controller, generates a sequence of active power alternative values ​​for the wind turbine cluster it controls through an MPC algorithm, and distributes active power output instructions to each wind turbine in the wind turbine cluster it controls based on the active power distribution value received from the central controller.

2. The wind farm active power distribution system of reliability assessment combined with hierarchical MPC according to claim 1 is characterized in that: The operation and environmental data of each wind turbine in the wind farm include active power sampling values ​​and wind speed sampling values ​​of each wind turbine at each sampling time; Furthermore, the evaluation unit evaluates the reliability of each wind turbine in the wind farm by obtaining an average utilization vector and a reliability matrix of each wind turbine in the wind farm, wherein the average utilization vector and the reliability matrix are determined based on the relationship between an ideal output model of each wind turbine and an operating condition probability model.

3. The wind farm active power distribution system of reliability assessment combined with hierarchical MPC according to claim 2 is characterized in that: The ideal output model is: Where v is the real-time wind speed, P(v) is the ideal active power value of the wind turbine when the wind speed is v, and v in 、v out are the cut-in wind speed and cut-out wind speed of the fan, respectively, v rate , P rate are the wind speed rating and active power rating of the wind turbine respectively; The operating condition probability model is: Among them, P status,P(v) , P status,0 , P status,αP(v) are the probabilities of the fan operating normally, shutting down, and operating at a reduced rating according to the derating factor α, τ1, τ2, τ3 are the durations of the fan operating normally, shutting down, and operating at a reduced rating, λ, λ d are the transfer rates of the fan from normal operation to shutdown and derating operation, μ, μ d are the recovery rates of the wind turbine from shutdown and derating operation to normal operation, and U1, U2, and U3 are random numbers uniformly distributed between [0,1].

4. The wind farm active power distribution system of reliability assessment combined with hierarchical MPC according to claim 3 is characterized in that: The evaluation unit obtains the average utilization vector and reliability matrix of each wind turbine in the wind farm through the following steps: S21, obtaining operation and environmental data of each wind turbine in the wind farm; S22, select any fan; S23, pre-processing the operation and environmental data of the wind turbine, eliminating wind speeds below the cut-in wind speed v in Or greater than its cut-out wind speed v out The wind speed sampling value and the corresponding active power sampling value; S24, determine the expected active power value P of the wind turbine at each sampling time k based on the following formula: E (k): Among them, v S (k) is the wind speed sampling value of the wind turbine at sampling time k, K is the total number of sampling times after preprocessing, and the cumulative time of K sampling times is the total time for reliability assessment of the wind turbine; S25, determining the operating state n(k) of the fan at each sampling time k based on the following formula: Among them, P S (k) is the active power sampling value of the wind turbine at each sampling time k, round[] is used to find the nearest integer of the value in the brackets, and N+1 is the total number of possible operating states of the wind turbine; S26, determining the average utilization vector AUR of the wind turbine based on the following formula: S27, determine the reliability matrix R of the wind turbine based on the following formula: Among them, the elements in R is the probability that the operating state of the fan changes from n1 to n2 at two adjacent sampling moments; S28, replacing the wind turbine and returning to step S23 until all wind turbines in the wind farm are traversed.

5. The wind farm active power distribution system of reliability assessment combined with hierarchical MPC according to claim 1 is characterized in that: The central controller determines the active power target value of each wind turbine cluster through the MPC algorithm based on the cost function minimization objective shown in the following formula: Where Φ is the cost function, I is the total number of wind turbine clusters, T is the total length of the forecast period, is the active power target value of the i-th wind turbine cluster at the prediction time t, P WT,i is the active power reference value of the i-th wind turbine cluster at the prediction time t, are the shaft torque variation and thrust variation of the i-th wind turbine cluster at the prediction time t, Q p , Q T , Q F is the weight of each square term in the formula.

6. The wind farm active power distribution system of reliability assessment combined with hierarchical MPC according to claim 5 is characterized in that: P WT,i It is determined by the central controller according to the reliability evaluation result.

7. The wind farm active power distribution system of reliability assessment combined with hierarchical MPC according to claim 5 is characterized in that: The constraints of the central controller when searching for the active power target value of each wind turbine cluster are specifically: Among them, P WG The power grid output command received by the wind farm. is the initial value of active power of the i-th wind turbine cluster in the forecast period, is the active power increment of the i-th wind turbine cluster during the forecast period, is the active power rating of the i-th wind turbine cluster.

8. The wind farm active power distribution system of reliability assessment combined with hierarchical MPC according to claim 5 is characterized in that: Each local controller generates a sequence of candidate active power values ​​for the wind turbine cluster it controls through the following steps: The central controller decomposes the cost function Φ into the local objective function Φ of each wind turbine cluster as shown in the following formula: i (u i ) is in the form of: Among them, u i is the candidate value of active power of the i-th wind turbine cluster; The central controller sends the local objective function and active power target value of each wind turbine cluster to the corresponding local controller; Each local controller receives active power target value based on its received minΦ i (u i ) 为目标 , the MPC algorithm is used to generate a sequence of candidate active power values ​​corresponding to three consecutive prediction times t-1, t, and t+1 9. The wind farm active power distribution system of reliability assessment combined with hierarchical MPC according to claim 8, characterized in that: The central controller performs stability optimization on the active power candidate value sequence of each wind turbine cluster based on the following formula to determine the active power allocation value of each wind turbine cluster: in, is the active power distribution value of the i-th wind turbine cluster at the prediction time t, A, B, and C are The stability optimization coefficient.

10. The wind farm active power distribution system of reliability assessment combined with hierarchical MPC according to claim 9, characterized in that: The stability optimization coefficient is determined by searching for the minimum value of the stability optimization function f(A, B, C) using the gradient descent method. The stability optimization function is specifically: