CN110676872B - Wind power plant active power distribution method considering fatigue load of unit - Google Patents

Wind power plant active power distribution method considering fatigue load of unit Download PDF

Info

Publication number
CN110676872B
CN110676872B CN201910836163.6A CN201910836163A CN110676872B CN 110676872 B CN110676872 B CN 110676872B CN 201910836163 A CN201910836163 A CN 201910836163A CN 110676872 B CN110676872 B CN 110676872B
Authority
CN
China
Prior art keywords
unit
wind
power
wind power
layer
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910836163.6A
Other languages
Chinese (zh)
Other versions
CN110676872A (en
Inventor
刘颖明
王瑛玮
王树旗
王晓东
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenyang University of Technology
Original Assignee
Shenyang University of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shenyang University of Technology filed Critical Shenyang University of Technology
Priority to CN201910836163.6A priority Critical patent/CN110676872B/en
Publication of CN110676872A publication Critical patent/CN110676872A/en
Application granted granted Critical
Publication of CN110676872B publication Critical patent/CN110676872B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
    • H02J3/46—Controlling of the sharing of output between the generators, converters, or transformers
    • H02J3/48—Controlling the sharing of the in-phase component
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A30/00—Adapting or protecting infrastructure or their operation
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E10/00—Energy generation through renewable energy sources
    • Y02E10/70—Wind energy
    • Y02E10/72—Wind turbines with rotation axis in wind direction
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E10/00—Energy generation through renewable energy sources
    • Y02E10/70—Wind energy
    • Y02E10/76—Power conversion electric or electronic aspects

Landscapes

  • Engineering & Computer Science (AREA)
  • Power Engineering (AREA)
  • Supply And Distribution Of Alternating Current (AREA)
  • Wind Motors (AREA)

Abstract

The method comprises the steps that 1) a first layer is a wind power cluster distribution layer, the wind power cluster distribution layer calculates the power generation capacity by utilizing wind speed prediction information of each wind power plant, and cluster active reference values are distributed to the wind power plant distribution layer according to the proportion of the power generation capacity values calculated by each wind power plant; 2) The second layer is a wind power plant distribution layer, and is used for distributing cluster active reference values received in the first layer to different groups by adopting a priority sorting method according to a clustering analysis result; 3) And the third layer is a unit distribution layer, and active reference values received by various units are distributed to the units. The invention well solves the problems existing in the past, reduces the impact of high-permeability wind power on a power grid, strengthens the coordination and coordination in the wind power cluster and improves the output under the condition of not influencing or even reducing the fatigue load of a unit.

Description

Wind power plant active power distribution method considering fatigue load of unit
Technical Field
The invention belongs to the field of wind power plant application, and particularly relates to a wind power plant active power distribution method considering unit fatigue load. And the wind power plant layer clusters the units by using a fuzzy c-means clustering method according to the fatigue load generated by the low-speed shaft torque and the tower bending moment of the units in the plant, and distributes a wind power plant power reference value to the unit layer.
Background
With large-scale wind power application represented by wind power plant clusters, randomness, unpredictability and low controllability of wind power bring serious challenges to power balance and safe operation of a power grid. The existing wind power cluster active scheduling method is rough, so that the active output adjusting precision of the scheduled wind power cluster is poor, the output is low, and the change of the reference value of the power of a wind power plant also causes frequent actions of a unit, so that the fatigue load of the wind power unit (hereinafter referred to as the unit) is increased, and the service life of the unit is shortened. In order to reduce the impact of high-permeability wind power on a power grid, under the condition of not influencing or even reducing fatigue load of a unit, the coordination inside a wind power cluster is enhanced, the output is improved, and further, an active power optimal distribution method of a wind power plant cluster is researched.
Disclosure of Invention
The purpose of the invention is as follows:
the invention aims to provide a wind power plant active power distribution method considering unit fatigue load, which is used for solving the problem of rough wind power plant cluster active power distribution method and improving the wind power cluster output under the condition of not influencing the unit fatigue load.
The technical scheme is as follows:
a wind power plant active power distribution method considering unit fatigue loads is characterized in that: the method performs power allocation through three levels:
the method comprises the following specific steps:
1) The first layer is a wind power cluster distribution layer, the wind power cluster distribution layer calculates the power generation capacity by using the wind speed prediction information of each wind power plant, and the cluster active reference value is distributed to the wind power plant distribution layer according to the proportion of the power generation capacity value calculated by each wind power plant;
2) The second layer is a wind power plant distribution layer, and is used for distributing the cluster active reference values received in the first layer to different groups by adopting a priority sorting method according to the result of cluster analysis;
3) And the third layer is a unit distribution layer, and active reference values received by various units are distributed to the units.
And the unit layer uses different fatigue loads of various units as a target, calculates the power reference value of the units in real time by using a quadratic programming algorithm and distributes the power reference value to each unit.
The method for calculating the power which can be generated by utilizing the wind speed prediction information of the wind power plant in the first layer comprises the following steps: the method comprises the steps of firstly, decomposing a wind speed sequence into a series of relatively stable components by utilizing EMD (empirical mode decomposition), wherein the components are different in frequency to the greatest extent;
then according to different component frequencies, an RBF (radial basis function) kernel function is selected as a kernel function of the SVM (support vector machine) for predicting the high-frequency component, and the calpain is generated according to k (x, y) = exp (— | | x-y) | 2 /σ 2 ) Determining related parameters sigma (wherein x and y are input values of n dimension, sigma is kernel parameter), predicting kernel function of intermediate frequency component according to selected polynomial kernel function as SVM (support vector machine), and predicting kernel function according to k (x, y) = [ (alpha x.y) + c] d Determining related parameters c, d (wherein alpha is a slope, x and y are input values of n dimensions, c is a constant term, and d is a polynomial degree); selecting a linear kernel function as a kernel function of the low-frequency component predicted by the SVM (support vector machine), and determining a related parameter c according to k (x, y) = x.y + c;
the prediction model is obtained by the formula:
Figure BDA0002192222980000021
in the formula: r reg Is a structural risk function; | w | charging 2 Is a described kernel function; f (x) is the model complexity; c is a constant; r emp Is a risk of experience
Then inputting prediction data and processing the generated prediction result;
and finally, superposing the wind speed prediction results and inputting the wind speed prediction results into a power conversion curve to obtain a wind power prediction result.
The power that can be sent in the first step is the maximum power that can be sent in the wind-powered electricity generation field, and its calculation method is as follows:
firstly, calculating the power which can be generated by a single unit according to the predicted wind speed, wherein the formula is as follows:
Figure BDA0002192222980000031
in the formula, P rated Is rated power; ρ is the air density; v. of nac A predicted value of the wind speed of the engine room is obtained;
Figure BDA0002192222980000032
is the power that can be generated by the unit; c p Is a wind energy conversion rate value; r is the radius of the wind wheel;
the maximum power generation of the wind power plant is the sum of the power generation of all the units in the wind power plant, and the calculation is as the following formula:
Figure BDA0002192222980000033
in the formula (I), the compound is shown in the specification,
Figure BDA0002192222980000034
the power can be generated for the jth wind power plant; and n is the number of the units in the wind power plant.
The method for distributing the cluster active reference values to the wind power plant distribution layer in proportion in the first layer comprises the following steps:
the wind power cluster active power distribution method is characterized in that active power dispatching instructions are distributed to each wind power plant in proportion according to the power which can be generated by the wind power plant, so that the purpose of fair distribution is achieved. Such as the formula:
Figure BDA0002192222980000035
in the formula:
Figure BDA0002192222980000036
an active power value distributed for a jth wind power plant;
Figure BDA0002192222980000037
and m is the number of wind power plants.
In the second layer, the wind farm distribution layer is according to the formula:
Figure BDA0002192222980000038
in the formula: j. the design is a square g Is the rotational inertia of the generator; j. the design is a square r Is the rotor moment of inertia; t is g Is the generator torque; delta T s Low speed shaft torque; delta T a Is an aerodynamic torque; eta g Is the gear box speed ratio; j. the design is a square t Is equivalent inertia
M T ≈H*F T
In the formula: h is the height of the tower; f T Thrust of the tower barrel is obtained; m is a group of T Is a tower bending moment
And clustering the wind turbines by adopting a fuzzy c-means clustering method for the fatigue loads generated by the low-speed shaft torque and the tower drum bending moment obtained by the formula, and distributing the power reference value to the turbine layer by adopting a priority sorting method.
In the second step: the active power priority ordering distribution method of the wind power plant layer based on the fuzzy c-means clustering comprises the following steps:
the range of the optimal cluster number is set as:
Figure BDA0002192222980000041
(wherein n is the number of units); the initial clustering center merging method comprises the following steps: will be provided with
Figure BDA0002192222980000042
The distance between every two initial clustering centers is calculated, and the arithmetic mean value is calculated for the two closest clustering centers to obtain
Figure BDA0002192222980000043
An initial cluster center.
Extracting an average value of the output power of the unit; extracting a wind speed root-mean-square difference value as a characteristic value, and then obtaining a characteristic matrix of the wind power and the wind speed of the unit as follows:
Figure BDA0002192222980000044
in the matrix, P me(0-1) (i) Average value v after normalization of active power of ith unit RM(0-1) (i) And normalizing the root mean square difference value after the wind speed of the ith unit.
The characteristic matrix is used as input of a fuzzy c-means algorithm, the active power mean value 1MV and the wind speed root-mean-square difference value 1.2 are used as standards, the units in the wind power plant are clustered, and the clustering is divided into four types, wherein the first type is that the active power mean value is small (smaller than 1 MV) and the wind speed root-mean-square difference value is large (larger than 1.2) (PSVB); the second type is that the average value of active power is large, and the root mean square difference value of wind speed is small (PBVS); the third category is that the average value of active power is large, and the root mean square difference value of wind speed is large (PBVB); the fourth type is that the mean value of active power is small, and the root mean square difference value of wind speed is small (PSVS);
based on the power distribution of the wind power plant layer of a priority ordering method, firstly distributing power to a fourth type unit and a second type unit with small root mean square difference of wind speeds, and distributing residual power to the first type unit and the third type unit;
1) The arrangement sequence of the distributed active power reference values: sequence 1 is a fourth type and a second type of unit,
Figure BDA0002192222980000051
the active reference value assigned for the fourth class,
Figure BDA0002192222980000052
active reference values allocated for the second class; sequence 2 is a first type and a third type of units,
Figure BDA0002192222980000053
the active reference value assigned for the first class,
Figure BDA0002192222980000054
and allocating an active reference value for the third class.
2)
Figure BDA0002192222980000055
Active power can be generated for each sequence unit;
Figure BDA0002192222980000056
the power predicted value of the wind power plant j is obtained;
Figure BDA0002192222980000057
is each timeTaking the lower limit of total active power set for avoiding frequent start and stop of the unit in a sequence
Figure BDA0002192222980000058
P demand And the active power reference value of the wind power plant j is obtained.
In the third step, the unit layer uses the minimum fatigue load of different units as a target, calculates the power reference value of the unit in real time by using the existing quadratic programming algorithm and distributes the power reference value to each unit so as to minimize the whole fatigue load of the unit;
the optimal distribution of the unit layer based on the quadratic programming can effectively reduce the whole fatigue load of the unit.
The objective function is as follows:
Figure BDA0002192222980000059
wherein: n is the number of various units, a and b are respectively the optimized coefficients of the torque of the transmission chain and the bending moment load of the tower barrel, and different values are obtained according to the unit classification. Delta T s 2 For low-speed torque fluctuations of the unit, Δ M t 2 The tower bending moment fluctuates.
The constraint conditions are as follows:
1) The unit output limit is restricted, and in order to avoid the unit halt, the unit output lower limit is set to be 5% of the output, as shown in the formula (2)
Figure BDA00021922229800000510
In the formula (I), the compound is shown in the specification,
Figure BDA0002192222980000061
is the available active power of the unit number i,
Figure BDA0002192222980000062
active reference value distributed for No. i machine set
2) Force tracking constraints, e.g. equations (3) and (4)
Figure BDA0002192222980000063
Figure BDA0002192222980000064
In the formula (I), the compound is shown in the specification,
Figure BDA0002192222980000065
the active power value required by the unit layer.
The utility model provides a consideration unit fatigue load's wind-powered electricity generation field active power distribution system which characterized in that: the system comprises a wind power cluster distribution module, a wind power field distribution module and a unit distribution module;
the wind power cluster distribution module calculates the power generation capacity by using the wind speed prediction information of each wind power plant, and distributes the cluster active reference value to the wind power plant distribution module according to the proportion of the power generation capacity values calculated by each wind power plant;
the wind power plant distribution module distributes the cluster active reference values received in the wind power cluster distribution module to different groups by adopting a priority method according to the result of cluster analysis; the unit distribution module distributes the active reference values received by the various units to the units.
The advantages and effects are as follows:
in order to achieve the purpose of the invention, the invention adopts the following technical scheme that: the wind power cluster active power distribution method considering the fatigue load of the unit comprises the following steps:
the wind power plant cluster framework provided by the invention is mainly divided into three parts:
the first part is a wind power cluster distribution layer, and the first part mainly distributes a grid active power reference value to each wind power plant.
The cluster layer calculates the power generation capacity by using wind power plant wind speed prediction information, and distributes cluster active reference values to the wind power plant layer according to the power generation capacity ratio so as to realize the average distribution among the wind power plants.
The second part is a wind power plant distribution layer, and the second part is mainly used for distributing power reference values received by a wind power plant to different groups according to clustering analysis results.
The wind power field layer clusters the wind turbine generators by adopting a fuzzy C-means clustering method according to fatigue loads generated by low-speed shaft torque of the generators and bending moment of a tower drum, and distributes power reference values to the generator layer by adopting a priority ordering method so as to reduce the different types of fatigue loads of the different types of generators.
The third part is a unit distribution layer, and the third part mainly distributes the active reference values received by various units to the units.
The unit layer aims at minimizing different fatigue loads of various units, calculates the power reference value of the units in real time by using a quadratic programming algorithm and distributes the power reference value to each unit so as to minimize the overall fatigue load of the units.
The method comprises the steps of firstly, decomposing a wind speed sequence into a series of relatively stable components by utilizing EMD (empirical mode decomposition) so as to reduce the mutual influence among different characteristic information; then, a prediction model is established for each component by using an SVM method, and different kernel functions and related parameters are selected according to the self characteristics of each sequence to process each group of different data so as to improve the prediction precision of a single model. And finally, superposing the wind speed prediction results and inputting the wind speed prediction results into a power conversion curve to obtain a wind power prediction result.
The power which can be generated by a single unit is calculated through the predicted wind speed, and the formula is as follows:
Figure BDA0002192222980000071
in the formula, P rated Is rated power; ρ is the air density; v. of nac A predicted value of the wind speed of the engine room is obtained;
Figure BDA0002192222980000072
is the power that can be generated by the unit; and R is the radius of the wind wheel.
The maximum power generation of the wind power plant is the sum of the power generation of all the units in the wind power plant, and the calculation is as the following formula:
Figure BDA0002192222980000073
in the formula (I), the compound is shown in the specification,
Figure BDA0002192222980000074
the power can be generated for the jth wind power plant; and n is the number of the units in the wind power plant.
The active distribution method for the wind power cluster distribution is characterized in that active scheduling instructions are distributed to each wind power plant in proportion according to the power which can be generated by the wind power plant, so that the purpose of fair distribution is achieved. Such as the formula:
Figure BDA0002192222980000081
in the formula:
Figure BDA0002192222980000082
an active power value distributed for the jth wind power plant;
Figure BDA0002192222980000083
and (5) an active power scheduling instruction is given to the wind power cluster.
The active power priority sequencing distribution method of the wind power plant layer based on the fuzzy c-means clustering sets the range of the optimal clustering number as follows: (
Figure BDA0002192222980000084
Where n is the number of units). The initial clustering center merging method comprises the following steps: will be provided with
Figure BDA0002192222980000085
The distance between every two initial clustering centers is calculated, and the arithmetic mean value is calculated for the two clustering centers with the shortest distance to obtain
Figure BDA0002192222980000086
An initial cluster center.
Because the torque of the low-speed shaft of the unit is mainly related to the power of the unit, the average value of the output power of the unit is extracted; the bending moment of the tower of the unit is mainly related to the wind speed fluctuation of the unit, so that the root mean square difference of the wind speeds is extracted as a characteristic value. Then obtaining a characteristic matrix of the wind power and the wind speed of the unit as
Figure BDA0002192222980000087
In the matrix, P me(0-1) (i) Average value v after normalization of active power of ith unit RM(0-1) (i) And normalizing the root mean square difference value after the wind speed of the ith unit is normalized.
The characteristic matrix is used as the input of a fuzzy c-means algorithm, the units in the wind power plant are clustered, and the clustering is divided into four categories, wherein the first category is that the active power mean value is small, and the wind speed root mean square difference value (PSVB) is large; the second type is that the mean value of active power is large, and the root mean square difference value of wind speed is small (PBVS); the third category is that the average value of active power is large, and the root mean square difference value of wind speed is large (PBVB); the fourth category is small mean active power and small root mean square difference wind speed (PSVS).
And distributing power to the fourth type unit and the second type unit with small wind speed root mean square difference value and distributing residual power to the first type unit and the third type unit based on the power distribution of the wind power plant layer of the priority sorting method.
The sequence of the distributed active power reference values is as follows: sequence 1 is a fourth type and a second type of unit,
Figure BDA0002192222980000091
the active reference value assigned for the fourth class,
Figure BDA0002192222980000092
active reference values allocated for the second class; sequence 2 is a first type and a third type of units,
Figure BDA0002192222980000093
the active reference value assigned for the first class,
Figure BDA0002192222980000094
and allocating an active reference value for the third class.
Figure BDA0002192222980000095
Active power can be generated for each sequence unit;
Figure BDA0002192222980000096
the power predicted value of the wind power plant j is obtained;
Figure BDA0002192222980000097
taking the lower limit of the total active power set for avoiding frequent start and stop of the unit in each sequence
Figure BDA0002192222980000098
P demand And the active power reference value of the wind power plant j is obtained.
The method is characterized in that a fatigue load-based unit layer active power optimal distribution layer is adopted, and the optimization method is mainly used for reducing the integral delta M for a first-class unit (PSVB) t The load generated by the bending moment of the tower of the unit is reduced, and the second type of unit (PBVS) mainly considers the reduction of the integral delta T s The load generated by the torque of the low-speed shaft of the unit is reduced, and the delta T of the third type unit (PBVB) s And Δ M t Meanwhile, the fourth-class unit (PSVS) does not need an optimization algorithm.
The main calculation formula of the fatigue load calculation model is as follows:
Figure BDA0002192222980000099
Figure BDA00021922229800000910
Figure BDA00021922229800000911
Figure BDA00021922229800000912
in the formula: omega g Is the generator speed; β is a defined function related to θ; omega f Filtering the rotating speed of the generator; eta g The gear ratio of the gear box is adopted (the direct drive unit is 1); jt is equivalent inertia; t is a unit of a Is an aerodynamic torque; p g The actual generating capacity of the unit is obtained; k p 、K i Is a gain constant; tau is f Is the filter time constant; omega g-rated The rated rotation speed of the generator. Wherein the addition of 0 subscript indicates the value of the sample at the current time of the respective parameter, e.g. T a0 Is the aerodynamic torque at time k.
Figure BDA0002192222980000101
Figure BDA0002192222980000102
Figure BDA0002192222980000103
Figure BDA0002192222980000104
Through the shaft motion equation of the unit, the delta T can be obtained s Such as the formula:
Figure BDA0002192222980000105
in the formula: j. the design is a square g Is the rotational inertia of the generator; j. the design is a square r Is the rotor moment of inertia; t is g Is the generator torque.
Conversion to:
Figure BDA0002192222980000106
namely:
Figure BDA0002192222980000107
tower drum bending moment M t The calculation can be simplified as in the formula:
M T ≈H*F T
in the formula: h is the height of the tower; f T Thrust for tower
Based on the optimized distribution of the secondary planning to the unit layer, the unit layer optimization target is to reduce the whole fatigue load of the unit under the condition of tracking the output, and the load is optimized according to the delta T s (k + 1) and Δ M t (k + 1) (abbreviated as "Δ T s "and" Δ M t ") computational model, performing different target optimization for each type of unit. The objective function is as follows:
Figure BDA0002192222980000108
wherein: n is the number of various units, a and b are respectively the optimized coefficients of the torque of the transmission chain and the bending moment load of the tower barrel, and different values are obtained according to the unit classification.
The constraint conditions are as follows:
1. the output limit of the unit is restricted, in order to avoid the shutdown of the unit, the lower output limit of the unit is set to be 5% of the output of the unit, such as a formula:
Figure BDA0002192222980000111
in the formula (I), the compound is shown in the specification,
Figure BDA0002192222980000112
is the available active power of the unit number i,
Figure BDA0002192222980000113
and allocating an active reference value for the No. i unit.
2. Force tracking constraints, such as the formula:
Figure BDA0002192222980000114
Figure BDA0002192222980000115
in the formula (I), the compound is shown in the specification,
Figure BDA0002192222980000116
and the active power value required by the unit layer.
According to the method, a wind power plant layer clusters the units by using a fuzzy c-means clustering method according to the fatigue load generated by the low-speed shaft torque and tower bending moment of the units in the plant, and distributes a wind power plant power reference value to the unit layer.
The invention well solves the problems existing in the past, reduces the impact of high-permeability wind power on a power grid, strengthens the coordination and coordination in the wind power cluster and improves the cluster output under the condition of not influencing or even reducing the fatigue load of a unit, ensures that the output distribution is more uniform on the basis of fully considering the variation trend of the wind power, and simultaneously reduces the loss of a feeder line.
Drawings
FIG. 1 is a block diagram of overall control of a wind farm cluster;
fig. 2 is a control flow chart of two sequence units.
Detailed Description
A wind power plant active power distribution method considering unit fatigue loads is characterized in that: the method performs power distribution through three levels:
the method comprises the following specific steps:
1) The first layer is a wind power cluster distribution layer, the wind power cluster distribution layer calculates the power generation capacity by using wind speed prediction information of each wind power plant, and the cluster active power reference value is distributed to the wind power plant distribution layer according to the proportion of the power generation capacity values calculated by each wind power plant;
2) The second layer is a wind power plant distribution layer, and is used for distributing cluster active reference values received in the first layer to different groups by adopting a priority sorting method according to a clustering analysis result;
3) And the third layer is a unit distribution layer, and active reference values received by various units are distributed to the units.
And the unit layer uses different fatigue loads of various units as a target, calculates the power reference value of the units in real time by using a quadratic programming algorithm and distributes the power reference value to each unit.
The method for calculating the power which can be generated by utilizing the wind speed prediction information of the wind power plant in the first layer comprises the following steps: the method comprises the steps of firstly, decomposing a wind speed sequence into a series of relatively stable components by utilizing EMD (empirical mode decomposition), wherein the components are different in frequency to the greatest extent;
then according to different component frequencies, an RBF (radial basis function) kernel function is selected as a kernel function of the SVM (support vector machine) for predicting the high-frequency component, and the calpain is generated according to k (x, y) = exp (— | | x-y) | 2 /σ 2 ) Determining related parameters sigma (wherein x and y are input values of n dimensions, sigma is a kernel parameter), predicting kernel function of intermediate frequency component according to selected polynomial kernel function as SVM (support vector machine), and according to k (x, y) = [ (alpha x.y) + c] d Determining related parameters c, d (wherein alpha is a slope, x and y are input values of n dimensions, c is a constant term, and d is a polynomial degree); selecting a linear kernel function as a kernel function of the low-frequency component predicted by the SVM (support vector machine), and determining a related parameter c according to k (x, y) = x.y + c;
the prediction model is obtained by the formula:
Figure BDA0002192222980000121
in the formula: r reg Is a structural risk function; | w | charging 2 Is a described kernel function; f (x) is the model complexity; c is a constant; r is emp Is a risk of experience
Then inputting prediction data and processing the generated prediction result;
and finally, superposing the wind speed prediction results and inputting the wind speed prediction results into a power conversion curve to obtain a wind power prediction result.
The method comprises the following steps that the power which can be generated in the first step is the maximum power which can be generated in the wind power plant, and the calculation method comprises the following steps:
firstly, calculating the power which can be generated by a single unit according to the predicted wind speed, wherein the formula is as follows:
Figure BDA0002192222980000131
in the formula, P rated Is the rated power; ρ is the air density; v. of nac A predicted value of the wind speed of the engine room is obtained;
Figure BDA0002192222980000132
is the power that can be generated by the unit; c p Is a wind energy conversion rate value; r is the radius of the wind wheel;
the maximum power generation of the wind power plant is the sum of the power generation of all the units in the wind power plant, and the calculation is as the following formula:
Figure BDA0002192222980000133
in the formula (I), the compound is shown in the specification,
Figure BDA0002192222980000134
the power can be generated for the jth wind power plant; and n is the number of the units in the wind power plant.
The method for distributing the cluster active reference values to the wind power plant distribution layer in proportion is as follows:
the wind power cluster active power distribution method is characterized in that active power dispatching instructions are distributed to each wind power plant in proportion according to the power which can be generated by the wind power plant, so that the purpose of fair distribution is achieved. Such as the formula:
Figure BDA0002192222980000135
in the formula:
Figure BDA0002192222980000136
an active power value distributed for a jth wind power plant;
Figure BDA0002192222980000137
and (5) scheduling instructions for the active power of the wind power cluster, wherein m is the number of wind power plants.
In the second layer, the wind farm distribution layer is according to the formula:
Figure BDA0002192222980000141
in the formula: j is a unit of g Is the rotational inertia of the generator; j. the design is a square r Is the rotor moment of inertia; t is g Is the generator torque; delta T s Low speed shaft torque; delta T a Is an aerodynamic torque; eta g Is the gear box speed ratio; j is a unit of t Is equivalent inertia
M T ≈H*F T
In the formula: h is the height of the tower; f T Thrust of the tower barrel is obtained; m T Bending moment for tower
And clustering the wind turbines by adopting a fuzzy c-means clustering method for the fatigue loads generated by the low-speed shaft torque and the tower drum bending moment obtained by the formula, and distributing the power reference value to the turbine layer by adopting a priority sorting method.
In the second step: the active power priority ordering distribution method of the wind power plant layer based on the fuzzy c-means clustering comprises the following steps:
the range of the optimal cluster number is set as:
Figure BDA0002192222980000142
(wherein n is the number of units); the initial clustering center merging method comprises the following steps: will be provided with
Figure BDA0002192222980000143
The distance between every two initial clustering centers is calculated, and the arithmetic mean value is calculated for the two clustering centers with the shortest distance to obtain
Figure BDA0002192222980000144
An initial cluster center.
Extracting an average value of the output power of the unit; extracting a wind speed root-mean-square difference value as a characteristic value, and then obtaining a characteristic matrix of wind power and wind speed of the unit as follows:
Figure BDA0002192222980000145
in the matrix, P me(0-1) (i) Average value v after normalization of active power of ith unit RM(0-1) (i) And normalizing the root mean square difference value after the wind speed of the ith unit.
The characteristic matrix is used as the input of a fuzzy c-means algorithm, the unit in the wind power plant is clustered by taking an active power mean value 1MV and a wind speed root-mean-square difference value 1.2 as standards, and the unit is divided into four types, wherein the first type is that the active power mean value is small (smaller than 1 MV) and the wind speed root-mean-square difference value is large (larger than 1.2) (PSVB); the second type is that the average value of active power is large, and the root mean square difference value of wind speed is small (PBVS); the third category is that the average value of active power is large, and the root mean square difference value of wind speed is large (PBVB); the fourth type is that the mean value of active power is small, and the root mean square difference value of wind speed is small (PSVS);
the method comprises the steps that power of a wind power plant layer is distributed based on a priority method, power is distributed to a fourth type unit and a second type unit which are small in wind speed root-mean-square difference, and residual power is distributed to the first type unit and the third type unit;
3) The sequence of the distributed active power reference values is as follows: sequence 1 is a fourth type and a second type of unit,
Figure BDA0002192222980000151
the active reference value assigned for the fourth class,
Figure BDA0002192222980000152
active reference values allocated for the second class; sequence 2 is a first type and a third type of units,
Figure BDA0002192222980000153
the active reference value assigned for the first class,
Figure BDA0002192222980000154
and allocating an active reference value for the third class.
4)
Figure BDA0002192222980000155
Active power can be generated for each sequence unit;
Figure BDA0002192222980000156
a power predicted value of the wind farm j;
Figure BDA0002192222980000157
taking the lower limit of the total active power set for avoiding frequent start and stop of the unit in each sequence
Figure BDA0002192222980000158
P demand And the active power reference value of the wind power plant j is obtained.
In the third step, the unit layer uses the minimum fatigue load of different units as a target, calculates the power reference value of the unit in real time by using the existing quadratic programming algorithm and distributes the power reference value to each unit so as to minimize the whole fatigue load of the unit;
the optimal distribution of the unit layer based on the quadratic programming can effectively reduce the whole fatigue load of the unit.
The objective function is as follows:
Figure BDA0002192222980000159
wherein: n is the number of various units, a and b are respectively the optimized coefficients of the torque of the transmission chain and the bending moment load of the tower barrel, and different values are obtained according to the unit classification. Delta T s 2 For low-speed torque fluctuations, Δ M, of the unit t 2 The tower bending moment fluctuates.
The constraint conditions are as follows:
1) The unit output limit is restricted, in order to avoid the unit halt, the lower limit of the unit output is set to be 5% of the output, as shown in the formula (2)
Figure BDA0002192222980000161
In the formula (I), the compound is shown in the specification,
Figure BDA0002192222980000162
is the available active power of the unit number i,
Figure BDA0002192222980000163
active reference value distributed for No. i machine set
2) Force tracking constraints, e.g. equations (3) and (4)
Figure BDA0002192222980000164
Figure BDA0002192222980000165
In the formula (I), the compound is shown in the specification,
Figure BDA0002192222980000166
the active power value required by the unit layer.
The utility model provides a consider wind-powered electricity generation field active power distribution system of unit fatigue load which characterized in that: the system comprises a wind power cluster distribution module, a wind power field distribution module and a unit distribution module;
the wind power cluster distribution module calculates the power generation capacity by using the wind speed prediction information of each wind power plant, and distributes the cluster active reference value to the wind power plant distribution module according to the proportion of the power generation capacity values calculated by each wind power plant;
the wind power plant distribution module distributes the cluster active power reference values received in the wind power cluster distribution module to different groups by adopting a priority method according to the result of cluster analysis; the unit distribution module distributes the active reference values received by the various units to the units.
The present invention will be described in further detail below by way of examples with reference to the accompanying drawings, which are illustrative of the present invention and are not to be construed as limiting the present invention.
As shown in fig. 1, a wind farm cluster architecture mainly includes three parts:
the first part is a wind power cluster distribution layer, and the first part mainly distributes a grid active power reference value to each wind power plant.
The cluster layer calculates the power generation capacity by using wind power plant wind speed prediction information, and distributes the cluster active power reference value to the wind power plant layer according to the power generation capacity proportion so as to realize the average distribution among the wind power plants.
The second part is a wind power plant distribution layer, and the second part is mainly used for distributing power reference values received by a wind power plant to different groups according to clustering analysis results.
The wind power field layer clusters the wind turbine generators by adopting a fuzzy C-means clustering method according to fatigue loads generated by low-speed shaft torque of the generators and bending moment of a tower drum, and distributes power reference values to the generator layer by adopting a priority ordering method so as to reduce the different types of fatigue loads of the different types of generators.
The third part is a unit distribution layer, and the third part mainly distributes the active reference values received by various units to the units.
The unit layer aims at minimizing different fatigue loads of various units, calculates the power reference value of the units in real time by using a quadratic programming algorithm and distributes the power reference value to each unit so as to minimize the overall fatigue load of the units.
The invention discloses a cluster layer active power proportion distribution method based on EMD-SVM wind speed prediction information, which comprises the steps of firstly decomposing a wind speed sequence into a series of relatively stable components by utilizing EMD so as to reduce the mutual influence among different characteristic information; then, a prediction model is established for each component by using an SVM method, and different kernel functions and related parameters are selected according to the self characteristics of each sequence to process each group of different data so as to improve the prediction precision of a single model. And finally, superposing the wind speed prediction results and inputting the wind speed prediction results into a power conversion curve to obtain a wind power prediction result.
The power which can be generated by a single unit is calculated by the predicted wind speed, such as the formula:
Figure BDA0002192222980000171
in the formula, P rated Is rated power; ρ is the air density; v. of nac The predicted value of the wind speed of the engine room is taken;
Figure BDA0002192222980000172
is the power that can be generated by the unit; and R is the radius of the wind wheel.
The maximum power generation of the wind power plant is the sum of the power generation of all the units in the wind power plant, and the calculation is as the following formula:
Figure BDA0002192222980000173
in the formula (I), the compound is shown in the specification,
Figure BDA0002192222980000174
generating power for the jth wind power plant; and n is the number of the units in the wind power plant.
The wind power cluster active distribution method is characterized in that active scheduling instructions are distributed to each wind power plant in proportion according to the power which can be generated by the wind power plant, so that the purpose of fair distribution is achieved. Such as the formula:
Figure BDA0002192222980000181
in the formula:
Figure BDA0002192222980000182
an active power value distributed for a jth wind power plant;
Figure BDA0002192222980000183
and (5) an active power scheduling instruction is given to the wind power cluster.
For the power distribution effect of the wind power cluster, the method takes 1min as a control period to perform proportional distribution based on the power predicted value. When the control cycle of the cluster active power distribution method is 1min, the active power output is larger than that of a proportion distribution method which is updated once in 5min and 15 min. When the control period is 5min and 15min, because the power prediction has errors and the predicted value is the average value of 5min and 15min, the real-time output of the wind power plant cannot be represented, so that the wind power plant can obtain new distributed power every 5min and 15min, the output of the wind power plant can be changed during the period, the output cannot accurately follow the scheduling value, the adjustment precision is low, and if a shorter control period is adopted, the power adjustment difficulty is increased due to the limitation of communication delay and pitch angle adjustment time.
On the level of distribution of a wind power plant layer, the method divides the units in the wind power plant into four types of units by using a fuzzy c-means clustering algorithm according to the historical synchronous wind speed and output power data of each unit, and then performs power distribution on the four types of units by a priority method according to the unit classification result.
The active power priority ranking distribution method of the wind power plant layer based on the fuzzy c-means clustering sets the range of the optimal clustering number as: (
Figure BDA0002192222980000184
Where n is the number of units). The initial clustering center merging method comprises the following steps: will be provided with
Figure BDA0002192222980000185
The distance between every two initial clustering centers is calculated, and the arithmetic mean value is calculated for the two clustering centers with the shortest distance to obtain
Figure BDA0002192222980000186
An initial cluster center.
Because the torque of the low-speed shaft of the unit is mainly related to the power of the unit, extracting the average value of the output power of the unit; the bending moment of the tower barrel of the unit is mainly related to the fluctuation of the wind speed of the unit, so that the root mean square difference of the wind speed is extracted as a characteristic value. Then obtaining a characteristic matrix of wind power and wind speed of the unit as
Figure BDA0002192222980000191
In the matrix, P me(0-1) (i) Average value v after normalization of active power of ith unit RM(0-1) (i) And normalizing the root mean square difference value after the wind speed of the ith unit is normalized.
The characteristic matrix is used as input of a fuzzy c-means algorithm to cluster units in the wind power plant, the first type is that the mean value of active power is small, and the root mean square difference value of wind speed is large (PSVB); the second type is that the average value of active power is large, and the root mean square difference value of wind speed is small (PBVS); the third category is that the average value of active power is large, and the root mean square difference value of wind speed is large (PBVB); the fourth category is small mean value of active power and small root mean square difference of wind speed (PSVS).
The wind power plant layer power distribution based on the priority ranking method is characterized in that the wind speed root-mean-square difference is small, so that the fluctuation of the wind speed in the time interval is small, and the tower bending moment generated by the fluctuation of the wind speed is small, so that the unit layer distribution firstly considers the load generated by reducing the tower bending moment and then considers the load generated by reducing the low-speed shaft torque, the power is distributed to the fourth type unit and the second type unit with small wind speed root-mean-square difference, and the residual power is distributed to the first type unit and the third type unit.
The sequence of the distributed active power reference values is as follows: sequence 1 is a fourth type and a second type of unit,
Figure BDA0002192222980000192
the active reference value assigned for the fourth class,
Figure BDA0002192222980000193
active reference values allocated for the second class; sequence 2 is a first type and a third type of units,
Figure BDA0002192222980000194
the active reference value assigned for the first class,
Figure BDA0002192222980000195
and allocating an active reference value for the third class.
Figure BDA0002192222980000196
Active power can be generated for each sequence unit;
Figure BDA0002192222980000197
the power predicted value of the wind power plant j is obtained;
Figure BDA0002192222980000198
taking the lower limit of the total active power set for avoiding the frequent start and stop of the unit in each sequence
Figure BDA0002192222980000199
P demand And the active power reference value of the wind power plant j is obtained.
The allocation rule based on the prioritization method is shown in fig. 2.
The invention relates to a fatigue load-based optimal distribution layer of active power of a unit layer, which mainly considers the torque T of a low-speed shaft of a unit s And tower bending moment M t To reduce both fluctuations (i.e. to reduce Δ T) s And Δ M t ) The corresponding Damage Equivalent Load (DEL) can be effectively reduced, thereby reducing fatigue loads. The optimization method mainly considers the reduction of the integral delta M for the first type unit (PSVB) t The load generated by the bending moment of the tower of the unit is reduced, and the whole delta T of the second type unit (PBVS) is mainly considered to be reduced s The load generated by the torque of the low-speed shaft of the unit is reduced, and the delta T of the third type unit (PBVB) s And Δ M t Meanwhile, the fourth type unit (PSVS) does not need an optimization algorithm to distribute power.
The main calculation formula of the unit fatigue load calculation model is as follows:
by making a single-state machine set model x (k) = [ Delta omega ] g ,Δβ,Δω f ] T And establishing a state space equation and discretizing to obtain the value of the x (k + 1) time.
Figure BDA0002192222980000201
Figure BDA0002192222980000202
Figure BDA0002192222980000203
Figure BDA0002192222980000204
In the formula: omega g Is the generator speed; β is a defined function related to θ; omega f Filtering the rotation speed of the generator; eta g The gear ratio of the gear box is adopted (the direct drive unit is 1); j is a unit of t Is equivalent inertia; t is a unit of a Is the aerodynamic torque; p g The actual generating capacity of the unit is obtained; k is p 、K i Is a gain constant; tau is f Is the filter time constant; omega g-rated The rated rotating speed of the generator. Wherein the addition of 0 subscript indicates the value of the sample at the current time of the respective parameter, e.g. T a0 Is the aerodynamic torque at time k.
Figure BDA0002192222980000211
Figure BDA0002192222980000212
Figure BDA0002192222980000213
Figure BDA0002192222980000214
Through the shaft motion equation of the unit, the delta T can be obtained s Such as the formula:
Figure BDA0002192222980000215
in the formula: j. the design is a square g Is the rotational inertia of the generator; j. the design is a square r Is the rotor moment of inertia; t is g Is the generator torque.
Conversion to:
Figure BDA0002192222980000216
namely:
Figure BDA0002192222980000217
in the formula:
Figure BDA0002192222980000218
Figure BDA0002192222980000219
tower drum bending moment M t The calculation can be simplified by the formula:
M T ≈H*F T
in the formula: h is the height of the tower; f T Thrust for tower
Visible M t Mainly with F T It is related. Δ F T Is calculated as the formula:
ΔF T (k)＝C Ft x(k)
Figure BDA00021922229800002110
in the formula:
Figure BDA00021922229800002111
based on quadratic programming pairOptimizing and distributing a unit layer, wherein the optimization target of the unit layer is to reduce the integral fatigue load of the unit under the condition of tracking output, and the optimization target is to reduce the integral fatigue load of the unit according to the listed delta T s (k + 1) and Δ M t (k + 1) (abbreviated as "Δ T s "and" Δ M t ") computational model, performing different target optimization for each type of unit.
The objective function is as follows:
Figure BDA0002192222980000221
wherein: n is the number of various units, a and b are respectively the optimized coefficients of the torque of the transmission chain and the bending moment load of the tower barrel, and different values of constraint conditions are taken according to different unit classifications:
1. the output limit of the unit is restricted, in order to avoid the shutdown of the unit, the lower output limit of the unit is set to be 5% of the output of the unit, such as a formula:
Figure BDA0002192222980000222
in the formula (I), the compound is shown in the specification,
Figure BDA0002192222980000223
is the available active power of the unit number i,
Figure BDA0002192222980000224
active reference value allocated for No. i unit
2. Force tracking constraints, such as the formula:
Figure BDA0002192222980000225
Figure BDA0002192222980000226
in the formula (I), the compound is shown in the specification,
Figure BDA0002192222980000227
the active power value required by the unit layer.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
The advantages of the present invention are illustrated by comparing Damage Equivalent Load (DEL) results of the conventional proportional distribution method (PD) and the optimization method (OPT) of the present invention.
The first type of unit mainly considers reducing the integral delta M t The load generated by the bending moment of the tower barrel of the unit is reduced, and the optimization result is shown in the table 1.
TABLE 1 first class of unit load optimization results
Table1 Fatigue load optimization results of the first kind of turbines
Figure BDA0002192222980000231
Figure BDA0002192222980000241
As can be seen from Table 1, M is reduced in the main consideration t Under the condition of generated fatigue load, the tower bending moment DELs after the optimization method is adopted is reduced by 21.88% compared with the traditional method.
The second type of unit mainly considers the reduction of integral delta T s The load generated by the torque of the low-speed shaft of the unit is reduced. The optimization results are shown in table 2.
TABLE 2 second class of unit load optimization results
Table2 Fatigue load optimization results of the second kind of turbines
Figure BDA0002192222980000242
Figure BDA0002192222980000251
As can be seen from Table 2, the reduction of T is a major consideration s Under the generated fatigue load condition, the low-speed shaft torque DELs after the optimization method is reduced by 11.72 percent compared with the traditional method.
Third type unit delta T s And Δ M t Also, the results are shown in Table 3.
TABLE 3 load optimization results for the third type of unit
Table3 Fatigue load optimization results of the third kind of turbines
Figure BDA0002192222980000252
Figure BDA0002192222980000261
As can be seen from Table 3, while considering T s And M t Under the condition of generated fatigue load, the low-speed shaft torque DELs after the optimization method is reduced by 8.18 percent compared with the traditional method, and the optimization method is adoptedThe tower bending moment DELs after the method is reduced by 4.84% compared with that of the traditional method.
The results of the class iv crew DELs are shown in Table 4.
TABLE 4 optimization results of the fourth type of unit load
Table4 Fatigue load optimization results of the fourth kind of turbines
Figure BDA0002192222980000262
TABLE 5 Total load cases for two different distribution modes
Table5 Total fatigue load under two different modes of distribution
Figure BDA0002192222980000263
Figure BDA0002192222980000271
As can be seen from Table 5, although the first type of unit T is shown in Table 1 s Resulting fatigue loads and the second type of units M in Table 2 t The generated fatigue load is increased compared with the traditional method, but for the wind farm overall, the low-speed shaft torque DELs of the wind farm after the optimization method is adopted is still reduced by 0.63% in total, and the tower bending moment DELs after the optimization method is adopted is reduced by 12.98% compared with the traditional method.

Claims (9)

1. A wind power plant active power distribution method considering unit fatigue loads is characterized in that: the method performs power allocation through three levels:
the method comprises the following specific steps:
1) The first layer is a wind power cluster distribution layer, the wind power cluster distribution layer calculates the power generation capacity by using the wind speed prediction information of each wind power plant, and the cluster active reference value is distributed to the wind power plant distribution layer according to the proportion of the power generation capacity value calculated by each wind power plant;
2) The second layer is a wind power plant distribution layer, and is used for distributing the cluster active reference values received in the first layer to different groups by adopting a priority sorting method according to the result of cluster analysis;
3) The third layer is a unit distribution layer, and active reference values received by various units are distributed to the units;
the method for calculating the power which can be generated by utilizing the wind speed prediction information of the wind power plant in the first layer comprises the following steps:
firstly, decomposing a wind speed sequence into a series of relatively stable components by using EMD;
then, according to different component frequencies, an RBF kernel function is selected as a kernel function for predicting high-frequency components by the SVM, and according to k (x, y) = exp (— | | x-y) | circuitry 2 /σ 2 ) Determining a related parameter sigma, wherein x and y are input values of n dimensions, and sigma is a kernel parameter; according to the selection of a polynomial kernel function as a kernel function of SVM prediction of the intermediate frequency component, according to k (x, y) = [ (alpha x.y) + c] d Determining related parameters c and d, wherein alpha is a slope, x and y are input values of n dimensions, c is a constant term, and d is polynomial degree; selecting a linear kernel function as a kernel function of the SVM for predicting the low-frequency component, and determining a related parameter c according to k (x, y) = x.y + c;
the prediction model is obtained by the formula:
Figure FDA0003958557210000011
in the formula: r reg Is a structural risk function; | w | non-woven phosphor 2 Is a described kernel function; f (x) is the model complexity; c is a constant; r emp Is a risk of experience
Then inputting prediction data and processing the generated prediction result;
and finally, superposing the wind speed prediction results and inputting the wind speed prediction results into a power conversion curve to obtain a wind power prediction result.
2. The active power distribution method of the wind farm considering the fatigue load of the unit according to claim 1, characterized in that:
the method for calculating the maximum power in the first layer is as follows:
firstly, calculating the power which can be generated by a single unit according to the predicted wind speed, wherein the formula is as follows:
Figure FDA0003958557210000021
in the formula, P rated Is the rated power; ρ is the air density; v. of nac The predicted value of the wind speed of the engine room is taken;
Figure FDA0003958557210000022
is the power that can be generated by the unit; c p Is a wind energy conversion rate value; r is the radius of the wind wheel;
the maximum power generation of the wind power plant is the sum of the power generation of all the units in the wind power plant, and the calculation is as the following formula:
Figure FDA0003958557210000023
in the formula, P WFjavai The power can be generated for the jth wind power plant; and n is the number of the units in the wind power plant.
3. The wind farm active power distribution method considering the unit fatigue load according to claim 1, characterized in that: the method for distributing the cluster active reference values to the wind power plant distribution layer in proportion in the first layer comprises the following steps:
the wind power cluster active power distribution method is characterized in that active power dispatching instructions are distributed to each wind power plant in proportion according to the power which can be generated by the wind power plants, so that the purpose of fair distribution is achieved; such as the formula:
Figure FDA0003958557210000024
in the formula:
Figure FDA0003958557210000025
an active power value distributed for a jth wind power plant;
Figure FDA0003958557210000026
and m is the number of wind power plants.
4. The active power distribution method of the wind farm considering the fatigue load of the unit according to claim 1, characterized in that: in the second layer, the wind farm distribution layer is according to the formula:
Figure FDA0003958557210000027
in the formula: j. the design is a square g Is the rotational inertia of the generator; j. the design is a square r Is the rotor moment of inertia; t is g Is the generator torque; delta T s Low speed shaft torque; delta T a Is an aerodynamic torque; eta g Is the gear box speed ratio; j. the design is a square t Is equivalent inertia
M T ≈H*F T
In the formula: h is the height of the tower; f T Thrust of the tower barrel is obtained; m T Bending moment of the tower barrel;
and clustering the wind turbines by adopting a fuzzy c-means clustering method for the fatigue loads generated by the low-speed shaft torque and the tower drum bending moment obtained by the formula, and distributing the power reference value to the turbine layer by adopting a priority sorting method.
5. The active power distribution method of the wind power plant considering the fatigue load of the unit according to claim 4, characterized in that:
in the second layer: the active power priority ordering distribution method of the wind power plant layer based on the fuzzy c-means clustering comprises the following steps:
the range of the optimal cluster number is set as:
Figure FDA0003958557210000031
(wherein n is the number of units); the initial clustering center merging method comprises the following steps: will be provided with
Figure FDA0003958557210000032
The distance between every two initial clustering centers is calculated, and the arithmetic mean value is calculated for the two clustering centers with the shortest distance to obtain
Figure FDA0003958557210000033
An initial clustering center;
extracting an average value of the output power of the unit; extracting a wind speed root-mean-square difference value as a characteristic value, and then obtaining a characteristic matrix of wind power and wind speed of the unit as follows:
Figure FDA0003958557210000034
in the matrix, P me(0-1) (i) Average value v after normalization of active power of ith unit RM(0-1) (i) And normalizing the root mean square difference value after the wind speed of the ith unit is normalized.
6. The active power distribution method of the wind farm considering the fatigue load of the unit according to claim 5, characterized in that:
the characteristic matrix is used as input of a fuzzy c-means algorithm, the active power mean value 1MV and the wind speed root-mean-square difference value 1.2 are used as standards, the units in the wind power plant are clustered, and the clusters are divided into four types, wherein the first type is that the active power mean value is small and the wind speed root-mean-square difference value is large (PSVB); the second type is that the average value of active power is large, and the root mean square difference value of wind speed is small (PBVS); the third category is that the average value of active power is large, and the root mean square difference value of wind speed is large (PBVB); the fourth type is that the mean value of active power is small, and the root mean square difference value of wind speed is small (PSVS);
the method comprises the steps that power of a wind power plant layer is distributed based on a priority method, power is distributed to a fourth type unit and a second type unit which are small in wind speed root-mean-square difference, and residual power is distributed to the first type unit and the third type unit;
1) The arrangement sequence of the distributed active power reference values: sequence 1 is a fourth type and a second type of unit,
Figure FDA0003958557210000035
the active reference value assigned for the fourth class,
Figure FDA0003958557210000036
active reference values allocated for the second class; sequence 2 is a first type and a third type of units,
Figure FDA0003958557210000037
the active reference value assigned for the first class,
Figure FDA0003958557210000038
active reference values allocated for the third class;
2)
Figure FDA0003958557210000039
active power can be generated for each sequence unit;
Figure FDA00039585572100000310
a power predicted value of the wind farm j; p mini (i =1, 2) taking P as the lower limit of the total active power set for avoiding frequent start-stop of the unit in each sequence mini ＝10％P forfj ；P demand And the active power reference value of the wind power plant j is obtained.
7. The wind farm active power distribution method considering the unit fatigue load according to claim 1, characterized in that: in the third layer, the unit layer uses the minimum fatigue load of different units as a target, calculates the power reference value of the unit in real time by using the existing quadratic programming algorithm and distributes the power reference value to each unit so as to minimize the overall fatigue load of the unit;
the optimal distribution of the unit layer is based on the quadratic programming, so that the overall fatigue load of the unit can be effectively reduced;
the objective function is as follows:
Figure FDA0003958557210000041
wherein: n is the number of various units, a and b are respectively the optimized coefficients of the torque of the transmission chain and the bending moment load of the tower barrel, and different values are taken according to the unit classification; delta T s 2 For low-speed torque fluctuations, Δ M, of the unit t 2 The tower bending moment fluctuates.
8. The active power distribution method of the wind farm considering the fatigue load of the unit according to claim 7, characterized in that:
the optimal distribution constraint conditions of the unit layer based on quadratic programming are as follows:
1) The unit output limit is restricted, in order to avoid the unit halt, the lower limit of the unit output is set to be 5% of the output, as shown in the formula (2)
Figure FDA0003958557210000042
In the formula (I), the compound is shown in the specification,
Figure FDA0003958557210000043
active power available for the unit i, P refWT-i Active reference value allocated for No. i unit
2) Force tracking constraints, e.g. equations (3) and (4)
Figure FDA0003958557210000044
Figure FDA0003958557210000045
In the formula (I), the compound is shown in the specification,
Figure FDA0003958557210000046
and the active power value required by the unit layer.
9. The utility model provides a consider wind-powered electricity generation field active power distribution system of unit fatigue load which characterized in that: the system comprises a wind power cluster distribution module, a wind power field distribution module and a unit distribution module;
the wind power cluster distribution module calculates the power generation capacity by using the wind speed prediction information of each wind power plant, and distributes the cluster active reference value to the wind power plant distribution module according to the proportion of the power generation capacity values calculated by each wind power plant;
the wind power plant distribution module distributes the cluster active reference values received in the wind power cluster distribution module to different groups by adopting a priority method according to the result of cluster analysis;
the unit distribution module distributes the active reference values received by the various units to the units;
the specific allocation method performs power allocation through three levels:
the method comprises the following specific steps:
1) The first layer is a wind power cluster distribution layer, the wind power cluster distribution layer calculates the power generation capacity by using the wind speed prediction information of each wind power plant, and the cluster active reference value is distributed to the wind power plant distribution layer according to the proportion of the power generation capacity value calculated by each wind power plant;
2) The second layer is a wind power plant distribution layer, and is used for distributing the cluster active reference values received in the first layer to different groups by adopting a priority sorting method according to the result of cluster analysis;
3) The third layer is a unit distribution layer, and active reference values received by various units are distributed to the units;
the method for calculating the power which can be generated by utilizing the wind power plant wind speed prediction information in the first layer comprises the following steps:
firstly, decomposing a wind speed sequence into a series of relatively stable components by using EMD;
then selecting RBF kernel function as kernel function of SVM predicting high frequency component according to different component frequencies,calculating the routing according to k (x, y) = exp (- | | x-y) | 2 /σ 2 ) Determining a related parameter sigma, wherein x and y are input values of n dimensions, and sigma is a kernel parameter; according to the kernel function of the selected polynomial kernel function as the SVM for predicting the intermediate frequency component, according to k (x, y) = [ (alpha x.y) + c] d Determining related parameters c and d, wherein alpha is a slope, x and y are input values of n dimensions, c is a constant term, and d is polynomial degree; selecting a linear kernel function as a kernel function of the SVM for predicting the low-frequency component, and determining a related parameter c according to k (x, y) = x.y + c;
the prediction model is obtained by the formula:
Figure FDA0003958557210000051
in the formula: r reg Is a structural risk function; | w | non-woven phosphor 2 Is a described kernel function; f (x) is the model complexity; c is a constant; r emp Is a risk of experience
Then inputting prediction data and processing the generated prediction result;
and finally, superposing the wind speed prediction results and inputting the wind speed prediction results into a power conversion curve to obtain a wind power prediction result.
CN201910836163.6A 2019-09-05 2019-09-05 Wind power plant active power distribution method considering fatigue load of unit Active CN110676872B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910836163.6A CN110676872B (en) 2019-09-05 2019-09-05 Wind power plant active power distribution method considering fatigue load of unit

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910836163.6A CN110676872B (en) 2019-09-05 2019-09-05 Wind power plant active power distribution method considering fatigue load of unit

Publications (2)

Publication Number Publication Date
CN110676872A CN110676872A (en) 2020-01-10
CN110676872B true CN110676872B (en) 2023-01-20

Family

ID=69076003

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910836163.6A Active CN110676872B (en) 2019-09-05 2019-09-05 Wind power plant active power distribution method considering fatigue load of unit

Country Status (1)

Country Link
CN (1) CN110676872B (en)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111262257B (en) * 2020-02-26 2023-05-16 沈阳工业大学 Wind turbine frequency modulation method considering fatigue load of turbine
CN112821471B (en) * 2020-12-31 2022-09-30 哈尔滨工业大学 Auxiliary control method for wind turbine generator set participating in power grid frequency modulation considering fatigue load
CN113078690A (en) * 2021-04-30 2021-07-06 南京河大风电科技有限公司 Wind power plant automatic power generation control system and method considering unit fatigue difference
CN113394813B (en) * 2021-05-31 2023-01-31 南方海上风电联合开发有限公司 Method for calculating unit power instruction value of offshore wind farm and distributed scheduling method
CN114123357B (en) * 2021-11-24 2023-08-15 华能威宁风力发电有限公司 Wind farm AGC power optimization control method
CN116667344B (en) * 2023-07-31 2023-10-10 浙江大学 Self-adaptive local fatigue load balance scheduling method and device for offshore wind farm

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102570471A (en) * 2012-02-27 2012-07-11 山东大学 Method for controlling voltage of wind farm through hierarchical coordination of multiple reactive power sources
CN103545849A (en) * 2013-10-28 2014-01-29 国家电网公司 Framework with wind electricity participating in system frequency modulation control
CN104978608A (en) * 2015-06-24 2015-10-14 云南电网有限责任公司电力科学研究院 Wind power prediction apparatus and prediction method
CN108899930A (en) * 2018-07-09 2018-11-27 西华大学 Wind-powered electricity generation station equivalent modeling method based on Principal Component Analysis Method and hierarchical clustering algorithm
CN109245150A (en) * 2018-08-24 2019-01-18 中国电力科学研究院有限公司 A kind of power distribution method and system for wind-powered electricity generation cluster

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DK201570559A1 (en) * 2015-08-28 2017-03-27 Vestas Wind Sys As Methods and Systems for Generating Wind Turbine Control Schedules
US10928816B2 (en) * 2015-06-30 2021-02-23 Vestas Wind Systems A/S Methods and systems for generating wind turbine control schedules

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102570471A (en) * 2012-02-27 2012-07-11 山东大学 Method for controlling voltage of wind farm through hierarchical coordination of multiple reactive power sources
CN103545849A (en) * 2013-10-28 2014-01-29 国家电网公司 Framework with wind electricity participating in system frequency modulation control
CN104978608A (en) * 2015-06-24 2015-10-14 云南电网有限责任公司电力科学研究院 Wind power prediction apparatus and prediction method
CN108899930A (en) * 2018-07-09 2018-11-27 西华大学 Wind-powered electricity generation station equivalent modeling method based on Principal Component Analysis Method and hierarchical clustering algorithm
CN109245150A (en) * 2018-08-24 2019-01-18 中国电力科学研究院有限公司 A kind of power distribution method and system for wind-powered electricity generation cluster

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
基于经验模态分解和支持向量机的短期风电功率组合预测模型;叶林;《中国电机工程学报》;20111105;第31卷(第31期);全文 *
考虑机组疲劳载荷的风电集群有功功率分配方法;刘颖明;《太阳能学报》;20210531;第42卷(第5期);全文 *

Also Published As

Publication number Publication date
CN110676872A (en) 2020-01-10

Similar Documents

Publication Publication Date Title
CN110676872B (en) Wind power plant active power distribution method considering fatigue load of unit
CN108388973B (en) Virtual power plant self-adaptive robust scheduling optimization method
CN110571867B (en) Day-ahead optimal scheduling system method for virtual power plant considering wind power uncertainty
CN107039977B (en) Robust scheduling uncertainty set construction method for power system
Bitar et al. Coordinated control of a wind turbine array for power maximization
CN107909211B (en) Wind field equivalent modeling and optimization control method based on fuzzy c-means clustering algorithm
Wang et al. Electric vehicle fleet integration in a virtual power plant with large-scale wind power
CN111697578B (en) Multi-target energy storage-containing regional power grid operation control method
CN104810860A (en) Power distribution method and power distribution device in wind power plant
CN113193547A (en) Day-ahead-day cooperative scheduling method and system for power system considering uncertainty of new energy and load interval
CN107482692A (en) The method, apparatus and system of wind power plant real power control
CN108198091A (en) A kind of reactive optimization method of wind power system based on improvement particle cluster algorithm
CN114336785B (en) Distributed power supply group control and group dispatching control method and device based on grid clustering
CN116667389B (en) Novel wind-storage combined prediction correction control method for inertia lifting of power system
CN115409360A (en) Power scheduling method, system, equipment and storage medium for multi-machine type wind power plant
CN115275983A (en) Photovoltaic power fluctuation stabilizing control method based on distributed resource clustering
CN107732984B (en) Power control method for multi-model wind turbine generator mixed wind power plant
CN113112287A (en) Virtual power plant energy management method and device based on micro-grid scheduling
CN110350554B (en) Wind storage system auxiliary power grid primary frequency modulation control method based on series-parallel structure
CN112202202A (en) Wind power plant group coordination control method under multilayer hierarchical structure
CN111523947A (en) Virtual power plant power generation cost generation method
CN116937637A (en) Energy storage capacity distribution robust planning method for providing frequency support through multi-energy complementation
CN113410872B (en) Power optimization distribution algorithm between units in wind power plant
CN106849189B (en) Consider the micro-capacitance sensor economy and method for optimizing stability of renewable energy randomness
CN109787279B (en) Interval quantity solving method for uncertain scheduling problem of wind power grid-connected system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant