CN113323821A - Method for adjusting yaw control parameters of wind turbine model prediction - Google Patents

Method for adjusting yaw control parameters of wind turbine model prediction Download PDF

Info

Publication number
CN113323821A
CN113323821A CN202110653490.5A CN202110653490A CN113323821A CN 113323821 A CN113323821 A CN 113323821A CN 202110653490 A CN202110653490 A CN 202110653490A CN 113323821 A CN113323821 A CN 113323821A
Authority
CN
China
Prior art keywords
yaw
control
representing
follows
membership function
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.)
Granted
Application number
CN202110653490.5A
Other languages
Chinese (zh)
Other versions
CN113323821B (en
Inventor
宋冬然
李子群
杨建�
董密
孙尧
粟梅
杨迎港
刘俊波
涂燕萍
许杉敏
黄朝能
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.)
Central South University
Original Assignee
Central South University
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 Central South University filed Critical Central South University
Priority to CN202110653490.5A priority Critical patent/CN113323821B/en
Publication of CN113323821A publication Critical patent/CN113323821A/en
Application granted granted Critical
Publication of CN113323821B publication Critical patent/CN113323821B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F03—MACHINES OR ENGINES FOR LIQUIDS; WIND, SPRING, OR WEIGHT MOTORS; PRODUCING MECHANICAL POWER OR A REACTIVE PROPULSIVE THRUST, NOT OTHERWISE PROVIDED FOR
    • F03D—WIND MOTORS
    • F03D17/00—Monitoring or testing of wind motors, e.g. diagnostics
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F03—MACHINES OR ENGINES FOR LIQUIDS; WIND, SPRING, OR WEIGHT MOTORS; PRODUCING MECHANICAL POWER OR A REACTIVE PROPULSIVE THRUST, NOT OTHERWISE PROVIDED FOR
    • F03D—WIND MOTORS
    • F03D7/00—Controlling wind motors 
    • F03D7/02—Controlling wind motors  the wind motors having rotation axis substantially parallel to the air flow entering the rotor
    • F03D7/0204—Controlling wind motors  the wind motors having rotation axis substantially parallel to the air flow entering the rotor for orientation in relation to wind direction
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F03—MACHINES OR ENGINES FOR LIQUIDS; WIND, SPRING, OR WEIGHT MOTORS; PRODUCING MECHANICAL POWER OR A REACTIVE PROPULSIVE THRUST, NOT OTHERWISE PROVIDED FOR
    • F03D—WIND MOTORS
    • F03D7/00—Controlling wind motors 
    • F03D7/02—Controlling wind motors  the wind motors having rotation axis substantially parallel to the air flow entering the rotor
    • F03D7/04—Automatic control; Regulation
    • F03D7/042—Automatic control; Regulation by means of an electrical or electronic controller
    • F03D7/043—Automatic control; Regulation by means of an electrical or electronic controller characterised by the type of control logic
    • F03D7/046—Automatic control; Regulation by means of an electrical or electronic controller characterised by the type of control logic with learning or adaptive control, e.g. self-tuning, fuzzy logic or neural network
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F03—MACHINES OR ENGINES FOR LIQUIDS; WIND, SPRING, OR WEIGHT MOTORS; PRODUCING MECHANICAL POWER OR A REACTIVE PROPULSIVE THRUST, NOT OTHERWISE PROVIDED FOR
    • F03D—WIND MOTORS
    • F03D80/00—Details, components or accessories not provided for in groups F03D1/00 - F03D17/00
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Mechanical Engineering (AREA)
  • Sustainable Development (AREA)
  • Sustainable Energy (AREA)
  • Chemical & Material Sciences (AREA)
  • Combustion & Propulsion (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Feedback Control In General (AREA)

Abstract

The invention provides a method for adjusting a model prediction yaw control parameter of a wind turbine, which comprises the following steps: step 1, setting a yaw system of a wind turbine into a yaw mode adopting model predictive control; step 2, determining a yaw error of model prediction control, and establishing a yaw error m-step prediction model according to the yaw error; step 3, determining a control target of model predictive control, and carrying out normalization processing on the control target; step 4, determining a target function of model predictive control according to the control target after normalization processing; and 5, designing a weight coefficient evaluator, and optimizing parameters by adopting a fuzzy rule membership function association optimization strategy and a self-adaptive grid multi-target particle swarm optimization method. The method can dynamically adjust the weight coefficient of the model predictive control objective function according to the wind direction information, so that the wind turbine can more accurately track the wind direction change at a lower yaw rate, and the power capture loss of the wind turbine is reduced while the yaw rate is reduced.

Description

Method for adjusting yaw control parameters of wind turbine model prediction
Technical Field
The invention relates to the technical field of wind power generation, in particular to a method for adjusting a yaw control parameter of a wind turbine model prediction.
Background
With the increasing social demands, wind power generation is continuously developing. According to a 2019 global wind power development report issued by GWEC, the total installed capacity of the global wind power in 2019 reaches 650GW, and a fan is developing towards large-scale and high-capacity. Typical wind turbine control mainly includes pitch control, torque control and yaw control, wherein the related research of the first two accounts for about 80% of the wind turbine control field, and the yaw control only obtains limited attention. In general pitch and torque control research, the wind turbine is generally considered to be in a perfect yaw state, and the influence of the yaw on the wind turbine is not considered. However, with the development of the large-scale fan, problems such as power reduction and load increase caused by misalignment of yaw have not been ignored. According to one study result, the potential power loss due to yaw misalignment is about 2.7%. Meanwhile, according to the inspection and display of the countries such as Denmark, the failure rate of the yaw system accounts for about 12.5% of the total failure rate of the fan. Therefore, there is an urgent need to improve the yaw control performance of the wind turbine.
With the application of a light detection and ranging (LIDAR) technology in the field of wind energy, the defects of the traditional yaw control method are improved, wind speed and wind direction information in front of a fan can be obtained in advance through the LIDAR, and the yaw control effect can be well improved by applying the information to model prediction control. The model prediction control system has the advantages of explicit active processing constraint, good fusion of prediction wind information into the model and the like. However, the problem of model predictive control is that when there are multiple performance indicators, it is common practice to combine multiple performance indicators into one objective function by adding weight coefficients, which are usually set empirically, and the weight coefficients cannot be adjusted according to actual wind direction information, so that there is still room for improvement in power capture.
Fuzzy Logic Control (FLC) is abstract of approximate reasoning characteristics of human decision, and can be regarded as generalization of a rule-based expert system, fuzzy control is developed in multiple fields at present, and the fuzzy logic control is respectively adopted to control a wind power energy storage system, so that the wind power fluctuation is stabilized, and meanwhile, the deep charging and discharging of the system are avoided; the FLC and the fuzzy neural network are adopted to control the pitch of the large-scale fan, and the control effect is more effective than that of the PI. However, the advantages of the FLC are also the disadvantages, and the characteristics of the FLC which depend too much on expert experience cause the artificial set membership functions and fuzzy rules to have great influence on the control performance. Therefore, the optimization algorithm is adopted to optimize the membership function and the fuzzy rule, for example, the genetic algorithm, the ant colony algorithm and the particle swarm algorithm are adopted to optimize the membership function, so that the control effect of the FLC is improved.
Disclosure of Invention
The invention provides a method for adjusting a wind turbine model prediction yaw control parameter, and aims to solve the problems that the traditional parameter adjusting method has defects caused by experience setting, and a weight coefficient cannot be adjusted according to actual wind direction information in actual operation.
In order to achieve the above object, an embodiment of the present invention provides a method for adjusting a wind turbine model prediction yaw control parameter, including:
step 1, setting a yaw system of a wind turbine into a yaw mode adopting model predictive control;
step 2, determining a yaw error of model prediction control, and establishing a yaw error m-step prediction model according to the yaw error;
step 3, determining a control target of model predictive control, and carrying out normalization processing on the control target;
step 4, determining a target function of model predictive control according to the control target after normalization processing;
designing a weight coefficient evaluator, and optimizing parameters of the fuzzy inference weight coefficient evaluator by adopting a fuzzy rule and membership function association optimization strategy and a self-adaptive grid multi-target particle swarm algorithm;
and 6, determining a weight coefficient in the model predictive control objective function through the optimized fuzzy inference weight coefficient evaluator.
The scheme of the invention has the following beneficial effects:
the method for adjusting the wind turbine model prediction yaw control parameters in the embodiment of the invention adopts the fuzzy rule and membership function association optimization strategy to simplify the fuzzy rule and membership function optimization problem of the fuzzy inference weight coefficient evaluator, the optimization performance is ensured while the complexity of the optimization problem is reduced, the fuzzy inference weight coefficient evaluator is optimized by adopting a self-adaptive grid multi-target particle swarm algorithm to obtain a pareto curved surface, the optimal parameters of the fuzzy inference weight coefficient evaluator are decided according to the pareto curved surface, and the optimized fuzzy inference weight coefficient evaluator dynamically adjusts the weight coefficient of the model predictive control objective function according to the wind direction information, so that the wind direction change can be more accurately tracked by the fan at a lower yaw rate, and the power capture loss of the fan is reduced while the yaw rate is reduced.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a control block diagram of the present invention;
FIG. 3 is a graphical illustration of a limited set of yaw rates of the present invention;
FIG. 4 is a flow chart of the fuzzy rule-membership function association optimization strategy of the present invention;
FIG. 5 is a flow chart of an adaptive grid multi-objective particle swarm algorithm of the present invention;
fig. 6 is a schematic diagram of the fuzzy inference weight coefficient evaluator scheme of the present invention.
FIG. 7(a) is a diagram of the fuzzy input variable WD of the fuzzy inference weight coefficient evaluator of the present inventionavA schematic representation of the membership function of (a);
FIG. 7(b) is a diagram of the fuzzy input variable WD of the fuzzy inference weight coefficient evaluator of the present inventionstdA schematic representation of the membership function of (a);
FIG. 7(c) is a schematic diagram of the membership function of the fuzzy output variable ω of the fuzzy inference weight coefficient evaluator of the present invention;
FIG. 8 is a 24h wind diagram of the present invention;
FIG. 9(a) is a schematic diagram of the iterative result of the fuzzy rule-membership function sequential optimization strategy of the present invention;
FIG. 9(b) is a diagram illustrating the iterative result of the fuzzy rule-membership function mixed integer optimization strategy of the present invention;
FIG. 9(c) is a diagram illustrating the iterative result of the fuzzy rule-membership function association optimization strategy according to the present invention;
FIG. 10(a) is a schematic diagram of a pareto front edge obtained by the fuzzy rule-membership function sequential optimization strategy according to the present invention;
FIG. 10(b) is a schematic diagram of a Parritotr frontage obtained by the fuzzy rule-membership function mixed integer optimization strategy according to the present invention;
FIG. 10(c) is a schematic diagram of a pareto front edge obtained by the fuzzy rule-membership function association optimization strategy according to the present invention;
FIG. 11 is a comparison of pareto surfaces under different optimization strategies in accordance with the present invention;
FIG. 12 is a diagram illustrating the optimization results of different prediction step sizes m according to the present invention;
FIG. 13(a)) Fuzzy input variable WD for the optimized fuzzy inference weight coefficient evaluator of the present inventionavA schematic representation of the membership function of (a);
FIG. 13(b) is a diagram of the fuzzy input variable WD of the fuzzy inference weight coefficient evaluator optimized according to the present inventionstdA schematic representation of the membership function of (a);
FIG. 13(c) is a schematic diagram of a membership function of the fuzzy output variable ω of the optimized fuzzy inference weight coefficient evaluator of the present invention;
FIG. 14 is a schematic diagram of an optimized fuzzy regular surface of the present invention;
FIG. 15(a) is a schematic diagram illustrating the comparison of the power capture loss rate under model predictive control of the present invention with the power capture loss rate under normal model predictive control;
fig. 15(b) is a schematic diagram showing a comparison between the yaw rate under model predictive control of the present invention and the yaw rate under normal model predictive control.
Detailed Description
In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following detailed description is made with reference to the accompanying drawings and specific embodiments.
The invention provides a method for adjusting a model prediction yaw control parameter of a wind turbine, aiming at the problems that the existing parameter adjusting method has defects caused by experience setting and the weight coefficient cannot be adjusted according to actual wind direction information in actual operation.
As shown in fig. 1 to 15, an embodiment of the present invention provides a method for adjusting a wind turbine model predicted yaw control parameter, including: step 1, setting a yaw system of a wind turbine into a yaw mode adopting model predictive control; step 2, determining a yaw error of model prediction control, and establishing a yaw error m-step prediction model according to the yaw error; step 3, determining a control target of model predictive control, and carrying out normalization processing on the control target; step 4, determining a target function of model predictive control according to the control target after normalization processing; designing a weight coefficient evaluator, and optimizing parameters of the fuzzy inference weight coefficient evaluator by adopting a fuzzy rule and membership function association optimization strategy and a self-adaptive grid multi-target particle swarm algorithm; and 6, determining the weight coefficient in the model predictive control objective function through the optimized fuzzy reasoning weight coefficient evaluator.
In the method for adjusting the wind turbine model prediction yaw control parameter according to the above embodiment of the present invention, as shown in fig. 2, the yaw mode of the wind turbine is set to be the fixed rate yaw mode, the yaw rate is 0.5deg/s, a constraint is set on the change of the yaw rate in the yaw process, and the relevant settings of the model prediction control include: the method comprises the steps of (1) carrying out constraint, a limited control set, multiple targets and variable weight; the membership function and the fuzzy rule of the fuzzy inference weight coefficient evaluator are optimized by adopting a self-adaptive grid multi-target particle swarm algorithm, the iteration number of the self-adaptive grid multi-target particle swarm algorithm is set to be 50 generations, the population scale is 100, the optimization dimension is 13 dimensions, and 13-dimensional optimization variables are set as follows: with WDstdMembership functions of (a): the variables to be optimized comprise the bottom points on the right side of the leftmost triangular membership function and the bottom points on the left side of the rightmost triangular membership function, and the vertexes of the middle three triangular membership functions comprise five variables to be optimized, wherein the shapes of the middle three triangular membership functions are not changed and translate in a membership space along with the vertexes, and similarly, WD (weighted distribution) dataavThe membership function of (1) comprises 3 variables to be optimized; the omega membership function comprises 5 variables to be optimized, the final result of optimization is to form a pareto curved surface for selecting proper fuzzy inference weight coefficient evaluator parameters, the real wind direction information sampled every 1s in one day from a certain wind power plant is adopted, the sampling period of model predictive control is set to be 1s, the control period is set to be 60s, and the value range of the predicted step length is [1,6 ]]Setting input WD of fuzzy inference weight coefficient evaluatoravIs a membership function space of [0,15 ]]，WDavThe initial positions of the three membership functions are uniformly distributed in [0,15 ]]Internal, WDstdThe initial positions of the 5 membership functions are uniformly distributed in [0,15 ]]In the internal output omega, the initial positions of 5 membership functions are uniformly distributed at [0,0.1 ]]Selecting proper parameters of a fuzzy inference weight coefficient evaluator on the pareto curved surface; setting parameters of fuzzy inference weight coefficient evaluator and relevant parameters of model predictive controlAfter counting, the method for adjusting the wind turbine model prediction yaw control parameter by using the actual wind direction information is tested, and the method can be specifically described as follows: assuming that the current control period is the kth control period, WD is calculated according to the current position of the fan cabin and wind direction information in the k to k + m control periods in the futureavAnd WDstd(ii) a WD to be calculatedavAnd WDstdAs the input of a fuzzy inference weight coefficient evaluator, obtaining an output weight parameter omega (k) through fuzzy inference; taking the weight parameter omega (k) as one input of model predictive control, and combining set other model predictive control parameters and wind direction information to obtain a control sequence [ u (k), u (k +1), … u (k + m-1) through calculation]And using the first quantity u (k) as an input to the fan to control the fan yaw, and repeating the process in the next control cycle. WT in FIG. 2 represents a fan, and different weight coefficients are determined according to system states in each control period to obtain an optimal objective function, so that the performance of model predictive control can be improved.
Wherein, the step 1 specifically comprises: the yaw system of the wind turbine is a rigid yaw model, the rigid yaw model adopts a yaw control mode of model predictive control, the control quantity of the yaw system is a yaw rate, and the yaw rate has three possibilities: -0.5deg/s, 0deg/s and 0.5deg/s, generalizing the problem of yaw rate of a wind turbine yaw system into a three-element control set, as follows:
Figure BDA0003112783200000061
wherein,
Figure BDA0003112783200000062
the Yaw rate of a Yaw system of the wind turbine is represented, k represents a time, j represents the number of different Yaw conditions, the Yaw rate is designated by j, Yaw _ speed represents a Yaw rate value, Yaw _ speed is 0.5deg/s, in actual operation, the Yaw rate of the current time is influenced by the Yaw rate of the previous time, and the influence of the Yaw rate of the previous time is represented as:
Figure BDA0003112783200000063
wherein,
Figure BDA0003112783200000064
indicating the yaw rate at time k,
Figure BDA0003112783200000065
representing the yaw rate at time k + 1.
According to the method for adjusting the wind turbine model prediction yaw control parameters, disclosed by the embodiment of the invention, a yaw system of a modern large horizontal axis wind turbine can be divided into three different models: rigid yaw model, flexible yaw model and controlled yaw model, because of the limitation of equation (2), the wind turbine yaw system becomes a finite control set problem, and in a certain determined state of the current time, the control law of the system at the next time is an element in a finite set, for example, considering that the current time is 0 time, the yaw rate of the system at the next 1,2 and 3 times can be expressed as: as can be seen from fig. 3, when the yaw rate at time 0 is 0, there are three possible values for the yaw rate at time 1. Similarly, the control sequences of the system at the time 2 and the time 3 have 7 possible values and 17 possible values respectively, so that the yaw rate sequence of the wind turbine yaw system can be traversed in a limited time. This yaw motion mode provides the basis for a finite set model predictive control.
Wherein, the step 2 specifically comprises: model predictive control of yaw error theta at a time tyeAs follows:
Figure BDA0003112783200000066
wherein, thetawd(t) represents the wind direction at time t, θnp(t) denotes the nacelle position at time t, θnp(t1) Represents t1The cabin position at the moment;
an m-step prediction model of the yaw error is obtained according to equation (3) as follows:
Figure BDA0003112783200000071
wherein, thetanp(k + m | k) represents the nacelle position at time k + m,
Figure BDA0003112783200000072
representing the yaw rate, T, between time k and time k + isDenotes a control period, and m denotes a prediction step size.
Wherein, the step 3 specifically comprises:
considering yaw error, predicting power capture E of fan in m stepscapAs follows:
Figure BDA0003112783200000073
where ρ represents the air density, Ar＝πR2Representing swept area, C, of the fan impeller surfacepRepresenting the aerodynamic power coefficient, V0Representing the equivalent wind speed at the hub;
time t of operation of yaw actuator within m-step prediction rangeyawAs follows:
Figure BDA0003112783200000074
wherein,
Figure BDA0003112783200000075
is a logic value which is 1 when the yaw rate is not 0 and is 0 when the yaw rate is 0; equations (5) and (6) are two control targets for yaw control.
Wherein, the step 3 further comprises: capturing the power of the fan under the m-step prediction EcapAnd predicting the action time t of the yaw actuator within the range of m stepsyawAs model predictionControl target of control, taking into account m-step prediction of power capture E of fancapAnd the action time t of the yaw actuator in the m-step prediction rangeyawThe difference of dimension exists, and the power capture E of the fan under the m-step prediction is respectively carried outcapAnd the action time t of the yaw actuator in the m-step prediction rangeyawNormalization processing was performed as follows:
Figure BDA0003112783200000076
where ξ represents the power capture loss rate due to yaw error, EidealRepresenting the power capture of the fan in an ideal case;
Figure BDA0003112783200000077
where ζ represents the rate of use of the yaw actuator, ttolRepresenting the total time the fan is running.
Wherein, the step 4 specifically comprises: an objective function QF of the model predictive control is obtained from the control targets ξ and ζ of the model predictive control, as follows:
Figure BDA0003112783200000081
where ω represents a weight coefficient for balancing between the power capture loss rate and the usage rate of the yaw actuators, and the weight coefficient ω has a value in the range of [0,1 ].
In the method for adjusting the wind turbine model prediction yaw control parameter according to the embodiment of the invention, the main aim of the yaw control problem of the large-scale wind turbine is to reduce the yaw error thetayeThe primary goal of yaw control is to improve the power capture of the wind turbine, another goal of yaw control is the actuation time of the yaw actuator, and there are three states for the yaw controller: positive action, negative action, and no action.
Wherein, the step 5 specifically comprises: taking the weighted average of the difference between the position of the cabin in the prediction step and the predicted wind direction and the root mean square of the weighted average of the difference between the position of the cabin in the prediction step and the predicted wind direction as the input of a fuzzy inference weight coefficient evaluator;
in the fuzzy inference weight coefficient evaluator, the region center calculation inference result obtains clear output by using a gravity center method, as follows:
Figure BDA0003112783200000082
wherein y represents a clear output, RiExpressing the inference result of the ith rule, XiThe output corresponding to the ith rule is shown, and n is the number of the rules;
weighted average WD of the difference between the nacelle position and the predicted wind direction within the prediction stepavThe calculation of (a) is as follows:
Figure BDA0003112783200000083
wherein WDavRepresents a weighted average of the difference between the nacelle position and the predicted wind direction within the prediction step, will WDavAs an input to a fuzzy inference weight coefficient evaluator; WD (k + i | k) represents the predicted yaw error at k + i time at k time;
root mean square WD of weighted average of differences between nacelle position and predicted wind direction within predicted step sizestdThe calculation of (a) is as follows:
Figure BDA0003112783200000091
wherein WDstdMean square root representing a weighted average of the difference between the nacelle position and the predicted wind direction within the prediction step, will WDstdAs another input to the fuzzy inference weight coefficient evaluator;
from the calculated input WDavAnd WDstdFuzzy inference weight coefficient evaluationThe weight coefficient ω is output from the multiplier.
In the wind turbine model predictive yaw control parameter adjusting method according to the above embodiment of the present invention, the fuzzy logic inference is a method that incorporates heuristic rules reflecting human expert experience into the control process, so that the selection of the weight coefficients becomes scientific and simple, the adopted fuzzy inference weight coefficient estimator (FDFE) is used to determine the weight coefficients in the model predictive control objective function, and the proposed scheme of the fuzzy inference weight coefficient estimator for weight coefficient estimation is shown in fig. 6. In the fuzzy inference weight coefficient estimator, a Mamdani type minimum inference method is adopted to be matched with a region Center (COA) defuzzification process for use so as to generate clear inference output. In FIG. 6, WDavAnd WDavAll represent the input of the fuzzy inference weight coefficient estimator, and are respectively the weighted average of the difference between the cabin position and the predicted wind direction in the predicted step size and the root mean square thereof. Unlike the conventional two-input fuzzy inference, the input of the fuzzy inference weight coefficient estimator does not adopt the mode of error and derivative of the error, because the yaw error directly affects the action of model predictive control, and the input mode of predicting the yaw error can very intuitively reflect the relation between the input and the output. From equation (11), it can be found that as the prediction step size increases, the weight of the predicted yaw error of the step decreases gradually, and the influence of the yaw error closer to the current time on the model predictive control operation in the predicted yaw error sequence increases.
Wherein, the step 5 further comprises: the optimization problem of the membership function and the fuzzy rule of the fuzzy inference weight coefficient evaluator is as follows:
min ferror(xmembership，xrule)
Figure BDA0003112783200000092
wherein f iserrorRepresenting an objective function, which is an error between the inference result and an ideal optimal value; x is the number ofmembershipRepresenting the optimal vector membership function, xruleRepresents the optimized vector fuzzy rule, ΩmFeasible domain, omega, representing the membership function of the optimization vectorrA feasible field representing an optimized vector fuzzy rule;
in the optimization of vector xmembershipIn the method, the adjusted parameters are divided into three types of quantity, shape and position distribution, the quantity m of input and output and the value range thereof, and the quantity num of membership functions are presetmfAs follows:
nummf＝[num1，num2，…，numm] (14)
wherein, nummfNumber of membership functions, nummThe number of membership functions contained for each input or output;
the types of membership functions are 8, and the type of each membership function is restricted
Figure BDA0003112783200000101
As follows:
Figure BDA0003112783200000102
wherein,
Figure BDA0003112783200000103
indicates num contained in the nth input or outputnA type vector of each membership function, wherein the value range of elements in the vector is [1,2, …,8 ]]Respectively corresponding to 8 membership function types;
the determination of the location and shape of each membership function is related to the type of each membership function, the location of each membership function is constrained as follows:
Figure BDA0003112783200000104
wherein,
Figure BDA0003112783200000105
indicates the nth inputNum contained in or outnThe location and shape of the kth of the membership functions;
assuming that the type of membership function is a triangular membership function, fpositionThe concrete expression is as follows:
Figure BDA0003112783200000106
wherein f ispositionRepresenting the position and shape functions of the membership functions; according to the formula (17), the position and the shape of the triangular membership function are determined by three parameters, and the three parameters are constrained by upper and lower limits of corresponding input or output;
the fuzzy rule is influenced by the number of input and output, the back piece of the fuzzy rule of fuzzy inference adopting the Mamdani model is a certain fuzzy set of output quantity, if s inputs exist in the fuzzy inference, the feasible domain constrained by the fuzzy rule is shown as follows:
Figure BDA0003112783200000111
wherein omegarSet of all fuzzy rules representing fuzzy inference of s-input, m-s-output.
Wherein, the step 5 further comprises: simplifying the optimization problem of the fuzzy rule and the membership function of the fuzzy inference weight coefficient evaluator by adopting a fuzzy rule-membership function association optimization strategy:
according to the formula (13) -formula (18), determining the input and output quantity and the language value quantity, and further determining the quantity num of the membership functionsmfAnd type constraint of each membership function
Figure BDA0003112783200000112
Assuming that the position of each membership function is uniquely determined by a certain vertex of the triangle, the optimization dimensionality of the membership function is reduced, and for the fuzzy rule, the order constraint of the language value of the output membership function is not considered, and the association optimization is carried out according to the optimization process of the membership functionAnd (3) conversion:
the optimization of the membership functions results in a change in the sequence of output linguistic values as follows:
B＝A·S (19)
wherein, A represents the original sequence, S represents the elementary matrix after elementary transformation, called transformation matrix;
the transformation of the fuzzy rule includes two methods: the first method transforms according to the S matrix, as shown below:
NEWrules[3×5]＝OLDrules[3×5]·S[5×5] (20)
the second method processes the values in the fuzzy rule matrix according to the variation of the linguistic values, which are assumed to be the [ seml VS VL ] sequence, as follows:
[S M L VS VL]→[VS(S) S(M) M(L) L(VS) VL(VL)] (21)
each linguistic value in the fuzzy rule matrix is processed accordingly with a certain probability according to equation (21).
According to the method for adjusting the wind turbine model prediction yaw control parameters, the input and output membership functions have sequential constraint, the membership function with smaller semantics needs to be located in front of the membership function with larger semantics, and the constraint is used for avoiding repeated search on one hand and ensuring the logic accuracy of fuzzy reasoning on the other hand. However, the linguistic value in the fuzzy rule is the linguistic value of the output membership function, if the output membership function is not restrained any more, and the current linguistic value sequence is taken as a reference for optimizing the fuzzy rule, and the fuzzy rule and the membership function are optimized in a correlation mode, so that the problem complexity can be greatly reduced, and the searching capability can be ensured. For example, the S language value is changed to VS with a small probability p, the probabilistic processing can improve the search capability on one hand and can also ensure that the optimal membership function is searched as much as possible under the current fuzzy rule, the fuzzy rule-membership function association optimization strategy is shown in fig. 4, the fuzzy rule-membership function association optimization strategy associates the membership function with the fuzzy rule to optimize, the optimization solving capability is improved as much as possible under the condition of reducing the complexity, in a certain iteration, the output semantic sequence is transformed into [ S M LVS VL ], and then the semantic sequence can be expressed as:
Figure BDA0003112783200000121
the method for adjusting the wind turbine model prediction yaw control parameter and the WD of the fuzzy inference weight coefficient estimator in the embodiments of the inventionav、WDstdInitial membership functions of the sum weight coefficients ω are shown in FIG. 7, WDavAnd WDstdAre all selected as symmetrical triangles, WD, with equal base and 50% overlap with adjacent membership functionsavAnd WDstdAll the membership functions of (1) are discussed in the range of [0deg, 15deg ]]Above definition, WDavThe membership function of (a) is: s, M and L, respectively, representing: small, medium and large; WDstdThe membership function of (a) is: VS, S, M, L and VL, respectively: very small, medium, large, and very large; the optimal value range of the weight coefficient omega is [0,0.1 ]]Therefore, the membership functions of the weight coefficients ω are discussed in the normalized domain [0,0.1 ]]As defined above, the membership function of the weight coefficient ω is: VS, S, M, L and VL, respectively: very small, medium, large, and very large; in combination with expert experience, the initial fuzzy rule table of the fuzzy inference weight coefficient evaluator includes 15 different rules, as shown in table 1:
TABLE 1 fuzzy rule Table
Figure BDA0003112783200000122
The fuzzy inference device adopts two inputs and one output, the inputs of the fuzzy inference device respectively have 3 and 5 logic linguistic variables, the outputs of the fuzzy inference device have 5 logic linguistic variables, and the types of the membership functions adopt sensitive and simple triangular membership functions.
The method for adjusting the wind turbine model prediction yaw control parameter according to the embodiment of the invention discusses two other optimization strategies: evaluation of fuzzy inference weight coefficient by fuzzy rule-membership function sequential optimization strategyAnd (3) optimizing by an estimator: 1. setting a default membership function and setting optimization parameters; 2. optimizing a fuzzy rule; 3. selecting a proper fuzzy rule; 4. optimizing the membership function; and in order to simplify the problem, when the membership functions are optimized, only the position optimization of the membership functions in a feasible domain is considered, and the shapes of the membership functions are not changed. The fuzzy rule-membership function sequence optimization strategy has the advantages of simplicity and convenience, and the optimization problem becomes very simple by optimizing the membership function and the fuzzy rule separately, but the defects are obvious: because the membership function is fixed in the first step of optimization, the optimized fuzzy rule can only be the optimal rule under the membership function, and the solution obtained in the second step of membership function optimization is probably not optimal or even probably negative optimization. And (3) optimizing the fuzzy inference weight coefficient evaluator by adopting a fuzzy rule-membership function mixed integer optimization strategy: simultaneously optimizing the fuzzy rule and the membership function; 1. combining the fuzzy rule with expert experience to obtain a fuzzy rule set, and simplifying a membership function; 2. and carrying out integral constraint on the fuzzy rule set, and carrying out low-dimensional constraint on the simplified membership function. The complexity of the fuzzy rule constraint can be greatly reduced by combining expert experience. Specifically, the method comprises the following steps: when WDavIs VS, WDstdAlso at VS, the output of the fuzzy inference weight coefficient evaluator should be VB; when WDavIs VB, WDstdWhen it is VS, the output of the fuzzy inference weight coefficient evaluator should be VS; when WDavIs B, WDstdAnd S, the output of the fuzzy inference weight coefficient evaluator should be one of VS, S and M. After expert experience is considered, the number of fuzzy rules is acceptable, the fuzzy rules and the membership functions are taken as an integer constraint to be considered in optimization of the membership functions, a fuzzy rule-membership function mixed integer optimization strategy is formed, the fuzzy rule-membership function mixed integer optimization strategy has acceptable complexity, but the expert experience limits the capability of optimization search, and negative optimization can be caused when the expert experience is wrong.
Wherein the steps5 further comprising: searching and solving the simplified optimization problem by adopting a self-adaptive grid multi-target particle swarm algorithm: the self-adaptive grid multi-target particle swarm algorithm comprises the following steps: the self-adaptive grid algorithm, the selection of global optimal particles and the Archive set truncation technology assume a two-dimensional target space optimization problem, and the self-adaptive grid algorithm calculates the search range of a target space after k iterations
Figure BDA0003112783200000131
And
Figure BDA0003112783200000132
the modulus of the grid is computed as follows:
Figure BDA0003112783200000133
Figure BDA0003112783200000134
wherein, Delta Y1 KRepresenting the modulus value of the target 1 of the kth iteration,
Figure BDA0003112783200000135
representing the modulus, Y, of the object 2 of the kth iteration1 KRepresents the calculated value of the target 1 for the k-th iteration,
Figure BDA0003112783200000136
representing the calculated value of the target 2 of the kth iteration, M representing the division of the target space into M grids, Y1And Y2Respectively, the values of the fitness function;
and traversing the group to calculate the grid number of each particle, and calculating the grid number of the particle i as follows:
Figure BDA0003112783200000141
wherein,
Figure BDA0003112783200000143
a grid number representing the particle i with respect to the target 1,
Figure BDA0003112783200000144
grid number, Y, representing particle i with respect to target 21 iRepresents the calculated value of the target 1 for the particle i,
Figure BDA0003112783200000145
represents the calculated value of target 2 for particle i, Int represents rounding; obtaining grid density information of each grid according to the grid number of each particle; selecting global optimal particles according to the grid density information of each grid;
global search experience g corresponding to each particle in populationbestAs follows:
Figure BDA0003112783200000146
wherein,
Figure BDA0003112783200000147
represents the corresponding global optimum value of the jth particle in the kth iteration, F (i) represents the number of particles of the grid where the ith particle is located,
Figure BDA0003112783200000148
representing a set of particles in the Archive set that are better than the population particle j,
Figure BDA0003112783200000149
the calculation of (a) is as follows:
Figure BDA00031127832000001410
wherein,
Figure BDA00031127832000001412
representing pareto dominance relationshipsThat is, i is the dominant solution of j, and when there are a plurality of particles satisfying the condition, the Archive set particle having the largest potential is selected as gbestAs follows:
Figure BDA00031127832000001411
wherein, PkRepresenting the particle population at the k iteration, wherein formula (26) represents that the particles with the most dominant population particles are selected from all the particles meeting formula (24), and when the particles with the same potential exist, one particle is randomly selected;
the Archive set truncation technology adopted by the self-adaptive grid multi-target particle swarm algorithm is based on the density information of the grid, and when the density of the grid exceeds a threshold value, the particles with the minimum potential are deleted according to a formula (26).
According to the method for adjusting the wind turbine model prediction yaw control parameters, simulation is carried out based on a Matlab/Simulink platform, firstly, three different optimization strategies are operated and debugged on the Matlab platform, and the optimization parameters of a fuzzy inference weight coefficient evaluator are obtained; then, simulation is carried out on a Simulink self-built fan module, AGA-MOPSO is a self-adaptive grid multi-target particle swarm algorithm, in order to compare the difference between different control strategies, the parameters of the wind turbine model prediction yaw control parameter adjusting method are all set to be the same value, and the parameters of the wind turbine model prediction yaw control parameter adjusting method are shown in a table 2:
TABLE 2 parameter table of wind turbine model prediction yaw control parameter adjusting method
Figure BDA0003112783200000142
Figure BDA0003112783200000151
The wind direction data adopted when the parameters of the fuzzy inference weight coefficient evaluator are optimized are from actual wind direction data in one day of a certain operating wind power plant, and the sampling time is 1s, as shown in fig. 8.
According to the method for adjusting the wind turbine model prediction yaw control parameters, different optimization strategy results are compared and analyzed: the optimization results of different optimization strategies are discussed by taking the case that the predicted step length m is 6 as an example, and the optimization results when m is 6 are obtained according to the previously set adaptive grid multi-target particle swarm algorithm parameters, as shown in fig. 9. Where the abscissa represents yaw actuator usage and the ordinate represents power capture loss rate. The open circles represent particles on the pareto front surface, while the solid origin represents the distribution of particles throughout the iteration. It can be seen that after 50 iterations, the particles in the three optimization strategies finally converge to the pareto frontier, thus proving the effectiveness of the different optimization strategies and optimization algorithms for the optimization problem.
The method for adjusting the predictive yaw control parameter of the wind turbine model according to the above embodiment of the invention, as shown in fig. 10, the pareto frontage when ω changes from 0.01 to 0.1 in 0.001 steps in a normal MPC and obtained with different optimization strategies is plotted in the case of m-6, it can be seen that as the value of ω increases, the yaw actuator usage rate for the conventional model predictive control decreases and the power capture loss rate increases and also constitutes a pareto front surface, this is because, in the objective function of the model predictive control, the larger ω is, the larger the proportion of the yaw actuator portion is, the larger the use rate of the yaw actuator becomes the main control target, and on the other hand, comparing the two pareto fronts, the fronts optimized by the three optimization strategies are superior to the fronts formed by the common model predictive control, and the method for adjusting the wind turbine model predictive yaw control parameters is proved to be superior to the common model predictive control.
In the method for adjusting the wind turbine model prediction yaw control parameter according to the embodiment of the invention, in fig. 11, the pareto curved surfaces under different optimization strategies are drawn in the same graph, so that it can be found that the pareto curved surface obtained by the fuzzy rule-membership function association optimization strategy is optimal, and the pareto curved surface obtained by the fuzzy rule-membership function sequence optimization strategy is worst.
In the method for adjusting the wind turbine model predicted yaw control parameter according to the above embodiment of the present invention, fig. 12 shows the optimization results of all values of the prediction step length m, since the results of different optimization strategies are similar, the optimization results under the fuzzy rule-membership function association optimization strategy are selected for discussion, and it can be seen from fig. 12 that when m is 1, the prediction step length is shorter, the predicted wind information obtained by the fuzzy inference weight coefficient estimator is less, the actual wind speed and wind direction have larger fluctuation due to the existence of turbulence, and in addition, the yaw system has larger time lag, so that the effect of the method for adjusting the wind turbine model predicted yaw control parameter when m is 1 is not ideal relative to m 2, 3, 4, 5, 6, and when the prediction step length m is continuously increased, the effect of the method for adjusting the wind turbine model predicted yaw control parameter is continuously increased, the method comprises the steps that the predicted wind direction information obtained by a fuzzy inference weight coefficient evaluator is increased along with the increase of m, the wind turbine model starts yawing in a plurality of control cycles to compensate for time lag, the improvement of the performance of the method for adjusting the predicted yaw control parameters of the wind turbine model is continuously reduced along with the increase of m, the most obvious improvement effect is achieved from m-1 to m-2, the pareto curved surfaces of m-5 and m-6 are basically overlapped, and the performance improvement of the method for adjusting the predicted yaw control parameters of the wind turbine model reaches a threshold value when m-6.
In the method for adjusting the wind turbine model prediction yaw control parameter according to the above embodiment of the present invention, a point in fig. 10(c) is selected, the prediction step length m is 6, and the optimization parameters and the optimization results of the fuzzy inference weight coefficient evaluator corresponding to the point are shown in table 3:
TABLE 3 optimization parameters and results Table
x1 x2 x3 x4 x5
3.3147444 2.1905863 5.0108080 4.3392312 3.5177704
x6 x7 x8 x9 x10
5.4050607 4.3172474 5.3360159 5.3509208 6.3286873
x11 x12 x13 y1 y2
3.8476847 3.5440369 5.4369245 0.0428547 0.0423611
Obtaining a membership function of the optimized fuzzy inference weight coefficient evaluator, such as fig. 13(a), 13(b) and 13(c), wherein a letter above each membership function represents the semantics of the optimized membership function, and the letter in the parenthesis is the original semantics, so that the semantics of part of the membership functions are changed, which represents that the fuzzy rule is optimized to a certain extent while the membership functions are optimized, and the optimized fuzzy rule, such as table 4, and the corresponding rule curve, such as fig. 14, are obtained;
TABLE 4 fuzzy rule Table after optimization
Figure BDA0003112783200000161
Figure BDA0003112783200000171
In fig. 15(a) and 15(b), FDWE-MPC is model predictive control for adjusting weight by using the optimized fuzzy inference weight coefficient evaluator, MPC is conventional model predictive control, fig. 15(a) is a schematic diagram of power capture loss rate, and fig. 15(b) is a schematic diagram of yaw rate, and it can be seen that the model predictive control for adjusting weight by using the optimized fuzzy inference weight coefficient evaluator reduces about 0.3% of power capture loss under the condition of reducing about 1% of yaw rate, and fig. 15(a) and 15(b) prove the effectiveness of the method for adjusting the wind turbine model predictive yaw control parameters.
The wind turbine model prediction yaw control parameter adjusting method according to the above embodiment of the present invention adopts an optimization strategy of associating fuzzy rules with membership functions to reduce the complexity of the optimization problem and simultaneously ensure the optimization performance, and can optimize the fuzzy rules and the membership functions simultaneously, and a majority of regions of a pareto curved surface obtained by optimizing a fuzzy inference weight coefficient evaluator by using an adaptive mesh multi-objective group algorithm dominates a pareto curve under the normal model prediction yaw control, and the wind turbine model prediction yaw control parameter adjusting method can dynamically adjust the weight coefficient of a model prediction control target function according to wind direction information, and can obtain smaller power capture loss under the same yaw rate, obtain lower yaw rate under the same power capture loss and obtain smaller power capture loss under the lower yaw rate, the wind direction change can be more accurately tracked by the lower yaw rate of the cabin position of the fan, the yaw rate is reduced, the yaw error is reduced, and the power capture loss of the fan is reduced.
In the method for adjusting the wind turbine model prediction yaw control parameter according to the above embodiment of the present invention, the abscissa time in fig. 8 represents time; the wind direction of the ordinate represents the wind direction, and the unit is degree; the abscissa activator usage ratio in fig. 9, 10, 11, 12 and 15 represents the yaw actuator usage rate, and the ordinate power fractional ratio represents the power loss rate.
The foregoing is a preferred embodiment of the present invention, and it should be noted that it is obvious to those skilled in the art that various modifications and improvements can be made without departing from the principle of the present invention, and these modifications and improvements should be construed as the protection scope of the present invention.

Claims (10)

1. A method for adjusting a wind turbine model prediction yaw control parameter is characterized by comprising the following steps:
step 1, setting a yaw system of a wind turbine into a yaw mode adopting model predictive control;
step 2, determining a yaw error of model prediction control, and establishing a yaw error m-step prediction model according to the yaw error;
step 3, determining a control target of model predictive control, and carrying out normalization processing on the control target;
step 4, determining a target function of model predictive control according to the control target after normalization processing;
designing a weight coefficient evaluator, and optimizing parameters of the fuzzy inference weight coefficient evaluator by adopting a fuzzy rule and membership function association optimization strategy and a self-adaptive grid multi-target particle swarm algorithm;
and 6, determining a weight coefficient in the model predictive control objective function through the optimized fuzzy inference weight coefficient evaluator.
2. The method for adjusting the wind turbine model predictive yaw control parameter as claimed in claim 1, wherein the step 1 specifically comprises:
the yaw system of the wind turbine is a rigid yaw model, the rigid yaw model adopts a yaw control mode of model predictive control, the control quantity of the yaw system is a yaw rate, and the yaw rate has three possibilities: -0.5deg/s, 0deg/s and 0.5deg/s, generalizing the problem of yaw rate of a wind turbine yaw system into a three-element control set, as follows:
Figure FDA0003112783190000011
wherein,
Figure FDA0003112783190000012
the Yaw rate of a Yaw system of the wind turbine is represented, k represents a time, j represents the number of different Yaw conditions, the Yaw rate is designated by j, Yaw _ speed represents a Yaw rate value, Yaw _ speed is 0.5deg/s, in actual operation, the Yaw rate of the current time is influenced by the Yaw rate of the previous time, and the influence of the Yaw rate of the previous time is represented as:
Figure FDA0003112783190000013
wherein,
Figure FDA0003112783190000021
indicating the yaw rate at time k,
Figure FDA0003112783190000022
representing the yaw rate at time k + 1.
3. The method for adjusting the wind turbine model predictive yaw control parameter as claimed in claim 2, wherein the step 2 specifically comprises:
model predictive control of yaw error theta at a time tyeAs follows:
Figure FDA0003112783190000023
wherein, thetaωd(t) represents the wind direction at time t, θnp(t) denotes the nacelle position at time t, θnp(t1) represents t1The cabin position at the moment;
an m-step prediction model of the yaw error is obtained according to equation (3) as follows:
Figure FDA0003112783190000024
wherein, thetanp(k + m | k) represents the nacelle position at time k + m,
Figure FDA0003112783190000025
representing the yaw rate, T, between time k and time k + isDenotes a control period, and m denotes a prediction step size.
4. The method for adjusting the wind turbine model predictive yaw control parameter as claimed in claim 3, wherein the step 3 specifically comprises:
considering yaw error, predicting power capture E of fan in m stepscapAs follows:
Figure FDA0003112783190000026
where ρ represents the air density, Ar＝πR2Representing swept area, C, of the fan impeller surfacepRepresenting the aerodynamic power coefficient, V0Representing the equivalent wind speed at the hub;
time t of operation of yaw actuator within m-step prediction rangeyawAs follows:
Figure FDA0003112783190000027
wherein,
Figure FDA0003112783190000028
is a logical value which is 1 when the yaw rate is not 0 and is 0 when the yaw rate is 0; equations (5) and (6) are two control targets for yaw control.
5. The method for adjusting parameters of predictive yaw control of a wind turbine model according to claim 4, wherein the step 3 further comprises:
capturing the power of the fan under the m-step prediction EcapAnd the action time t of the yaw actuator in the m-step prediction rangeyawAs a control target of model predictive control, the power capture E of the fan in consideration of m-step predictioncapAnd the action time t of the yaw actuator in the m-step prediction rangeyawDimension difference exists, and the power capture E of the fan under the m-step prediction is respectively carried outcapAnd the action time t of the yaw actuator in the m-step prediction rangeyawNormalization processing was performed as follows:
Figure FDA0003112783190000031
where ξ represents the power capture loss rate due to yaw error, EidealRepresenting the power capture of the wind turbine under ideal conditions;
Figure FDA0003112783190000032
where ζ represents the rate of use of the yaw actuator, ttolRepresenting the total time the fan is running.
6. The method for adjusting the wind turbine model predictive yaw control parameter as claimed in claim 5, wherein the step 4 specifically comprises:
an objective function QF of the model predictive control is obtained from the control targets ξ and ζ of the model predictive control, as follows:
Figure FDA0003112783190000033
where ω represents a weight coefficient to balance the power capture loss rate and the usage rate of the yaw actuator, and the weight coefficient ω has a value in the range of [0,1 ].
7. The method for adjusting the wind turbine model predictive yaw control parameter as claimed in claim 6, wherein the step 5 specifically comprises:
taking the weighted average of the difference between the position of the cabin in the prediction step and the predicted wind direction and the root mean square of the weighted average of the difference between the position of the cabin in the prediction step and the predicted wind direction as the input of a fuzzy inference weight coefficient evaluator;
in the fuzzy inference weight coefficient evaluator, the region center calculation inference result obtains clear output by using a gravity center method, as follows:
Figure FDA0003112783190000041
wherein y represents a clear output, RiIndicating the ith ruleInference result, XiThe output corresponding to the ith rule is shown, and n represents the number of the rules;
weighted average WD of the difference between the nacelle position and the predicted wind direction within the prediction stepavThe calculation of (a) is as follows:
Figure FDA0003112783190000042
wherein WDavRepresents a weighted average of the difference between the nacelle position and the predicted wind direction within the prediction step, will WDavAs an input to a fuzzy inference weight coefficient evaluator; WD (k + i | k) represents the predicted yaw error at time k + i at time k;
root mean square WD of weighted average of differences between nacelle position and predicted wind direction within predicted step sizestdThe calculation of (a) is as follows:
Figure FDA0003112783190000043
wherein WDstdRoot mean square representing a weighted average of the difference between the nacelle position and the predicted wind direction over a prediction step, WDstdAs another input to the fuzzy inference weight coefficient evaluator;
from the calculated input WDavAnd WDstdThe fuzzy inference weight coefficient evaluator outputs a weight coefficient omega.
8. The method of claim 7, wherein the step 5 further comprises:
the optimization problem of the membership function and the fuzzy rule of the fuzzy inference weight coefficient evaluator is as follows:
min ferror(xmembership，xrule)
Figure FDA0003112783190000044
wherein f iserrorRepresenting an objective function, which is an error between the inference result and an ideal optimal value; x is the number ofmembershipRepresenting the optimal vector membership function, xruleRepresents the optimized vector fuzzy rule, ΩmRepresenting the feasible domain, Ω, of the membership function of the optimization vectorrA feasible field representing an optimized vector fuzzy rule;
in the optimization of vector xmembershipIn the method, the adjusted parameters are divided into three types of quantity, shape and position distribution, the quantity m of input and output and the value range thereof are preset, and the quantity num of membership functions is presetmfAs follows:
nummf＝[num1，num2，…，numm] (14)
wherein, nummfNumber of membership functions, nummThe number of membership functions contained for each input or output;
the types of membership functions are 8, and the type of each membership function is restricted
Figure FDA0003112783190000051
As follows:
Figure FDA0003112783190000052
wherein,
Figure FDA0003112783190000053
indicates num contained in the nth input or outputnA type vector of each membership function, wherein the value range of elements in the vector is [1,2, …,8 ]]Respectively corresponding to 8 membership function types;
the determination of the location and shape of each membership function is related to the type of each membership function, the location of each membership function is constrained as follows:
Figure FDA0003112783190000054
wherein,
Figure FDA0003112783190000055
indicating num of the nth input or outputnThe location and shape of the kth of the membership functions;
assuming that the type of membership function is a triangular membership function, fpositionThe concrete expression is as follows:
Figure FDA0003112783190000056
wherein f ispositionRepresenting the position and shape functions of the membership functions; according to the formula (17), the position and the shape of the triangular membership function are determined by three parameters, and the three parameters are constrained by upper and lower limits of corresponding input or output;
the fuzzy rule is influenced by the number of input and output, the back piece of the fuzzy rule of fuzzy inference adopting the Mamdani model is a certain fuzzy set of output quantity, if s inputs exist in the fuzzy inference, the feasible domain of the fuzzy rule constraint is as follows:
Figure FDA0003112783190000061
wherein omegarSet of all fuzzy rules representing fuzzy inference of s-input, m-s-output.
9. The method of claim 8, wherein the step 5 further comprises:
simplifying the optimization problem of the fuzzy rule and the membership function of the fuzzy inference weight coefficient evaluator by adopting a fuzzy rule-membership function association optimization strategy:
determining the sum of the input and output quantities according to formula (13) -formula (18)Number of linguistic values, and thus the number of membership functions nummfAnd type constraint of each membership function
Figure FDA0003112783190000062
Assuming that the position of each membership function is uniquely determined by a certain vertex of the triangle, the optimization dimensionality of the membership function is reduced, and for the fuzzy rule, the order constraint of the language value of the output membership function is not considered, and the association optimization is carried out according to the optimization process of the membership function:
the optimization of the membership functions results in a change in the sequence of output linguistic values as follows:
B＝A·S (19)
wherein, A represents the original sequence, S represents the elementary matrix after elementary transformation, called transformation matrix;
the transformation of the fuzzy rule includes two methods: the first method transforms according to the S matrix as follows:
NEWrules[3×5]＝OLDrules[3×5]·S[5×5] (20)
the second method processes the values in the fuzzy rule matrix according to the variation of the linguistic values, which are assumed to be the [ seml VS VL ] sequence, as follows:
[S M L VS VL]→[VS(S) S(M) M(L) L(VS) VL(VL)] (21)
and (3) correspondingly processing each language value in the fuzzy rule matrix with a certain probability according to a formula (21).
10. The method of claim 9, wherein the step 5 further comprises:
searching and solving the simplified optimization problem by adopting a self-adaptive grid multi-target particle swarm algorithm: the self-adaptive grid multi-target particle swarm algorithm comprises the following steps: the self-adaptive grid algorithm, the selection of global optimal particles and the Archive set truncation technology assume a two-dimensional target space optimization problem, and the self-adaptive grid algorithm calculates the search range of a target space after k iterations
Figure FDA0003112783190000063
And
Figure FDA0003112783190000064
the modulus of the grid is computed as follows:
Figure FDA0003112783190000071
Figure FDA0003112783190000072
wherein, Delta Y1 KRepresenting the modulus value of the target 1 of the kth iteration,
Figure FDA0003112783190000073
representing the modulus, Y, of the object 2 of the kth iteration1 KRepresents the calculated value of the target 1 for the k-th iteration,
Figure FDA0003112783190000074
representing the calculated value of the object 2 of the kth iteration, M representing the division of the object space into M grids, Y1And Y2Respectively, the values of the fitness function;
and traversing the group to calculate the grid number of each particle, and calculating the grid number of the particle i as follows:
Figure FDA0003112783190000075
wherein,
Figure FDA0003112783190000076
a grid number representing the particle i with respect to the target 1,
Figure FDA0003112783190000077
a grid number representing the particle i with respect to the target 2,
Figure FDA0003112783190000078
represents the calculated value of the target 1 for the particle i,
Figure FDA0003112783190000079
represents the calculated value of target 2 for particle i, Int represents rounding; obtaining grid density information of each grid according to the grid number of each particle; selecting global optimal particles according to the grid density information of each grid;
global search experience g corresponding to each particle in populationbestAs follows:
Figure FDA00031127831900000710
wherein,
Figure FDA00031127831900000711
represents the corresponding global optimum value of the jth particle in the kth iteration, F (i) represents the number of particles of the grid where the ith particle is located,
Figure FDA00031127831900000712
representing a set of particles in the Archive set that are better than the population particle j,
Figure FDA00031127831900000713
the calculation of (a) is as follows:
Figure FDA00031127831900000714
wherein,
Figure FDA00031127831900000715
indicating a pareto dominant relationship, i.e. i is the dominant solution of j, when there are a plurality of particles satisfying the conditionWhen selecting Archive aggregate particles with the largest potential as gbestAs follows:
Figure FDA00031127831900000716
wherein, PkRepresenting the particle population at the k iteration, wherein the formula (26) represents that the particles with the most dominant population particles are selected from all the particles meeting the formula (24), and when the particles with the same potential exist, one particle is randomly selected;
the Archive set truncation technology adopted by the self-adaptive grid multi-target particle swarm algorithm is based on the density information of the grid, and when the density of the grid exceeds a threshold value, the particles with the minimum potential are deleted according to a formula (26).
CN202110653490.5A 2021-06-11 2021-06-11 Method for adjusting yaw control parameters of wind turbine model prediction Active CN113323821B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110653490.5A CN113323821B (en) 2021-06-11 2021-06-11 Method for adjusting yaw control parameters of wind turbine model prediction

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110653490.5A CN113323821B (en) 2021-06-11 2021-06-11 Method for adjusting yaw control parameters of wind turbine model prediction

Publications (2)

Publication Number Publication Date
CN113323821A true CN113323821A (en) 2021-08-31
CN113323821B CN113323821B (en) 2022-10-21

Family

ID=77420586

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110653490.5A Active CN113323821B (en) 2021-06-11 2021-06-11 Method for adjusting yaw control parameters of wind turbine model prediction

Country Status (1)

Country Link
CN (1) CN113323821B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115983054A (en) * 2023-03-21 2023-04-18 中车山东风电有限公司 Capability evaluation and load processing method of wind generating set and terminal
CN116701830A (en) * 2023-04-26 2023-09-05 清华大学 Pareto front edge solution optimization method based on fuzzy rule and stability reasoning control
CN117293817A (en) * 2023-10-10 2023-12-26 华润电力技术研究院有限公司 Power generation parameter prediction method and device

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101498282A (en) * 2008-02-01 2009-08-05 北京能高自动化技术有限公司 Yaw control method for large-sized wind-driven generator group
CN104485681A (en) * 2015-01-05 2015-04-01 成都鼎智汇科技有限公司 Monitoring method for wind power plant energy storage system
CN105119285A (en) * 2015-09-17 2015-12-02 上海交通大学 Wind power storage coordination multi-objective optimization control method based on dynamic weighting
CN105117593A (en) * 2015-08-18 2015-12-02 河海大学 Wavelet transform and particle swarm optimized grey model-based short-term wind speed forecasting method
CN106677984A (en) * 2016-12-29 2017-05-17 北京金风科创风电设备有限公司 Method, equipment and system for yaw control of wind generating set
CN108105030A (en) * 2017-12-28 2018-06-01 中国船舶重工集团海装风电股份有限公司 A kind of yaw calibration method based on wind turbine sensor
CN108547736A (en) * 2018-03-27 2018-09-18 中南大学 The Yaw control method of wind speed and direction prediction technique and wind power generating set
CN108869175A (en) * 2018-06-15 2018-11-23 中南大学 A kind of Yaw control method of the Model Predictive Control MPC based on multi-step prediction
CN110705760A (en) * 2019-09-19 2020-01-17 广东工业大学 Photovoltaic power generation power prediction method based on deep belief network
CN110857683A (en) * 2018-08-22 2020-03-03 阿里巴巴集团控股有限公司 Yaw control method, device and equipment of wind driven generator

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101498282A (en) * 2008-02-01 2009-08-05 北京能高自动化技术有限公司 Yaw control method for large-sized wind-driven generator group
CN104485681A (en) * 2015-01-05 2015-04-01 成都鼎智汇科技有限公司 Monitoring method for wind power plant energy storage system
CN105117593A (en) * 2015-08-18 2015-12-02 河海大学 Wavelet transform and particle swarm optimized grey model-based short-term wind speed forecasting method
CN105119285A (en) * 2015-09-17 2015-12-02 上海交通大学 Wind power storage coordination multi-objective optimization control method based on dynamic weighting
CN106677984A (en) * 2016-12-29 2017-05-17 北京金风科创风电设备有限公司 Method, equipment and system for yaw control of wind generating set
CN108105030A (en) * 2017-12-28 2018-06-01 中国船舶重工集团海装风电股份有限公司 A kind of yaw calibration method based on wind turbine sensor
CN108547736A (en) * 2018-03-27 2018-09-18 中南大学 The Yaw control method of wind speed and direction prediction technique and wind power generating set
CN108869175A (en) * 2018-06-15 2018-11-23 中南大学 A kind of Yaw control method of the Model Predictive Control MPC based on multi-step prediction
CN110857683A (en) * 2018-08-22 2020-03-03 阿里巴巴集团控股有限公司 Yaw control method, device and equipment of wind driven generator
CN110705760A (en) * 2019-09-19 2020-01-17 广东工业大学 Photovoltaic power generation power prediction method based on deep belief network

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
刘婷婷: "风电功率预测模型的优选准则与融合策略研究", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115983054A (en) * 2023-03-21 2023-04-18 中车山东风电有限公司 Capability evaluation and load processing method of wind generating set and terminal
CN115983054B (en) * 2023-03-21 2023-08-11 中车山东风电有限公司 Capability assessment and load processing method for wind generating set and terminal
CN116701830A (en) * 2023-04-26 2023-09-05 清华大学 Pareto front edge solution optimization method based on fuzzy rule and stability reasoning control
CN117293817A (en) * 2023-10-10 2023-12-26 华润电力技术研究院有限公司 Power generation parameter prediction method and device
CN117293817B (en) * 2023-10-10 2024-06-07 华润电力技术研究院有限公司 Power generation parameter prediction method and device

Also Published As

Publication number Publication date
CN113323821B (en) 2022-10-21

Similar Documents

Publication Publication Date Title
CN113323821B (en) Method for adjusting yaw control parameters of wind turbine model prediction
CN112733462B (en) Ultra-short-term wind power plant power prediction method combining meteorological factors
CN110888317A (en) PID controller parameter intelligent optimization method
Wang et al. A novel combined forecasting model based on neural networks, deep learning approaches, and multi-objective optimization for short-term wind speed forecasting
Yong An improved artificial fish swarm algorithm for optimal operation of cascade reservoirs
Lv et al. A newly combination model based on data denoising strategy and advanced optimization algorithm for short-term wind speed prediction
Lu et al. Study on the optimal hydropower generation of Zhelin reservoir
CN113741184B (en) Wind turbine load prediction control method based on sliding-mode observer
CN114322199A (en) Ventilation system autonomous optimization operation regulation and control platform and method based on digital twins
CN114298140A (en) Wind power short-term power prediction correction method considering unit classification
CN115239029B (en) Wind power prediction method and system considering power time sequence and meteorological dependent characteristics
Li et al. Wind power prediction based on PSO-Kalman
CN116526473A (en) Particle swarm optimization LSTM-based electrothermal load prediction method
Zhang et al. Research on bulbous bow optimization based on the improved PSO algorithm
CN113937808A (en) Improved sparrow search algorithm-based distributed power supply site selection and volume fixing optimization method
CN113141012A (en) Power grid power flow regulation and control decision reasoning method based on deep deterministic strategy gradient network
CN113294297B (en) Variable weight adjusting method for wind turbine generator nonlinear model prediction torque control
Shen et al. Short-term load forecasting of power system based on similar day method and PSO-DBN
CN114036506A (en) Method for detecting and defending false data injection attack based on LM-BP neural network
CN116070498A (en) Handheld rapid damage assessment device
CN115716469A (en) Output power distribution control method of hybrid power system
CN110518644B (en) Reactive power optimization method and device for wind power plant and wind power plant
CN114139783A (en) Wind power short-term power prediction method and device based on nonlinear weighted combination
CN111859780A (en) Micro-grid operation optimization method and system
CN111709578A (en) Short-time ship traffic flow prediction method and device and storage medium

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