CN110932274A - Power system measurement and load parameter analysis and identification method - Google Patents

Power system measurement and load parameter analysis and identification method Download PDF

Info

Publication number
CN110932274A
CN110932274A CN201911307051.8A CN201911307051A CN110932274A CN 110932274 A CN110932274 A CN 110932274A CN 201911307051 A CN201911307051 A CN 201911307051A CN 110932274 A CN110932274 A CN 110932274A
Authority
CN
China
Prior art keywords
load
algorithm
parameter
parameters
load model
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.)
Withdrawn
Application number
CN201911307051.8A
Other languages
Chinese (zh)
Inventor
程海军
原琳
冮明颖
姜丕杰
屈丹
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.)
Liaoning University of Technology
Original Assignee
Liaoning 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 Liaoning University of Technology filed Critical Liaoning University of Technology
Priority to CN201911307051.8A priority Critical patent/CN110932274A/en
Publication of CN110932274A publication Critical patent/CN110932274A/en
Withdrawn legal-status Critical Current

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
    • 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
    • H02J3/12—Circuit arrangements for ac mains or ac distribution networks for adjusting voltage in ac networks by changing a characteristic of the network load
    • H02J3/14—Circuit arrangements for ac mains or ac distribution networks for adjusting voltage in ac networks by changing a characteristic of the network load by switching loads on to, or off from, network, e.g. progressively balanced loading
    • 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
    • Y02B—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO BUILDINGS, e.g. HOUSING, HOUSE APPLIANCES OR RELATED END-USER APPLICATIONS
    • Y02B70/00—Technologies for an efficient end-user side electric power management and consumption
    • Y02B70/30—Systems integrating technologies related to power network operation and communication or information technologies for improving the carbon footprint of the management of residential or tertiary loads, i.e. smart grids as climate change mitigation technology in the buildings sector, including also the last stages of power distribution and the control, monitoring or operating management systems at local level
    • Y02B70/3225—Demand response systems, e.g. load shedding, peak shaving
    • 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
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S20/00—Management or operation of end-user stationary applications or the last stages of power distribution; Controlling, monitoring or operating thereof
    • Y04S20/20—End-user application control systems
    • Y04S20/222—Demand response systems, e.g. load shedding, peak shaving

Abstract

A method for measuring, analyzing and identifying load parameters of an electric power system relates to a method for identifying parameters of the electric power system, and comprises the following steps: firstly, clustering and analyzing loads according to load measurement information, and determining the installation site of a load recording device; secondly, constructing a comprehensive load model of the parallel static characteristic load of the motor, and establishing an identification target function; then, parameter identification is carried out through the measured data of the load recording device by adopting a hybrid optimization algorithm combining an ant colony algorithm and a gradient algorithm; performing parameter optimization on the multiple identification results by using unbiased and optimal estimation characteristics of the Kriging algorithm; and finally, counting the load models of different sites to form a load model parameter library. The invention improves the speed of parameter identification, realizes parameter optimization for multiple identification results by adopting a Kriging algorithm, and improves the precision of parameter identification. The method can provide a model foundation for research in the fields of special power planning, power system stability analysis, power system energy efficiency evaluation and the like.

Description

Power system measurement and load parameter analysis and identification method
Technical Field
The present invention relates to a method for identifying parameters of an electrical power system, and more particularly, to a method for measuring and analyzing load parameters of an electrical power system.
Background
The development of the power system and the realization of large-area networking enable the scale of the power system to be enlarged day by day, the structure of a power grid to be more complex, obvious benefits are brought to national economy, meanwhile, the operation point of the power system is enabled to be closer to a stable limit, and the risk that the whole system is subjected to voltage instability and even voltage breakdown is increased continuously. Meanwhile, under the background of the smart grid, the real-time operation data brought by informatization and interaction of the power system and the grid connection of various distributed power supplies bring new challenges to the operation, planning and design of the power system. The power system simulation technology is one of the important scientific tools for analyzing and researching the operation mechanism of the power system, and has become an indispensable means in the planning, operation and control of the power system. The simulation research and various specific analyses of the power system are established on the basis of corresponding mathematical models, good load modeling and parameter identification are realized, and the method has important significance in the operation, control and calculation of the power system.
At present, load modeling faces two key problems in the practical process: the load is time-varying, and even for the same load node, the load composition changes along with time and season, and different load characteristics are presented at different moments; and secondly, the regional difference of the load, the regional distribution of the power load is very wide, and the load characteristics of each load node are different due to different load compositions. From the viewpoint of improving the accuracy of the load model, if the load model can be established for different load nodes, various load models in different time periods are certainly optimal, but in practice, the point is difficult to achieve, and the load model adopted by the power grid simulation calculation requirement is simplified as little as possible.
At present, the load parameter identification in China is mainly a method for actually measuring the load by adopting a load characteristic recorder. The regional distribution of the power load is very wide, a 22OkV load substation of a provincial power grid can also have dozens of hundreds, and the comprehensive load characteristics of each load node are different due to different load compositions. How to select a reasonable installation point for installing the load recording device from a plurality of load points is to classify loads with load characteristics close to or similar to those of different load points and perform cluster analysis.
Aiming at the development trend of load modeling of a wide area power system and integral modeling of the power system in the future, when the number of parameters to be identified is increased greatly, the algorithm is required to have higher efficiency. Meanwhile, due to the requirement of online or even real-time development of load modeling work of the power system, the calculation speed of parameter identification is required to be increased. These requirements have prompted the need for further research into the power load model parameter identification method.
The power load model is a nonlinear mathematical model, most of the parameter identification methods of nonlinear systems are based on optimization methods at present, and the main process is to find a group of optimal parameter vectors so as to minimize a preset error objective function value, wherein the error objective function is a function of parameters to be identified. There are several algorithms to solve the power system optimization problem: such as gradient algorithm, the local search efficiency is very high, but the global robustness is poor; the simulated evolutionary algorithm has strong global search capability, but poor local search efficiency. The power load model parameter identification algorithm mainly comprises a genetic algorithm, an ant colony algorithm, a particle swarm algorithm, a gradient algorithm and the like, wherein the ant colony algorithm simulates the foraging behavior of real ants in nature, and has the advantages of good overall performance, positive feedback and cooperativity, but long required calculation time.
Disclosure of Invention
The invention aims to provide a method for analyzing and identifying power system measurement and load parameters, which is a method for identifying a power system measurement and load model based on multi-algorithm fusion, realizes power system energy efficiency evaluation, optimizes power load identification parameters and improves parameter identification precision. The method is suitable for being applied to the field of electric power special planning, the technical field of electric power system stability analysis and the field of electric power system energy efficiency evaluation.
The purpose of the invention is realized by the following technical scheme:
a method for measuring, analyzing and identifying load parameters of a power system comprises the following steps:
s1: determining the installation point of the load recording device according to the clustering analysis of the load measurement information;
s2: the load model adopts a comprehensive load model, and the structure of the load model is a motor parallel static characteristic load model which reflects the dynamic characteristic of the load; the model has 14 parameters, namely the stator resistance of the induction motor
Figure DEST_PATH_IMAGE001
And stator reactance
Figure 928799DEST_PATH_IMAGE002
Excitation reactance
Figure DEST_PATH_IMAGE003
Rotor resistance
Figure 659995DEST_PATH_IMAGE004
And rotor reactance
Figure DEST_PATH_IMAGE005
Time constant of inertia of rotor
Figure 715675DEST_PATH_IMAGE006
Torque equation constants A and B, and proportion of initial active power of induction motor to total initial active power of load
Figure DEST_PATH_IMAGE007
Initial load factor of induction motor
Figure 532322DEST_PATH_IMAGE008
And the proportion coefficient of the active power and the reactive power of the constant impedance and the constant power in the static characteristic load model
Figure DEST_PATH_IMAGE009
The initial active power of the motor is P, and the P is the active power consumed by the measured load point in the transient process;
Figure 331650DEST_PATH_IMAGE010
is a rated initial load factor of the load,
Figure DEST_PATH_IMAGE011
is the rated capacity of the induction motor,
Figure 855036DEST_PATH_IMAGE012
is the reference voltage of the load and is,
Figure DEST_PATH_IMAGE013
is the initial value of the load bus voltage in the transient process;
Figure 347197DEST_PATH_IMAGE014
is an independent parameter vector to be identified of the load model,
Figure DEST_PATH_IMAGE015
to pass steady state conditions of the motor and
Figure 656999DEST_PATH_IMAGE016
the obtained vector of the identification parameter is then calculated,
Figure DEST_PATH_IMAGE017
is the initial transient voltage of the motor and,
Figure 994439DEST_PATH_IMAGE018
is the slip ratio of the motor and is,
Figure DEST_PATH_IMAGE019
for the synchronous reactance between the stator and the rotor,
Figure 637910DEST_PATH_IMAGE020
is the load factor;
s3: carrying out load model parameter identification, and specifically comprising the following steps:
s3.1: obtaining a record of the actual load dynamics at the site by means of a load recording device, from
Figure DEST_PATH_IMAGE021
Start sampling at a time to
Figure 35393DEST_PATH_IMAGE022
The sampling is finished at the moment, and the sampling is counted
Figure DEST_PATH_IMAGE023
Next, the process of the present invention,
Figure 826632DEST_PATH_IMAGE024
the actual measured value of the time is recorded as
Figure DEST_PATH_IMAGE025
Corresponding to the identified load model
Figure 967763DEST_PATH_IMAGE026
The response value at the time is
Figure DEST_PATH_IMAGE027
；
S3.2: setting independent parameter phasors to be identified
Figure 528058DEST_PATH_IMAGE028
The following objective functions are optimized by adopting a hybrid optimization algorithm:
Figure DEST_PATH_IMAGE029
the basic idea of the hybrid optimization algorithm is that the ant colony algorithm and the gradient algorithm are combined, the characteristics that the ant colony algorithm is good in global optimization performance and fast in local convergence of the gradient algorithm are comprehensively utilized, and a new parameter identification algorithm is formed; starting iteration by adopting an ant colony algorithm, selecting a plurality of current superior ants at a certain proper moment, and switching to a gradient algorithm for search iteration until convergence;
information quantity function in ant colony algorithm
Figure 299704DEST_PATH_IMAGE030
And the number of ants per sub-interval
Figure DEST_PATH_IMAGE031
Determined according to the following formula:
Figure 312660DEST_PATH_IMAGE032
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE033
is the peak value of pheromone content between two ants and is an objective function
Figure 257482DEST_PATH_IMAGE034
The reciprocal of (a);
Figure DEST_PATH_IMAGE035
is a compression factor;
Figure 672283DEST_PATH_IMAGE036
the number of ants;
Figure DEST_PATH_IMAGE037
is as follows
Figure DEST_PATH_IMAGE039
Only ants and the second
Figure 677148DEST_PATH_IMAGE040
The distance between the ants is only the distance between the ants,
Figure DEST_PATH_IMAGE041
，
Figure 177399DEST_PATH_IMAGE042
presentation pair
Figure DEST_PATH_IMAGE043
Pheromone progression in dimensional space
Figure 598017DEST_PATH_IMAGE043
The sum of the pheromone contents in the whole space is obtained by the calculus;
the iterative formula of the gradient method is determined as follows:
Figure 132903DEST_PATH_IMAGE044
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE045
as an objective function in
Figure 308669DEST_PATH_IMAGE043
The gradient of the dimensional space is determined,
Figure 233900DEST_PATH_IMAGE046
is a step size factor;
s4: load model parameter optimization
According to what is built
Figure DEST_PATH_IMAGE047
The load model parameter set is respectively estimated for each row of parameters by adopting a Kriging algorithm, so that the aim of parameter optimization is fulfilled;
s5: load feature library update
Respectively carrying out parameter optimization of Kriging algorithm on 14 parameters to be identified of the comprehensive load model, and calculating unbiased and optimal estimated values of all the parameters under actual measurement data to obtain a group of load model parameters reflecting statistical characteristics of the actual measurement data; when a plurality of new measured data are recorded on site, curve identification is carried out on the measured data, then newly identified load model parameters are added into a load parameter set, and parameter optimization of the Kriging algorithm is carried out again to replace the original model parameters;
the power system measurement and load parameter analysis and identification method and the load clustering method adopt a SOM neural network clustering analysis method, and the connection weight adjustment of a winning neuron and a neighbor neuron thereof is determined according to the following formula:
Figure 520525DEST_PATH_IMAGE048
wherein
Figure DEST_PATH_IMAGE049
Represents a learning rate, and
Figure 644339DEST_PATH_IMAGE050
；
Figure DEST_PATH_IMAGE051
is a function of the field of the winning neuron,
Figure 928690DEST_PATH_IMAGE052
；
Figure DEST_PATH_IMAGE053
and
Figure 403533DEST_PATH_IMAGE054
respectively the positions of the winning node and other nodes j in the two-dimensional topological space of the output layer;
Figure DEST_PATH_IMAGE055
is the range of the neighborhood.
In the step S3.2, the switching conditions of different algorithms in the hybrid optimization algorithm are slowed down because the ant colony algorithm gradually approaches the optimal solution during local iteration, so that the algorithm is switched according to the fact that the relative value between the optimal objective function values between two adjacent ant colony algorithms is less than a certain degree, that is, the method is determined according to the following formula:
Figure 511427DEST_PATH_IMAGE056
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE057
the value of the objective function is,
Figure 755327DEST_PATH_IMAGE058
in order to be able to perform the number of iterations,
Figure DEST_PATH_IMAGE059
to switch the threshold.
The electric power system measurement and load parameter analysis distinguishIdentification method, based on load model parameter optimization, established
Figure 945000DEST_PATH_IMAGE060
The load model parameter set is respectively estimated for each row of parameters by adopting a Kriging algorithm, so that the aim of parameter optimization is fulfilled; for a certain load model parameter variable
Figure DEST_PATH_IMAGE061
Estimation of Kriging algorithm
Figure 907139DEST_PATH_IMAGE062
Determined as follows:
Figure DEST_PATH_IMAGE063
wherein the content of the first and second substances,
Figure 535567DEST_PATH_IMAGE064
the weight coefficient represents the degree of contribution of each recognition value to the estimation value. To ensure that the estimation is unbiased and optimal,
Figure 837235DEST_PATH_IMAGE064
the value of (a) is determined by:
Figure DEST_PATH_IMAGE065
wherein the content of the first and second substances,
Figure 260126DEST_PATH_IMAGE066
lagrange coefficients.
The invention has the advantages and effects that:
1. according to the load measurement information, the SOM neural network method is adopted to realize cluster analysis on the power load, and a relatively excellent load recording device installation place is given, and is representative;
2. according to the method, a hybrid optimization algorithm combining an ant colony algorithm and a gradient algorithm is adopted to identify the parameters of the load model, algorithm switching conditions are given, the characteristics of good global optimization performance and fast local convergence of the gradient algorithm of the ant colony algorithm are comprehensively utilized, and the speed of parameter identification is increased;
3. the invention utilizes the unbiased and optimal estimation characteristics of the Kriging algorithm to the multiple identification results, realizes the optimization of load identification parameters and improves the accuracy of parameter identification.
Drawings
FIG. 1 is a flow chart of an embodiment of a method for identifying load parameters of an electrical power system based on multi-algorithm fusion according to the present invention;
FIG. 2 is a flow chart of a hybrid optimization algorithm for load parameter identification in the present invention.
Detailed Description
The following description of the embodiments of the present invention is provided in order to better understand the present invention for those skilled in the art with reference to the accompanying drawings. It is to be expressly noted that in the following description, a detailed description of known functions and designs will be omitted when it may obscure the subject matter of the present invention.
The invention comprises the following steps:
firstly, determining the installation point of a load recording device according to the clustering analysis of load measurement information; a cluster analysis method adopting an SOM neural network comprises the following steps: initializing a network, providing an input mode, calculating the distance between a connection weight vector and the input mode, determining a winning neuron, adjusting the connection weight of the winning neuron and a neighbor neuron thereof, and selecting a new input mode; repeatedly calculating the distance between the connection weight vector and the input mode, determining a winning neuron, and adjusting the connection weight of the winning neuron and the neighbor neurons thereof; until all samples are learned once, updating the learning rate and the neighborhood function, completing one-time network learning, and returning to the step of providing the input mode; until reaching the set learning times;
constructing a load model, wherein the structure of the load model is a motor parallel static characteristic load model, and the load model comprises 14 independent parameters to be identified;
the method adopts a hybrid optimization algorithm combining an ant colony algorithm and a gradient algorithm to identify the parameters of the load model, and comprises the following specific steps: initializing ant colony, calculating information quantity, calculating distribution of ant number in subintervals, determining moving direction and moving, calculating a target function, and judging whether algorithm switching is met; if yes, entering a gradient algorithm step; if not, entering an information quantity calculation step and a gradient algorithm and judging whether the optimal value is reached or not; if yes, entering a step of outputting an identification result; if not, entering a gradient algorithm step;
outputting the identification result;
optimizing the parameters of the load model by using a Kriging algorithm;
and updating the load characteristic library.
Examples
Fig. 1 is a flowchart of an embodiment of a method for identifying load parameters of an electrical power system based on multi-algorithm fusion according to the present invention. As shown in fig. 1, the method for identifying load parameters of a power system based on multi-algorithm fusion of the present invention comprises the following steps:
s1: according to the clustering analysis of the load measurement information, determining the installation point of the load recording device:
how to select a reasonable installation point for installing the load recording device from a plurality of load points is to classify loads with load characteristics close to or similar to those of different load points and perform cluster analysis. The invention adopts a SOM neural network clustering analysis method, which comprises the following steps:
s1.1: initializing the network, determining the number n of input layer neurons, the number m of competition layer neurons and the adopted topological structure, and connecting each output layer with weight value
Figure DEST_PATH_IMAGE067
Is assigned to [0, l]And setting the network learning times T according to the random value in the interval, and stopping learning when the total learning times of the network reaches T.
S1.2: providing a new input mode
Figure 709562DEST_PATH_IMAGE068
And inputting the data into the network.
S1.3: computing connection weight vectors
Figure DEST_PATH_IMAGE069
And input mode
Figure 141680DEST_PATH_IMAGE070
The distance between
Figure DEST_PATH_IMAGE071
Figure 32276DEST_PATH_IMAGE072
S1.4: determining winning neurons, and input patterns
Figure DEST_PATH_IMAGE073
The neuron with the smallest distance is the winning neuron.
If so
Figure 891648DEST_PATH_IMAGE074
The vector of connecting weights representing the winning neuron C and the input neuron is:
Figure DEST_PATH_IMAGE075
s1.5: adjusting the connection weights of winning neurons and their neighbor neurons
Figure 562800DEST_PATH_IMAGE076
Wherein
Figure DEST_PATH_IMAGE077
Represents a learning rate, and
Figure 798610DEST_PATH_IMAGE078
is a domain function around the winning neuron, and adopts Gauss neighborhood function
Figure DEST_PATH_IMAGE079
Figure 871608DEST_PATH_IMAGE080
And
Figure DEST_PATH_IMAGE081
respectively the positions of the winning node and other nodes j in the two-dimensional topological space of the output layer,
Figure 839564DEST_PATH_IMAGE082
reflecting the range of the neighborhood.
S1.6: and (5) selecting a new input mode, and repeating the steps (3), (4) and (5) until all samples are learned.
S1.7: updating learning rates and neighborhood functions
Figure DEST_PATH_IMAGE083
Figure 998013DEST_PATH_IMAGE084
Monotonically decreases over time, thereby ensuring convergence of the learning process.
Figure DEST_PATH_IMAGE085
Figure 771934DEST_PATH_IMAGE086
Is that
Figure DEST_PATH_IMAGE087
The initial value of (a) is,
Figure 699439DEST_PATH_IMAGE088
as a time constant, take
Figure DEST_PATH_IMAGE089
。
As the number of iterations increases in the sequence,
Figure 838296DEST_PATH_IMAGE087
with an exponential decrease, the topological neighborhood scales shrink in a corresponding manner, which means that as the number of iterations increases, the incentive of the winning neuron to the neighborhood neuron decreases, thereby reinforcing its own advantage in responding to a certain pattern.
S1.8: order to
Figure 484041DEST_PATH_IMAGE090
And returning to the step (2) until
Figure DEST_PATH_IMAGE091
Until now.
S2: building a load model
The load model adopts a comprehensive load model, has a structure of a motor parallel static characteristic load model, is widely applied to power system simulation and actual identification, and has the advantages of strong organic property and capability of better reflecting the dynamic characteristic of the load. The model has 14 parameters, namely the stator resistance of the induction motor
Figure 796073DEST_PATH_IMAGE092
And stator reactance
Figure DEST_PATH_IMAGE093
Excitation reactance
Figure 781347DEST_PATH_IMAGE094
Rotor resistance
Figure DEST_PATH_IMAGE095
And rotor reactance
Figure 159282DEST_PATH_IMAGE096
Time constant of inertia of rotor
Figure DEST_PATH_IMAGE097
Torque equation constants A and B, and proportion of initial active power of induction motor to total initial active power of load
Figure 292323DEST_PATH_IMAGE098
InducingInitial load factor of motor
Figure DEST_PATH_IMAGE099
And the proportion coefficient of the active power and the reactive power of the constant impedance and the constant power in the static characteristic load model
Figure 408046DEST_PATH_IMAGE100
The initial active power of the motor is P, and the P is the active power consumed by the measured load point in the transient process;
Figure 247827DEST_PATH_IMAGE099
is a rated initial load factor of the load,
Figure DEST_PATH_IMAGE101
is the rated capacity of the induction motor,
Figure 790803DEST_PATH_IMAGE102
is the reference voltage of the load and is,
Figure DEST_PATH_IMAGE103
is the initial value of the load bus voltage during the transient process.
Figure 145561DEST_PATH_IMAGE104
Is an independent parameter vector to be identified of the load model,
Figure DEST_PATH_IMAGE105
to pass steady state conditions of the motor and
Figure 64976DEST_PATH_IMAGE106
the obtained vector of the identification parameter is then calculated,
Figure DEST_PATH_IMAGE107
is the initial transient voltage of the motor and,
Figure 759262DEST_PATH_IMAGE108
is the slip ratio of the motor and is,
Figure DEST_PATH_IMAGE109
for the synchronous reactance between the stator and the rotor,
Figure 473140DEST_PATH_IMAGE110
is the load factor.
S3: performing load model parameter identification
The power load model is a nonlinear mathematical model, most of the parameter identification methods of nonlinear systems are based on optimization methods at present, and the main process is to find a group of optimal parameter vectors so as to minimize a preset error objective function value, wherein the error objective function is a function of parameters to be identified. In order to find the minimum error between the objective function and the identification value, the parameter identification criterion of the power load model is as follows:
Figure DEST_PATH_IMAGE111
wherein
Figure 580774DEST_PATH_IMAGE112
For the start time of the sampling to be,
Figure DEST_PATH_IMAGE113
for the end of sampling, total sampling
Figure 38300DEST_PATH_IMAGE114
Next, the process is carried out.
Figure DEST_PATH_IMAGE115
And
Figure 587093DEST_PATH_IMAGE116
is a time of day
Figure 737451DEST_PATH_IMAGE118
And the output response obtained by identifying the model.
The invention adopts a hybrid optimization algorithm combining an ant colony algorithm and a gradient algorithm, namely, the ant colony algorithm is adopted to start iteration, a plurality of current superior ants are selected at a certain proper moment, and the ant colony optimization algorithm is switched to the gradient algorithm to search iteration until convergence. Fig. 2 is a flow chart of a hybrid optimization algorithm for load parameter identification according to the present invention, and the specific process is as follows:
s3.1: initializing ant colony, equally dividing each dimension of solution space into
Figure DEST_PATH_IMAGE119
The initial distribution of ants is one per sub-interval, with a total of
Figure 66802DEST_PATH_IMAGE120
A plurality of;
s3.2: calculation of information amount, the
Figure DEST_PATH_IMAGE121
The coordinates of the positions of the ants are
Figure 265702DEST_PATH_IMAGE122
Then the amount of information it carries is calculated as follows:
Figure DEST_PATH_IMAGE123
wherein the content of the first and second substances,
Figure 731318DEST_PATH_IMAGE119
is the peak value of pheromone content between two ants and is an objective function
Figure 52578DEST_PATH_IMAGE124
The reciprocal of (a);
Figure DEST_PATH_IMAGE125
is a compression factor;
Figure 806907DEST_PATH_IMAGE126
is as follows
Figure 606236DEST_PATH_IMAGE128
Only ants and the second
Figure DEST_PATH_IMAGE129
The distance between the ants is only the distance between the ants,
Figure 129621DEST_PATH_IMAGE130
；
s3.3: by means of information distribution functions
Figure DEST_PATH_IMAGE131
And (4) obtaining the distribution condition of the total information quantity of the current ant colony in each subspace through integration, and determining the distribution of the number of ants in each subinterval according to the proportion of the total information quantity of the current ant colony to the total problem sum and the current ant colony scale. The number of ants in each sub-interval is calculated according to the following formula:
Figure 684100DEST_PATH_IMAGE132
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE133
presentation pair
Figure 716603DEST_PATH_IMAGE134
Pheromone progression in dimensional space
Figure 257306DEST_PATH_IMAGE134
And (5) re-integrating the obtained pheromone content sum in the whole space.
S3.4: and determining the moving direction of the ant colony according to the difference between the ant colony distribution condition and the current ant colony distribution condition in each subinterval, moving the ant colony and changing the current ant coordinate. After the ants finish the integral movement once, corresponding information quantity distribution calculation and ant colony moving operation are carried out;
s3.5: calculating an objective function;
s3.6: judging whether algorithm switching is met; if yes, go to step S3.7; if not, go to step S3.2. Considering that the speed of the ant colony algorithm gradually approaching to the optimal solution is reduced during local iteration, the method provided by the invention is based on the optimal target function between two adjacent steps of the ant colony algorithmThe algorithm is switched with a relative value of the numerical drop less than a certain degree, i.e. determined as follows:
Figure DEST_PATH_IMAGE135
wherein the content of the first and second substances,
Figure 963094DEST_PATH_IMAGE136
the value of the objective function is,
Figure DEST_PATH_IMAGE137
in order to be able to perform the number of iterations,
Figure 360577DEST_PATH_IMAGE138
to switch the threshold.
S3.7: the gradient algorithm is characterized in that the gradient method is that the negative gradient direction of an objective function is taken as the search direction of each step of iteration, the optimal step length of the negative gradient direction is taken for each step of iteration, and the iteration formula is determined according to the following formula:
Figure DEST_PATH_IMAGE139
wherein the content of the first and second substances,
Figure 89499DEST_PATH_IMAGE140
as an objective function in
Figure DEST_PATH_IMAGE141
The gradient of the dimensional space is determined,
Figure 230630DEST_PATH_IMAGE142
is the step size factor.
S3.8: judging whether the optimal value is reached; if yes, go to step S3.9; if not, entering step S3.7;
s3.9: and outputting the identification result.
After the actual measurement record of the load dynamic characteristics of a field is obtained through a load recording device, the parameter identification of a comprehensive load model is carried out on the actual measurement data item by utilizing a hybrid optimization algorithm to respectively obtain 14 parameters, and the parameters are established
Figure DEST_PATH_IMAGE143
The load model parameter set of (a), namely:
Figure 790924DEST_PATH_IMAGE144
s4: model parameter optimization
According to what is built
Figure DEST_PATH_IMAGE145
The load model parameter set is estimated by adopting a Kriging algorithm for each row of parameters respectively, so that the purpose of parameter optimization is achieved. The Kriging algorithm is a method for unbiased and optimal estimation of parameter values in a limited region from parameter correlation and variability in a statistical sense, and is a specific sliding weighted average method. For comprehensive load model parameters
Figure 359309DEST_PATH_IMAGE146
And the secondary identification results have correlation, and are suitable for unbiased and optimal estimation by adopting a Kriging algorithm.
For a certain load model parameter variable
Figure DEST_PATH_IMAGE147
To proceed with
Figure 575527DEST_PATH_IMAGE148
The result of this identification is respectively
Figure DEST_PATH_IMAGE149
Then the Kriging algorithm estimates the parameter
Figure 520349DEST_PATH_IMAGE150
Comprises the following steps:
Figure DEST_PATH_IMAGE151
wherein the content of the first and second substances,
Figure 935150DEST_PATH_IMAGE152
the weight coefficient represents the degree of contribution of each recognition value to the estimation value. Two conditions must be satisfied when the weight coefficient is solved, one is to ensure that the estimation is unbiased, namely the mathematical expectation of the deviation is zero; the second is optimal, i.e. the estimated variance is minimal.
Guarantee unbiased, optimal estimation, i.e. guarantee the following:
Figure DEST_PATH_IMAGE153
solving the linear equation set of the above formula to obtain the weight coefficient
Figure 940015DEST_PATH_IMAGE154
And lagrange coefficient
Figure DEST_PATH_IMAGE155
Obtaining an estimated value of a certain load parameter
Figure 440266DEST_PATH_IMAGE156
While the estimated variance is known
Figure DEST_PATH_IMAGE157
Determined as follows:
Figure 860883DEST_PATH_IMAGE158
s5: load feature library update
And (3) respectively carrying out parameter optimization of Kriging algorithm on 14 parameters to be identified of the comprehensive load model, and calculating unbiased and optimal estimated values of all the parameters under the actual measurement data to obtain a group of load model parameters reflecting the statistical characteristics of the actual measurement data. When a plurality of new measured data are recorded on site, curve identification is carried out on the measured data, then the newly identified load model parameters are added into a load parameter set, and the parameter optimization of the Kriging algorithm is carried out again to replace the original model parameters.

Claims (4)

1. A method for measuring, analyzing and identifying load parameters of an electric power system is characterized by comprising the following steps:
s1: determining the installation point of the load recording device according to the clustering analysis of the load measurement information;
s2: the load model adopts a comprehensive load model, and the structure of the load model is a motor parallel static characteristic load model which reflects the dynamic characteristic of the load; the model has 14 parameters, namely the stator resistance of the induction motor
Figure DEST_PATH_IMAGE002
And stator reactance
Figure DEST_PATH_IMAGE004
Excitation reactance
Figure DEST_PATH_IMAGE006
Rotor resistance
Figure DEST_PATH_IMAGE008
And rotor reactance
Figure DEST_PATH_IMAGE010
Time constant of inertia of rotor
Figure DEST_PATH_IMAGE012
Torque equation constants A and B, and proportion of initial active power of induction motor to total initial active power of load
Figure DEST_PATH_IMAGE014
Initial load factor of induction motor
Figure DEST_PATH_IMAGE016
And the proportion coefficient of the active power and the reactive power of the constant impedance and the constant power in the static characteristic load model
Figure DEST_PATH_IMAGE018
For the initial active power of the motor,p is the active power consumed by the measured load point in the transient process;
Figure DEST_PATH_IMAGE020
is a rated initial load factor of the load,
Figure DEST_PATH_IMAGE022
is the rated capacity of the induction motor,
Figure DEST_PATH_IMAGE024
is the reference voltage of the load and is,
Figure DEST_PATH_IMAGE026
is the initial value of the load bus voltage in the transient process;
Figure DEST_PATH_IMAGE028
is an independent parameter vector to be identified of the load model,
Figure DEST_PATH_IMAGE030
to pass steady state conditions of the motor and
Figure DEST_PATH_IMAGE032
the obtained vector of the identification parameter is then calculated,
Figure DEST_PATH_IMAGE034
is the initial transient voltage of the motor and,
Figure DEST_PATH_IMAGE036
is the slip ratio of the motor and is,
Figure DEST_PATH_IMAGE038
for the synchronous reactance between the stator and the rotor,
Figure DEST_PATH_IMAGE040
is the load factor;
s3: carrying out load model parameter identification, and specifically comprising the following steps:
s3.1: obtaining a record of the actual load dynamics at the site by means of a load recording device, from
Figure DEST_PATH_IMAGE042
Start sampling at a time to
Figure DEST_PATH_IMAGE044
The sampling is finished at the moment, and the sampling is counted
Figure DEST_PATH_IMAGE046
Next, the process of the present invention,
Figure DEST_PATH_IMAGE048
the actual measured value of the time is recorded as
Figure DEST_PATH_IMAGE050
Corresponding to the identified load model
Figure DEST_PATH_IMAGE052
The response value at the time is
Figure DEST_PATH_IMAGE054
；
S3.2: setting independent parameter phasors to be identified
Figure DEST_PATH_IMAGE056
The following objective functions are optimized by adopting a hybrid optimization algorithm:
Figure DEST_PATH_IMAGE058
the basic idea of the hybrid optimization algorithm is that the ant colony algorithm and the gradient algorithm are combined, the characteristics that the ant colony algorithm is good in global optimization performance and fast in local convergence of the gradient algorithm are comprehensively utilized, and a new parameter identification algorithm is formed; starting iteration by adopting an ant colony algorithm, selecting a plurality of current superior ants at a certain proper moment, and switching to a gradient algorithm for search iteration until convergence;
information quantity function in ant colony algorithm
Figure DEST_PATH_IMAGE060
And the number of ants due to each subinterval is determined according to the following formula:
Figure DEST_PATH_IMAGE062
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE064
is the peak value of pheromone content between two ants and is an objective function
Figure DEST_PATH_IMAGE066
The reciprocal of (a);
Figure DEST_PATH_IMAGE068
is a compression factor;
Figure DEST_PATH_IMAGE070
the number of ants;
Figure DEST_PATH_IMAGE072
is as follows
Figure DEST_PATH_IMAGE074
Only ants and the second
Figure DEST_PATH_IMAGE076
The distance between the ants is only the distance between the ants,
Figure DEST_PATH_IMAGE078
，
Figure DEST_PATH_IMAGE080
presentation pair
Figure DEST_PATH_IMAGE082
Pheromone progression in dimensional space
Figure 771502DEST_PATH_IMAGE082
The sum of the pheromone contents in the whole space is obtained by the calculus;
the iterative formula of the gradient method is determined as follows:
Figure DEST_PATH_IMAGE084
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE086
as an objective function in
Figure 806323DEST_PATH_IMAGE082
The gradient of the dimensional space is determined,
Figure DEST_PATH_IMAGE088
is a step size factor;
s4: load model parameter optimization
According to what is built
Figure DEST_PATH_IMAGE090
The load model parameter set is respectively estimated for each row of parameters by adopting a Kriging algorithm, so that the aim of parameter optimization is fulfilled;
s5: load feature library update
Respectively carrying out parameter optimization of Kriging algorithm on 14 parameters to be identified of the comprehensive load model, and calculating unbiased and optimal estimated values of all the parameters under actual measurement data to obtain a group of load model parameters reflecting statistical characteristics of the actual measurement data; when a plurality of new measured data are recorded on site, curve identification is carried out on the measured data, then the newly identified load model parameters are added into a load parameter set, and the parameter optimization of the Kriging algorithm is carried out again to replace the original model parameters.
2. The method of claim 1, wherein the load clustering method is a cluster analysis method using an SOM neural network, and the adjustment of the connection weights of winning neurons and their neighbor neurons is determined according to the following formula:
Figure DEST_PATH_IMAGE092
wherein
Figure DEST_PATH_IMAGE094
Represents a learning rate, and
Figure DEST_PATH_IMAGE096
is a function of the field of the winning neuron,
Figure DEST_PATH_IMAGE098
and
Figure DEST_PATH_IMAGE100
respectively the positions of the winning node and other nodes j in the two-dimensional topological space of the output layer;
Figure DEST_PATH_IMAGE102
is the range of the neighborhood.
3. The method as claimed in claim 1, wherein in step S3.2, the switching conditions of different algorithms in the hybrid optimization algorithm are slow, and therefore the algorithm is switched according to the fact that the relative value between the optimal objective function values between two adjacent ant colony algorithms is less than a certain degree, that is, the relative value is determined according to the following formula:
Figure DEST_PATH_IMAGE104
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE106
the value of the objective function is,
Figure DEST_PATH_IMAGE108
in order to be able to perform the number of iterations,
Figure DEST_PATH_IMAGE110
to switch the threshold.
4. The method of claim 1, wherein the method is based on load model parameter optimization
Figure DEST_PATH_IMAGE112
The load model parameter set is respectively estimated for each row of parameters by adopting a Kriging algorithm, so that the aim of parameter optimization is fulfilled; for a certain load model parameter variable
Figure DEST_PATH_IMAGE114
Estimation of Kriging algorithm
Figure DEST_PATH_IMAGE116
Determined as follows:
Figure DEST_PATH_IMAGE118
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE120
the weight coefficient represents the contribution degree of each identification value to the estimation value, and in order to ensure that the estimation is unbiased and optimal,
Figure 235949DEST_PATH_IMAGE120
the value of (a) is determined by:
Figure DEST_PATH_IMAGE122
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE124
lagrange coefficients.
CN201911307051.8A 2019-12-18 2019-12-18 Power system measurement and load parameter analysis and identification method Withdrawn CN110932274A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911307051.8A CN110932274A (en) 2019-12-18 2019-12-18 Power system measurement and load parameter analysis and identification method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911307051.8A CN110932274A (en) 2019-12-18 2019-12-18 Power system measurement and load parameter analysis and identification method

Publications (1)

Publication Number Publication Date
CN110932274A true CN110932274A (en) 2020-03-27

Family

ID=69864216

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911307051.8A Withdrawn CN110932274A (en) 2019-12-18 2019-12-18 Power system measurement and load parameter analysis and identification method

Country Status (1)

Country Link
CN (1) CN110932274A (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111725804A (en) * 2020-06-10 2020-09-29 国网新疆电力有限公司 Comprehensive load model parameter identification method and computer readable storage medium
CN111835015A (en) * 2020-08-15 2020-10-27 杨银平 Power equipment coordination control system
CN111835016A (en) * 2020-08-15 2020-10-27 杨银平 Power grid instability analysis method
CN111835014A (en) * 2020-08-15 2020-10-27 杨银平 Multi-node power grid voltage stabilization control system
CN112039051A (en) * 2020-06-30 2020-12-04 长沙理工大学 Real-time modeling method for accessing double-fed wind driven generator into substation bus load
CN112419093A (en) * 2020-11-26 2021-02-26 清华大学 Load model characteristic parameter extraction method and device based on clustering algorithm
CN112836320A (en) * 2020-12-15 2021-05-25 国网江苏省电力有限公司 Power load online modeling correction method
CN112883632A (en) * 2021-01-13 2021-06-01 合肥工业大学 Lithium battery equivalent circuit model parameter identification method based on improved ant colony algorithm
WO2021243930A1 (en) * 2020-06-05 2021-12-09 中国电力科学研究院有限公司 Method for identifying composition of bus load, and machine-readable storage medium

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110125685A1 (en) * 2009-11-24 2011-05-26 Rizvi Syed Z Method for identifying Hammerstein models
CN103872678A (en) * 2014-03-06 2014-06-18 国家电网公司 Load model identification method based on transformer substation measurement
CN104880992A (en) * 2014-12-23 2015-09-02 湘潭大学 Magnetic-control arc sensor parameter optimization method based on Kriging surrogate model
CN106055918A (en) * 2016-07-26 2016-10-26 天津大学 Power system load data identification and recovery method
CN106650206A (en) * 2016-09-28 2017-05-10 湖南老码信息科技有限责任公司 Prediction method of high blood pressure based on incremental neural network model and prediction system

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110125685A1 (en) * 2009-11-24 2011-05-26 Rizvi Syed Z Method for identifying Hammerstein models
CN103872678A (en) * 2014-03-06 2014-06-18 国家电网公司 Load model identification method based on transformer substation measurement
CN104880992A (en) * 2014-12-23 2015-09-02 湘潭大学 Magnetic-control arc sensor parameter optimization method based on Kriging surrogate model
CN106055918A (en) * 2016-07-26 2016-10-26 天津大学 Power system load data identification and recovery method
CN106650206A (en) * 2016-09-28 2017-05-10 湖南老码信息科技有限责任公司 Prediction method of high blood pressure based on incremental neural network model and prediction system

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
秦川 等: "电力负荷模型参数辨识的混合优化算法" *
高铭 等: "支持向量机在地统计学中的应用研究" *

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2021243930A1 (en) * 2020-06-05 2021-12-09 中国电力科学研究院有限公司 Method for identifying composition of bus load, and machine-readable storage medium
CN111725804A (en) * 2020-06-10 2020-09-29 国网新疆电力有限公司 Comprehensive load model parameter identification method and computer readable storage medium
CN112039051A (en) * 2020-06-30 2020-12-04 长沙理工大学 Real-time modeling method for accessing double-fed wind driven generator into substation bus load
CN111835015A (en) * 2020-08-15 2020-10-27 杨银平 Power equipment coordination control system
CN111835016A (en) * 2020-08-15 2020-10-27 杨银平 Power grid instability analysis method
CN111835014A (en) * 2020-08-15 2020-10-27 杨银平 Multi-node power grid voltage stabilization control system
CN111835016B (en) * 2020-08-15 2021-11-02 杨银平 Power grid instability analysis method
CN112419093A (en) * 2020-11-26 2021-02-26 清华大学 Load model characteristic parameter extraction method and device based on clustering algorithm
CN112836320A (en) * 2020-12-15 2021-05-25 国网江苏省电力有限公司 Power load online modeling correction method
CN112836320B (en) * 2020-12-15 2022-10-25 国网江苏省电力有限公司 Online modeling correction method for power load
CN112883632A (en) * 2021-01-13 2021-06-01 合肥工业大学 Lithium battery equivalent circuit model parameter identification method based on improved ant colony algorithm

Similar Documents

Publication Publication Date Title
CN110932274A (en) Power system measurement and load parameter analysis and identification method
CN108155648A (en) Method for estimating state based on the infinite Extended Kalman filter of adaptive H
CN112510707B (en) Power distribution station power impedance topological graph generation method and system
Nichiforov et al. Deep learning techniques for load forecasting in large commercial buildings
CN103544525A (en) Method for identifying parameters of synchronous wind-driven generators on basis of improved particle swarm optimization algorithm
Dong et al. Applying the ensemble artificial neural network-based hybrid data-driven model to daily total load forecasting
CN110380444B (en) Capacity planning method for distributed wind power orderly access to power grid under multiple scenes based on variable structure Copula
CN110837915B (en) Low-voltage load point prediction and probability prediction method for power system based on hybrid integrated deep learning
CN107103184A (en) A kind of high-voltage cable joint temperature predicting method
Xuemei et al. Particle swarm optimization-based LS-SVM for building cooling load prediction
Han et al. Multi‐objective robust dynamic VAR planning in power transmission girds for improving short‐term voltage stability under uncertainties
CN112994017A (en) Distributed photovoltaic power supply site selection optimization method based on power distribution network probability load flow calculation
CN104811336A (en) Intelligent optimization-based distributed network flow predicting method
Dong et al. Short-term building cooling load prediction model based on DwdAdam-ILSTM algorithm: A case study of a commercial building
Yang et al. Multi-objective optimization of contactor’s characteristics based on RBF neural networks and hybrid method
CN113177600B (en) Adaptive robust state estimation method for power system
CN112446175B (en) Distributed power supply site selection optimization method based on power distribution network probability load flow calculation
Jiang et al. Two‐level area‐load modelling for OPF of power system using reinforcement learning
CN115688535A (en) Power data combined interpolation method and system based on waveform similarity analysis
Nichiforov et al. Efficient Load Forecasting Model Assessment for Embedded Building Energy Management Systems
Shen et al. A hybrid algorithm based on beluga whale optimization-forgetting factor recursive least square and improved particle filter for the state of charge estimation of lithium-ion batteries
CN113361187A (en) Power grid line operation safety optimization method, system and storage medium
Matsumoto et al. Connection Phase Estimation of Pole Mounted Distribution Transformers by Integer Form of Population Based Incremental Learning Considering Measurement Errors and Outliers by Correntropy
Ming et al. Electrical load prediction in energy internet via linear correlation coefficient approach
CN117277446B (en) Multi-target power distribution network planning method and 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
WW01 Invention patent application withdrawn after publication

Application publication date: 20200327

WW01 Invention patent application withdrawn after publication