CN115296309A - Wind, light, water, fire and storage combined secondary frequency modulation method based on real-time inertia estimation - Google Patents
Wind, light, water, fire and storage combined secondary frequency modulation method based on real-time inertia estimation Download PDFInfo
- Publication number
- CN115296309A CN115296309A CN202211226765.8A CN202211226765A CN115296309A CN 115296309 A CN115296309 A CN 115296309A CN 202211226765 A CN202211226765 A CN 202211226765A CN 115296309 A CN115296309 A CN 115296309A
- Authority
- CN
- China
- Prior art keywords
- region
- frequency
- inertia
- power
- wind
- 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
Links
Images
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/24—Arrangements for preventing or reducing oscillations of power in networks
- H02J3/241—The oscillation concerning frequency
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- H02J3/28—Arrangements for balancing of the load in a network by storage of energy
- H02J3/32—Arrangements for balancing of the load in a network by storage of energy using batteries with converting means
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
- H02J3/381—Dispersed generators
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
- H02J3/46—Controlling of the sharing of output between the generators, converters, or transformers
- H02J3/466—Scheduling the operation of the generators, e.g. connecting or disconnecting generators to meet a given demand
-
- 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
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- 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
- H02J2300/00—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation
- H02J2300/20—The dispersed energy generation being of renewable origin
- H02J2300/22—The renewable source being solar energy
- H02J2300/24—The renewable source being solar energy of photovoltaic origin
-
- 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
- H02J2300/00—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation
- H02J2300/20—The dispersed energy generation being of renewable origin
- H02J2300/28—The renewable source being wind energy
-
- 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
- H02J2300/00—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation
- H02J2300/40—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation wherein a plurality of decentralised, dispersed or local energy generation technologies are operated simultaneously
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses a wind, light, water, fire and storage combined secondary frequency modulation method based on real-time inertia estimation, which comprises the following steps of: establishing a frequency modulation model of each device of the wind, light, water and fire storage, and establishing a tie line and a frequency response model thereof; establishing a regional frequency modulation state space equation model containing wind, light, water and fire storage; designing a virtual inertia control link, and designing a closed-loop controller of the regional wind, light, water and fire storage secondary robust frequency controller meeting the given performance index on the basis; a real-time inertia estimation method of an interconnected power system based on multivariate random forest regression is provided. The invention has the beneficial effects that: after the scheme is adopted, the method can be used for adjusting the system frequency aiming at the dynamic performance of the system frequency under large disturbance and the like, effectively reduces the influence of the change of the system operation condition and the inertia lack on the secondary frequency modulation of the new energy unit participating in the power grid, and integrally improves the frequency modulation performance of the system.
Description
Technical Field
The invention belongs to the technical field of power grid frequency control, and particularly relates to a wind, light, water, fire and storage combined secondary frequency modulation method based on real-time inertia estimation.
Background
The traditional secondary frequency modulation of the power system means that a generator set provides enough adjustable capacity and a certain adjusting rate, and the frequency is tracked in real time under the allowable adjusting deviation so as to meet the requirement of system frequency stability. The secondary frequency modulation can realize the frequency adjustment without difference and can monitor and adjust the power of the tie line. With the gradual increase of the proportion of new energy, stored energy and other resources accessed to the power grid, the frequency change caused by a large amount of active power transmitted to the power grid needs to be improved by effectively controlling and utilizing the new energy.
However, the renewable power generation has strong uncertainty and volatility, and the current renewable energy power prediction accuracy cannot meet the operation requirement, so that the safety and quality of the power grid frequency are seriously affected. The frequency anomaly will have serious consequences for the safe operation of the generator and the system and for the user. Therefore, the frequency must be controlled efficiently. At present, most of traditional frequency modulation units are thermal power units and hydroelectric power units, the units have certain inherent defects, for example, when participating in secondary frequency modulation, in the face of access of a large number of new energy stations and stored energy into a power grid, frequency control operation of the traditional power grid faces various more complex problems, the current secondary frequency modulation scheme is generally based on traditional control, a power electronic inverter of the current secondary frequency modulation scheme has the characteristics of low inertia and small damping, but large load sudden change can cause the frequency and voltage of the power grid to deviate from an allowable range, and potential risks are brought to the system. With the increase of the new energy input, the generated power of the synchronous generator is reduced, the rotational inertia of the system is reduced, when unbalanced power occurs, the frequency characteristic of the system is deteriorated, and the frequency fluctuation is increased. Therefore, how to estimate the rotational inertia of the system on line based on the frequency constraint condition and then perform early warning on the bearing capacity of the new energy according to the equivalent rotational inertia of the system is a technical problem to be solved urgently. And through real-time inertia estimation, the problems of low inertia and small damping of the micro-grid can be solved, and strong frequency support is provided for the power grid.
Therefore, the scheduling method which can consider the peak regulation capability of a fan, photovoltaic, hydroelectric and energy storage and reasonably select a new energy unit to participate in system peak regulation so as to ensure the economical efficiency and reliability of the operation of the power system is very significant is provided for solving the problems that the wind power plant is influenced by wind speed, the load fluctuation is frequent, the frequency response speed is slow and the like in the secondary frequency modulation of the existing power grid. The key point for solving the problems is to establish a combined secondary frequency modulation method based on real-time inertia estimation and a robust cooperative frequency modulation model covering five types of energy sources of wind, light, water, fire and storage.
Aiming at the participation of new energy in frequency modulation, single station level control is mainly used at present, automatic power generation control and local frequency signals of a power grid are used as the basis, control design is carried out through a local traditional classical control theory of a station, and the effects of power distribution, frequency deviation reduction, power grid inertia improvement, power grid safe and stable operation capacity improvement and the like are achieved.
In the face of access of a large number of well-spraying type new energy stations and stored energy into a power grid, frequency control operation of a traditional power grid faces various more complex problems, the current secondary frequency modulation scheme generally only considers the utilization of a special communication line for signal transmission, and does not fully consider the inevitable cooperative control problem of a large number of frequency modulation resources such as the new energy stations and the stored energy in secondary frequency modulation along with the continuous promotion of electric power marketization, and the frequency control problem is not favorable for stable operation of the frequency of the whole power grid when the frequency modulation resources participate in secondary frequency modulation of the power grid. The existing scheme also establishes the relation between the virtual inertia parameters and the real-time inertia and the given frequency target of the interconnected power system less. As a result, when the system inertia changes or disturbances occur, the frequency modulation performance will be affected, and therefore adaptive virtual inertia parameters need to be established to avoid the problem.
Disclosure of Invention
The invention aims to provide a wind, light, water, fire and storage combined secondary frequency modulation method based on real-time inertia estimation, which can be used for adjusting system frequency aiming at the dynamic performance, dynamic out-of-limit risk and the like of the system frequency under large disturbance, effectively reduces the influence of the change of the system operation condition and inertia deficiency on the secondary frequency modulation of a new energy unit participating in power grid, and integrally improves the frequency modulation performance of the system.
The technical scheme of the invention is as follows: a wind, light, water, fire and storage combined secondary frequency modulation method based on real-time inertia estimation is characterized in that a secondary frequency modulation state space model of a system is established, a virtual inertia control link and a robust frequency controller closed-loop controller are designed, virtual inertia parameters are adjusted in a self-adaptive mode through real-time inertia estimation, and dynamic performance of system frequency under large disturbance can be effectively enhanced aiming at actual scenes such as system operation condition change and inertia deficiency; the method comprises the following steps:
and 6, when the operation condition and inertia of the interconnected power system change, changing the virtual inertia parameters through self-adaptation to meet the frequency modulation performance of the system, and reducing the risk of system frequency out-of-limit under a large disturbance accident.
Further, in step 1, the hydroelectric power generating unit model comprises:
a speed regulator model:
in the formula (I), the compound is shown in the specification,iis a region number, ΔL gi Is a regioniThe position variation of the speed regulator valve of the thermal power generating unit,sin order to be a differential operator, the method comprises the following steps of,T gi is a regioniThe time constant of the speed regulator of the thermal power generating unit,γ kGi is a regioniPower signal distribution coefficient, delta, of thermal power generating unitsP ci Is a regioniThe control signals of the system are sent to the system,K SAGi is a regioniSag factor, Δ, of thermal power generating unitsf i Is a regioniFrequency of (a)u Gi A control signal is sent to the thermal power generating unit;
the turbine model is as follows:
in the formula,. DELTA.P mi Is a regioniThermal power generating unitThe amount of change in the output power is,Y chi is a regioniThe time constant of the thermal power unit turbine,sis a differential operator, the same is carried out below;
for the wind power station model, a variable-speed wind turbine generator is adopted to participate in system frequency regulation, and the simplified model is expressed as follows:
in the formula (I), the compound is shown in the specification,indicating areaiThe variation of the rotating speed of the rotor of the fan,J Wti is a regioniThe comprehensive inertia coefficient of the fan is obtained,N gi is a regioniThe ratio of the fan to the gear box,indicating areaiVariation of pitch angle, Δ, of fanu Wi Indicating areaiA control signal of the wind power station is sent,α kWi is a regioniPower signal distribution coefficient, delta, of a wind farmP ci Is a regioniSystem control signal, Δv Wmi Is a regioniThe wind speed variation of (2);
wherein the content of the first and second substances,differential of (2)Can be expressed as follows:
in the formula (I), the compound is shown in the specification,andK piI respectively representing regionsiProportional and integral coefficients of the fan PI controller,K ci indicating areaiThe correction coefficient of the fan is changed according to the change of the fan,T gi indicating areaiThe mechanical torque of the fan is changed,indicating areaiThe variation of the rotating speed of the fan generator;
the photovoltaic power station active power reduction is adopted to participate in system frequency adjustment, the photovoltaic power station is equivalent to a first-order inertia link, and a dynamic response model is as follows:
in the formula,. DELTA.P iPV Is a regioniThe photovoltaic power station outputs the variation of active power,Y iPV is a regioniResponse time constant, Δ, of photovoltaic plantu iPV Indicating areaiA control signal of the photovoltaic station is sent,γ iPV is a regioniThe power signal distribution coefficient of the photovoltaic station;
for the energy storage power station, the transfer function in the energy storage power station is equivalent to a first-order inertia link as follows:
in the formula,. DELTA.P iBESS Is a regioniThe energy storage power station outputs the variation of active power,Y iBESS indicating areaiResponse time constant, Δ, of energy storage power stationc Bi Is a regioniA control signal of the energy storage power station,γ Bi is a regioniThe power signal distribution coefficient of the energy storage power station;
state of charge of energy storage batterySOCThe operating state and the regulation and control capacity of the energy storage unit are estimated by adopting an ampere-hour integration methodSOCThe calculation formula is as follows:
in the formula (I), the compound is shown in the specification,f(SOC i (t)) representstTime zoneiState of charge of energy storage power stationSOC,f(SOC i0 ) Is a regioniInitial state of charge of energy storage power stationSOC,For the power loss factor of the energy storage power station,P iBESS is a regioniThe energy storage power station outputs active power,S icap, is a regioniRated capacity of the energy storage power station;
after the frequency modulation model of each device is established, a tie line model is established as follows:
in the formula,. DELTA.P tie,i Is an implantation regioniTotal tie line power, ΔP tie,ij Is a regioniAndjthe power of the interconnect link of (a),T ij is a regioniAnd areajInterconnection gain, Δf i And Δf j Are respectively regionsiAndjthe frequency of (a) of (b) is,sin order to be a differential operator, the method comprises the following steps of,Nis the number of regions, ΔACE i Is a regioniThe control error of (2) is determined,β i is a regioniFrequency deviation factor of, ΔP ci Is a regioniA system control signal;
establishing a frequency response model of the regional power grid as follows:
in the formula,. DELTA.f i Is a regioniThe frequency of (a) of (b) is,M INEi is a regioniThe coefficient of inertia is determined by the measured value of the mass,D i is a regioniDamping coefficient, ΔP mi Is a regioniVariation of output power, delta, of thermal power generating unitsP wi Is a regioniVariation of output power, delta, of wind turbineP iPV Is a regioniVariation of output power, delta, of a photovoltaic power stationP iBESS Is a regioniOutput active power variation, delta, of energy storage power stationP di Is a regioniAmount of change in load, ΔP tie,i Is an implantation regioniTotal tie line power.
Further, in the step 2, a regional frequency modulation state space equation model including wind, light, water and fire storage is established as follows:
in the formula (I), the compound is shown in the specification,x(t) Is composed oftThe overall state vector of the time system is,A sys is a matrix of the whole system, and the system,Bis an integral control matrix and is provided with a plurality of control matrixes,u(t) Is composed oftThe time system is used for controlling the vector as a whole,B ω in the form of an overall perturbation matrix,ω(t) Is composed oftThe integral disturbance vector of the time system;
the overall vector contains the following specific quantities:
in the formula (I), the compound is shown in the specification,the state vector of the variation of the rotating speed of the fan rotor is represented,a state vector representing the variation of the fan generator speed,representing state of variation of pitch angle of fanVector, ΔP m For the output power variation state vector, delta, of the thermal power generating unitP g Adjusting a state vector, delta, of a change in valve position for a thermal power plant governorP BESS Outputting active power variation state vector, delta, for energy storage power stationSOCRepresenting the state of charge, Δ, of an energy storage power stationfFor each of the region frequency state vectors,integrating the state vector, Δ, for the region control errorP tie Is the total tie-line power state vector, Δ, of the implanted regionu W Representing control vectors, Δ, of wind power plant control signalsu G Control vector, delta, for thermal power plant control signalsu B Controlling the vector, Δ, for the control signal of the energy-storing power stationP d For disturbance vectors of load variation, Δv m For wind speed disturbance vector, superscriptTRepresenting a transposition;
the characteristic equation of the overall system is as follows:
in the formula (I), the compound is shown in the specification,for the characteristic equation function representation, det represents determinant,A sys is a matrix of the whole system and is,Ithe unit matrix is represented by a matrix of units,are the frequency domain coefficients of the characteristic polynomial,nis the number of characteristic polynomials.
Further, in step 3, the model of the virtual inertia control link is represented by the following equation:
in the formula,. DELTA.P ref (s) Being reference to inverting unitssThe transformation function represents the energy shortage of the system obtained by current calculation, the output of the transformation function is transmitted to the coordinated frequency modulation units such as the wind, light, water, fire and storage and the like in a signal form,H vi andD vi is a virtual rotor parameter, Δf(s) Is the system frequency variation deltafThe laplace transform function of;
the overall transfer function is expressed as:
in the formula (I), the compound is shown in the specification,R i andK int respectively, a virtual droop constant and a virtual integral gain;
based on the virtual inertial transfer function, the frequency outer loop is designed as follows:
in the formula (I), the compound is shown in the specification,for the description of the frequency variation of the system,is the nominal frequency of the system and is,J p andD p virtual inertia and damping coefficient, Δ, of the system, respectivelyP R For wind, light, water and fire storage of active power regulating quantity, deltaP unc For uncontrolled disturbances inside and outside the system, ΔPThe system active power output quantity is obtained;
on the basis, a state feedback robust controller is further designedM=KxTo ensure that the system is disturbed by the outsideiStability under the condition and maximally reducing disturbance to system regulated output functionyThe closed loop stability control of the frequency is realized, and the corresponding closed loop system is as follows:
in the formula (I), the compound is shown in the specification,y(k) Is composed ofkThe output function of the time of day system,Cin order to output the matrix for the state,D ω in order to perturb the output matrix,Kin order to achieve the gain,x(k) Is composed ofkThe state variable input by the system at the moment,Din order to control the output matrix,i(k) Is composed ofkDisturbance variables of the time system;
robust gain matrixKThe design method is as follows:
the robust performance constraint inequality of the system is:
wherein the content of the first and second substances,Ain the form of a matrix of parameters,WandXis a positive definite matrix to be solved, T represents the matrix transposition,Gin the form of a matrix of parameters,a parameter matrix is designed for the controller,γ 1 for the controller transfer function minimum rejection ratio,Iis a matrix of the units,Zin order for the controller to evaluate the output matrix,C 2 designing a parameter matrix for the controller;
positive definite matrix if optimal solution existsWAndXthe system is progressively stabilized and the gain parameter is controlled to beK=X(W) -1 I.e. byu(t)=X(W) -1 xFor co-operating frequency modulationA controller; the linear matrix inequality is used for solving the control parameters, so that the solving complexity is reduced, and the optimal control parameters are ensured.
Further, in step 4, a real-time inertia estimation method of the interconnected power system based on multivariate random forest regression is provided, and PMU data and environmental information are used for estimating the real-time inertia of the power system;
training a multivariate regression model for system inertia estimation using available inertial data, load curves, features extracted in environmental frequency measurements at different locations, and weather data; for applications with large amounts of training data, system inertia is estimated using multivariate random forest regression, MRFR, as a machine learning model;
MRFR is a set of regression trees trained by guided sampling and random feature selection, and it mainly comprises the following steps:
1) From training sample setsSIn the random extractionmObtaining a new sample pointS 1 -S n A training subset;
2) Training a CART regression tree by using training subsets, wherein in the training process, the segmentation rule of each node is that all the characteristics are randomly selectedkIs then characterized bykSelecting an optimal cutting point from the characteristics to divide left and right subtrees;
3) Obtaining a plurality of CART regression tree models through the step 2), wherein the final prediction result of each CART regression tree is the mean value from the sample point to the leaf node;
4) The prediction result of the multivariate random forest is the average value of the prediction results of all CART regression trees;
the use of MRFR includes both offline training, where the MRFR is trained using available offline data, and online applications, where the trained MRFR will receive online measurements and extracted features and use them to estimate the total inertia of the power system.
Further, in step 5, based on a deep neural network method, establishing an input-output relationship between a gain parameter of the controller and a corresponding target of real-time inertia and a given frequency of the interconnected power system;
after the real-time inertia is obtained through estimation, a key problem is that the virtual inertia required by an area is accurately calculated under the condition of giving the real-time inertia and the frequency response target of an interconnected power system, and because the relation between the real-time working condition of the system and the lowest point of the frequency response is nonlinear, the relation among the total inertia, the area virtual inertia and the lowest point of the system frequency of the interconnected system is accurately modeled by using a machine learning method such as a deep neural network method;
input data of the neural network are the total inertia of the interconnected system and the system frequency lowest point target, and are converted into a two-dimensional mapping and pooling layer in the convolutional layer, and finally a complete connection layer comprises an output control signal and a regional virtual inertia regulation value. The convolutional layer comprises a plurality of convolution kernels, each element corresponds to a weight coefficient and a deviation value, the size of each element is determined by the convolution kernels, and the convolutional layer can be used for extracting the characteristics of normalized input data and is described as follows:
in the formula (I), the compound is shown in the specification,O n j, the output of the convolutional layer is shown,fin order to map the layers of the image,U i n,-1 represents a parameterized spatial filter, the specific actions of which are automatically learned from the data during the training of the network,M n ij, which represents the input of the convolutional layer,b n j, in the case of a real number bias term,represents a cross product;
the function of the pooling layer is to select features and filter layers of information from the convolved output data, which is expressed as follows:
in the formula (I), the compound is shown in the specification,Q n+1,j the water is output from the pool layer,gin order to pool the mapping, the mapping is performed,k n+1,j andr n+1,j respectively representing the multiplicative deviation and the additive deviation,representing a pooling function;
the filter layer can thus be expressed as:
in the formula (I), the compound is shown in the specification,p、qrespectively representing the length and width of the filter layer; the pooling operation is to reduce the output of the upper layers, which makes the transformation of the predicted input data more robust,J i,j is output for the filtering layer, and is output by the filtering layer,S p,q in order to filter the parameters of the filter,b i+p ,j+q--11 is a real number bias term;
finally, the complete connection layer can expand the feature map into vectors and output values of each complete connection layer; passing to the output requires defining a loss function to accomplish this classification; the fully connected layer performs a non-linear combination, and this process is described as follows:
in the formula (I), the compound is shown in the specification,Y n represents the output of the fully-connected layer,fa non-linear function representing a fully connected layer,w n andp n respectively representing the weight and deviation of the fully connected layer;
for a given system, simulation data or historical data can be used for training a machine learning model, and the trained machine learning model is an adaptive model and realizes accurate mapping of input and output; and estimating a virtual inertia setting value of the target with the lowest point of the given frequency by using information such as a pre-training model and system inertia.
The invention has the beneficial effects that: after the scheme is adopted, the method can be used for adjusting the system frequency aiming at the dynamic performance of the system frequency under large disturbance and the like, effectively reduces the influence of the change of the system operation condition and the inertia lack on the secondary frequency modulation of the new energy unit participating in the power grid, and integrally improves the frequency modulation performance of the system. The invention provides a wind, light, water, fire and storage combined secondary frequency modulation method based on real-time inertia estimation. After the wind power station and the photovoltaic station are connected to a power grid, although active output can be carried out to participate in frequency adjustment, the active output cannot be stably output due to the influence of environmental uncertainty such as wind speed/illumination and large disturbance scenes, the influence of the change of the active output on the participation of a new energy unit in the secondary frequency modulation of the power grid can be effectively reduced, and the frequency modulation performance of the system is integrally improved. The real-time inertia estimation method of the interconnected power system based on the multivariate random forest regression is provided, and the real-time inertia of the power system is estimated by utilizing PMU data and environmental information, so that the real-time inertia of the system can be adjusted and compensated. The invention also establishes the input-output relationship between the virtual inertia parameters and the real-time inertia of the interconnected power system and the corresponding target (such as the lowest frequency point) of the given frequency by using a deep neural network method, thereby formulating the adjustment rule of the virtual inertia.
Drawings
FIG. 1 is a flow chart of wind, light, water, fire and storage combined secondary frequency modulation control design.
Fig. 2 is a graph showing frequency changes before and after zone control in the presence of large disturbances.
Detailed Description
As shown in fig. 1 and fig. 2, the invention is operated and implemented in such a way that a wind, light, water, fire and storage combined secondary frequency modulation method based on real-time inertia estimation can be expressed as follows:
and 6, when the operation condition and inertia of the interconnected power system change, the virtual inertia parameters can be changed in a self-adaptive manner to meet the frequency modulation performance of the system, and the system frequency out-of-limit risk under a large disturbance accident is reduced.
In the step 1, establishing a frequency modulation model of each device of the wind, light, water and fire storage; establishing a tie line and a frequency response model thereof;
the frequency modulation model of each device is specifically as follows:
a hydroelectric-thermal power unit model comprising:
a speed regulator model:
in the formula (I), the compound is shown in the specification,iis a region number, ΔL gi Is a regioniThe position variation of the speed regulator valve of the thermal power generating unit,sin order to be a differential operator, the system is,T gi is a regioniThe time constant of the speed regulator of the thermal power generating unit,γ kGi is a regioniPower signal distribution coefficient, delta, of thermal power generating unitsP ci Is a regioniThe control signal of the system is sent to the computer,K SAGi is a regioniSag factor, Δ, of thermal power generating unitsf i Is a regioniFrequency of (a)u Gi A control signal is sent to the thermal power generating unit;
the turbine model is as follows:
in the formula,. DELTA.P mi Is a regioniThe output power variable quantity of the thermal power generating unit,Y chi is a regioniThe time constant of the thermal power unit turbine,sis a differential operator, the same is carried out below;
for the wind power station model, a variable-speed wind turbine generator is adopted to participate in system frequency regulation, and the simplified model is expressed as follows:
in the formula (I), the compound is shown in the specification, i indicating areaiThe variation of the rotating speed of the fan rotor,J Wti is a regioniThe comprehensive inertia coefficient of the fan is as follows,N gi is a regioniThe ratio of the fan gearbox to the fan gearbox,indicating areaiVariation of pitch angle, Δ, of fanu Wi Indicating areaiA control signal of the wind power station is sent,α kWi is a regioniPower signal distribution coefficient, delta, of a wind farmP ci Is a regioniSystem control signal, Δv Wmi Is a regioniThe wind speed variation of (2);
wherein, the first and the second end of the pipe are connected with each other,is differentiated byCan be expressed as follows:
in the formula (I), the compound is shown in the specification,andK piI respectively represent regionsiThe proportional and integral coefficients of the fan PI controller,K ci indicating areaiThe correction coefficient of the fan is changed according to the change of the fan,T gi indicating areaiThe mechanical torque of the fan is changed into the torque,indicating areaiThe variation of the rotating speed of the fan generator;
the photovoltaic power station active power reduction is adopted to participate in system frequency adjustment, the photovoltaic power station is equivalent to a first-order inertia link, and a dynamic response model is as follows:
in the formula,. DELTA.P iPV Is a regioniThe photovoltaic power station outputs the variation of active power,Y iPV is a regioniResponse time constant, Δ, of photovoltaic plantu iPV Indicating areaiA control signal of the photovoltaic station is sent,γ iPV is a regioniA photovoltaic station power signal distribution coefficient;
for the energy storage power station, the transfer function in the energy storage power station is equivalent to a first-order inertia link as follows:
in the formula,. DELTA.P iBESS Is a regioniThe energy storage power station outputs the variation of active power,Y iBESS indicating areaiResponse time constant, Δ, of energy storage power stationc Bi Is a regioniA control signal of the energy storage power station,γ Bi is a regioniThe power signal distribution coefficient of the energy storage power station;
state of charge of energy storage batterySOCThe operating state and the regulation and control capacity of the energy storage unit are estimated by adopting an ampere-hour integration methodSOCThe calculation formula is as follows:
in the formula (I), the compound is shown in the specification,f(SOC i (t)) representstTime zoneiState of charge of energy storage power stationSOC,f(SOC i0 ) Is a regioniInitial state of charge of energy storage power stationSOC,For the power loss coefficient of the energy storage power station,P iBESS is a regioniThe energy storage power station outputs active power,S icap, is a regioniRated capacity of the energy storage power station;
after the frequency modulation model of each device is established, a tie line model is established as follows:
in the formula,. DELTA.P tie,i Is an implantation regioniTotal tie line power, ΔP tie,ij Is a regioniAndjthe power of the interconnect link of (a),T ij is a regioniAnd areajInterconnection gain, Δf i And Δf j Are respectively regionsiAndjthe frequency of (a) of (b) is,sin order to be a differential operator, the method comprises the following steps of,Nis the number of regions, ΔACE i Is a regioniThe control error of (2) is set,β i is a regioniFrequency deviation factor of, ΔP ci Is a regioniA system control signal;
the frequency response model of the regional power grid is established as follows:
in the formula,. DELTA.f i Is a regioniThe frequency of (a) of (b) is,M INEi is a regioniThe coefficient of inertia is a function of the mass of the motor,D i is a regioniDamping coefficient, ΔP mi Is a regioniVariation of output power, delta, of thermal power generating unitsP wi Is a regioniVariation of output power, delta, of wind turbineP iPV Is a regioniVariation of output power, delta, of a photovoltaic power stationP iBESS Is a regioniVariation of output active power, delta, of energy storage power stationP di Is a regioniAmount of change in load, ΔP tie,i Is an implantation regioniTotal tie line power.
In the step 2, a regional frequency modulation state space equation model containing wind, light, water and fire storage is established;
the regional frequency modulation state space equation model containing the wind, light, water and fire storage is established as follows:
in the formula (I), the compound is shown in the specification,x(t) Is composed oftThe overall state vector of the time system is,A sys is a matrix of the whole system and is,Bis an integral control matrix and is provided with a plurality of control matrixes,u(t) Is composed oftThe time of day is the overall control vector of the system,B ω in the form of an overall perturbation matrix,ω(t) Is composed oftThe integral disturbance vector of the time system;
the overall vector contains the following specific quantities:
in the formula (I), the compound is shown in the specification,the state vector of the variation of the rotating speed of the fan rotor is shown, f a state vector representing the variation of the fan generator speed,state vector, Δ, representing variation of pitch angle of a fanP m For the output power variation state vector, delta, of the thermal power generating unitP g For speed regulator of thermal power generating unitAdjusting the state vector of the change in valve position, ΔP BESS Outputting active power variation state vector, delta, for energy storage power stationSOCRepresenting the state of charge vector, Δ, of an energy storage power stationfFor each of the region frequency state vectors,integrating the state vector, Δ, for the region control errorP tie Is the total tie-line power state vector, Δ, of the injection regionu W Representing control vectors, Δ, of wind power plant control signalsu G Control vector, delta, for thermal power generating unitsu B For controlling vectors, Δ, for energy-storing power station control signalsP d For disturbance vectors of load variation, Δv m For wind speed disturbance vector, superscriptTRepresenting a transpose;
the characteristic equation of the overall system is as follows:
in the formula (I), the compound is shown in the specification,for the characteristic equation function representation, det represents determinant,A sys is a matrix of the whole system, and the system,Ithe unit matrix is represented by a matrix of units,are the frequency domain coefficients of the characteristic polynomial,nis the number of characteristic polynomials.
In the step 3, designing a virtual inertia control link, and designing a closed-loop controller of the regional wind, light, water and fire storage secondary robust frequency controller meeting the given performance index on the basis;
the model of the virtual inertial control element is represented by the following equation:
in the formula,. DELTA.P ref (s) Being reference to inverting unitssThe transformation function represents the energy shortage of the system obtained by current calculation, the output of the transformation function is transmitted to the coordinated frequency modulation units such as wind, light, water, fire storage and the like in a signal form,H vi andD vi is a virtual rotor parameter, Δf(s) Is the system frequency variation deltafThe laplace transform function of (a);
the overall transfer function is expressed as:
in the formula (I), the compound is shown in the specification,R i andK int respectively, a virtual droop constant and a virtual integral gain;
based on the virtual inertial transfer function, the frequency outer loop is designed as follows:
in the formula (I), the compound is shown in the specification,for the description of the frequency variation of the system,in order to be the nominal frequency of the system,J p andD p virtual inertia and damping coefficient, respectively, of the systemP R For wind, light, water and fire storage of active power regulating quantity, deltaP unc For uncontrolled disturbances inside and outside the system, ΔPThe system active power output quantity is obtained;
on the basis, a state feedback robust controller is further designedM=KxTo ensure that the system is disturbed by the outsideiStability under the condition of the condition, and maximally reducing disturbance to system regulated output functionyTo realize closed-loop stable control of frequency, the corresponding closed-loop system is:
In the formula (I), the compound is shown in the specification,y(k) Is composed ofkThe output function of the time of day system,Cin order to output the matrix for the state,D ω in order to perturb the output matrix,Kin order to obtain the gain of the gain,x(k) Is composed ofkThe state variable input by the system at the moment,Din order to control the output matrix,i(k) Is composed ofkDisturbance variables of the time system;
robust gain matrixKThe design method is as follows:
the robust performance constraint inequality of the system is:
wherein, the first and the second end of the pipe are connected with each other,Ais a matrix of parameters, and is,WandXis a positive definite matrix to be solved, T represents the matrix transposition,Gin the form of a matrix of parameters,a parameter matrix is designed for the controller,γ 1 for the controller transfer function minimum rejection ratio,Iis a matrix of the units,Zthe output matrix is evaluated for the controller,C 2 designing a parameter matrix for the controller;
positive definite matrix if optimal solution existsWAndXthe system is progressively stabilized and the gain parameter is controlled to beK=X(W) -1 I.e. byu(t)=X(W) -1 xFor co-frequency modulationA controller; the linear matrix inequality is used for solving the control parameters, so that the solving complexity is reduced, and the optimal control parameters are ensured.
training a multivariate regression model for system inertia estimation using available inertial data, load curves, features extracted in ambient frequency measurements at different locations, and weather data; for applications with large amounts of training data, system inertia is estimated using multivariate random forest regression, MRFR, as a machine learning model;
MRFR is a set of regression trees trained by guided sampling and random feature selection, and it mainly comprises the following steps:
1) From a training sample setSIn the random extractionmObtaining a new sample pointS 1 -S n A training subset;
2) Training a CART regression tree by using a training subset, wherein in the training process, the segmentation rule of each node is that all features are randomly selected firstlykIs then characterized bykSelecting the optimal cutting point from the characteristics to divide left and right subtrees;
3) Obtaining a plurality of CART regression tree models through the step 2), wherein the final prediction result of each CART regression tree is the mean value from the sample point to the leaf node;
4) The prediction result of the multivariate random forest is the average value of the prediction results of all CART regression trees;
the use of MRFR includes both offline training, where the MRFR is trained using available offline data, and online applications, where the trained MRFR will receive online measurements and extracted features and use them to estimate the total inertia of the power system.
after the real-time inertia is obtained through estimation, a key problem is that the virtual inertia required by an area is accurately calculated under the condition of giving the real-time inertia and the frequency response target of an interconnected power system, and because the relation between the real-time working condition of the system and the lowest point of the frequency response is nonlinear, the relation among the total inertia, the area virtual inertia and the lowest point of the system frequency of the interconnected system is accurately modeled by using a machine learning method such as a deep neural network method;
input data of the neural network are the minimum point targets of the total inertia of the interconnected system and the system frequency, and are converted into a two-dimensional mapping and pooling layer in a convolution layer, and finally a complete connection layer comprises an output control signal and a regional virtual inertia regulation value. The convolutional layer comprises a plurality of convolution kernels, each element corresponds to a weight coefficient and a deviation value, the size of each element is determined by the convolution kernels, and the convolutional layer can be used for extracting the characteristics of normalized input data and is described as follows:
in the formula (I), the compound is shown in the specification,O n j, the output of the convolution layer is shown,fin order to map the layers of the image,U i n,-1 represents a parameterized spatial filter, the specific actions of which are automatically learned from the data during the training of the network,M n ij, which represents the input of the convolutional layer,b n j, in the case of a real number bias term,represents a cross product;
the function of the pooling layer is to select features and filter layers of information from the convolved output data, which is expressed as follows:
in the formula (I), the compound is shown in the specification,Q n+1,j the water is output from the pond-forming layer,gin order to pool the mapping, the mapping is performed,k n+1,j andr n+1,j respectively representing the multiplicative deviation and the additive deviation,representing a pooling function;
the filter layer can thus be expressed as:
in the formula (I), the compound is shown in the specification,p、qrespectively representing the length and width of the filter layer; the pooling operation is to reduce the output of the upper layers, which makes the transformation of the predicted input data more robust,J i,j is the output of the filtering layer, and is the output of the filtering layer,S p,q in order to filter the parameters of the process,b i+p ,j+q--11 is a real number bias term;
finally, the complete connection layer can expand the feature map into vectors and output values of each complete connection layer; passing to the output requires defining a penalty function to accomplish this classification; the fully connected layer performs non-linear combination, and this process is described as follows:
in the formula (I), the compound is shown in the specification,Y n representing the output of the fully connected layer(s),fa non-linear function representing a fully connected layer,w n andp n respectively representing the weight and the deviation of the complete connection layer;
for a given system, simulation data or historical data can be used for training a machine learning model, and the trained machine learning model is an adaptive model and realizes accurate mapping of input and output; and estimating a virtual inertia set value of the target with the lowest point of the given frequency by using information such as a pre-training model and system inertia.
Claims (6)
1. A wind, light, water, fire and storage combined secondary frequency modulation method based on real-time inertia estimation is characterized by comprising the following steps: a secondary frequency modulation state space model of the system is established, a virtual inertia control link and a closed-loop robust frequency controller are designed, virtual inertia parameters are adjusted in a self-adaptive mode through real-time inertia estimation, and dynamic performance of system frequency under large disturbance can be effectively enhanced aiming at actual scenes of system operation condition change and inertia lack; the method comprises the following steps:
step 1, establishing a frequency modulation model of each device of the wind, light, water and fire storage; establishing a tie line and a frequency response model thereof;
step 2, establishing a regional frequency modulation state space equation model containing wind, light, water and fire storage;
step 3, designing a virtual inertia control link, and designing a closed-loop controller of the regional wind, light, water and fire storage secondary robust frequency controller meeting the given performance index on the basis of a virtual inertia control model;
step 4, providing an interconnected power system real-time inertia estimation method based on multivariate random forest regression, and estimating the real-time inertia of the power system by utilizing PMU data and environmental information;
step 5, establishing an input-output relation between the virtual inertia parameters and corresponding targets of real-time inertia and given frequency of the interconnected power system based on a deep neural network method;
and step 6, changing the operation condition and inertia of the interconnected power system, and changing the virtual inertia parameters in a self-adaptive manner to meet the frequency modulation performance of the system, so that the out-of-limit risk of the system frequency under a large disturbance accident is reduced.
2. The wind, light, water, fire and storage combined secondary frequency modulation method based on real-time inertia estimation as claimed in claim 1, wherein: in step 1, the frequency modulation model of each device is specifically:
a hydroelectric thermal power unit model comprising:
a speed regulator model:
in the formula (I), the compound is shown in the specification,iis a region number, ΔL gi Is a regioniThe position variation of the speed regulator regulating valve of the thermal power generating unit,sin order to be a differential operator, the method comprises the following steps of,T gi is a regioniThe time constant of the speed regulator of the thermal power generating unit,γ kGi is a regionDomainiPower signal distribution coefficient, delta, of thermal power generating unitsP ci Is a regioniThe control signal of the system is sent to the computer,K SAGi is a regioniSag factor, Δ, of thermal power generating unitsf i Is a regioniFrequency of (a)u Gi A control signal of the thermal power generating unit is obtained;
the turbine model is as follows:
in the formula,. DELTA.P mi Is a regioniThe output power variation of the thermal power generating unit,Y chi is a regioniThe time constant of the thermal power unit turbine,sis a differential operator;
for the wind power station model, a variable-speed wind turbine generator is adopted to participate in system frequency regulation, and the simplified model is expressed as follows:
in the formula (I), the compound is shown in the specification,indicating areaiThe variation of the rotating speed of the rotor of the fan,J Wti is a regioniThe comprehensive inertia coefficient of the fan is obtained,N gi is a regioniThe ratio of the fan to the gear box,indicating areaiVariation of pitch angle, Δ, of fanu Wi Indicating areaiA control signal of the wind power station is sent,α kWi is a regioniPower signal distribution coefficient, delta, of a wind farmP ci Is a regioniSystem control signal, Δv Wmi Is a regioniThe wind speed variation of (2);
wherein the content of the first and second substances,is differentiated byIs represented as follows:
in the formula (I), the compound is shown in the specification,andK piI respectively representing regionsiProportional and integral coefficients of the fan PI controller,K ci indicating areaiThe correction coefficient of the fan is set according to the specific formula,T gi indicating areaiThe mechanical torque of the fan is changed into the torque,indicating areaiThe variation of the rotating speed of the fan generator;
the photovoltaic power station active power reduction is adopted to participate in system frequency adjustment, the photovoltaic power station is equivalent to a first-order inertia link, and a dynamic response model is as follows:
in the formula,. DELTA.P iPV Is a regioniThe photovoltaic power station outputs the variation of the active power,Y iPV is a regioniResponse time constant, Δ, of photovoltaic plantu iPV Indicating areaiA control signal of the photovoltaic station is sent,γ iPV is a regioniThe power signal distribution coefficient of the photovoltaic station;
for the energy storage power station, the transfer function in the energy storage power station is equivalent to a first-order inertia link as follows:
in the formula,. DELTA.P iBESS Is a regioniThe energy storage power station outputs the variation of active power,Y iBESS indicating areaiResponse time constant, Δ, of energy storage power stationc Bi Is a regioniA control signal of the energy storage power station,γ Bi is a regioniThe power signal distribution coefficient of the energy storage power station;
state of charge of energy storage batterySOCThe operating state and the regulation and control capability of the energy storage unit are estimated by adopting an ampere-hour integration methodSOCThe calculation formula is as follows:
in the formula (I), the compound is shown in the specification,f(SOC i (t)) representstTime zoneiState of charge of energy storage power stationSOC,f(SOC i0 ) Is a regioniInitial state of charge of energy storage power stationSOC,For the power loss factor of the energy storage power station,P iBESS is a regioniThe energy storage power station outputs active power,S icap, is a regioniRated capacity of the energy storage power station;
after the frequency modulation model of each device is established, a tie line model is established as follows:
in the formula,. DELTA.P tie,i Is an implantation regioniTotal tie line power, ΔP tie,ij Is a regioniAndjthe power of the interconnect link of (a),T ij is a regioniAnd areajInterconnection gain, Δf i And Δf j Are respectively regionsiAndjthe frequency of (a) of (b) is,sin order to be a differential operator, the system is,Nis the number of regions, ΔACE i Is a regioniThe control error of (2) is set,β i is a regioniFrequency deviation factor of, ΔP ci Is a regioniA system control signal;
establishing a frequency response model of the regional power grid as follows:
in the formula,. DELTA.f i Is a regioniThe frequency of (a) of (b) is,M INEi is a regioniThe coefficient of inertia is determined by the measured value of the mass,D i is a regioniDamping coefficient, ΔP mi Is a regioniVariation of output power, delta, of thermal power generating unitsP wi Is a regioniVariation of output power, delta, of wind turbineP iPV Is a regioniVariation of output power, delta, of a photovoltaic power stationP iBESS Is a regioniVariation of output active power, delta, of energy storage power stationP di Is a regioniAmount of change in load, ΔP tie,i Is an implantation regioniTotal tie line power.
3. The wind, light, water, fire and storage combined secondary frequency modulation method based on real-time inertia estimation as claimed in claim 1, wherein:
in step 2, a regional frequency modulation state space equation model containing wind, light, water and fire storage is established as follows:
in the formula (I), the compound is shown in the specification,x(t) Is composed oftTime of day system global stateThe vector of the vector is then calculated,A sys is a matrix of the whole system, and the system,Bis an integral control matrix and is provided with a plurality of control matrixes,u(t) Is composed oftThe time system is used for controlling the vector as a whole,B ω in the form of an overall perturbation matrix,ω(t) Is composed oftThe integral disturbance vector of the time system;
the overall vector contains the following specific quantities:
in the formula (I), the compound is shown in the specification,the state vector of the variation of the rotating speed of the fan rotor is shown,a state vector representing the variation of the fan generator speed,state vector, Δ, representing variation of pitch angle of fanP m Output power variation state vector, delta, for thermal power generating unitsP g Adjusting a state vector, delta, of a change in valve position for a thermal power plant governorP BESS Outputting active power variation state vector, delta, for energy storage power stationSOCRepresenting the state of charge vector, Δ, of an energy storage power stationfFor each of the region frequency state vectors,integrating state vector, Δ, for regional control errorsP tie Is the total tie-line power state vector, Δ, of the implanted regionu W Representing control vectors, Δ, of wind power plant control signalsu G Control vector, delta, for thermal power generating unitsu B Controlling the vector, Δ, for the control signal of the energy-storing power stationP d Perturbation vector for load variation,Δv m For wind speed disturbance vector, superscriptTRepresenting a transpose;
the characteristic equation of the overall system is as follows:
in the formula (I), the compound is shown in the specification,for the characteristic equation function representation, det represents the determinant,A sys is a matrix of the whole system and is,Ithe matrix of the unit is expressed by,are the frequency domain coefficients of the characteristic polynomial,nis the number of characteristic polynomials.
4. The wind, light, water, fire and storage combined secondary frequency modulation method based on real-time inertia estimation as claimed in claim 1, wherein:
and 3, expressing a model of the virtual inertia control link by the following equation:
in the formula,. DELTA.P ref (s) Being reference to inverting unitsThe transformation function represents the energy shortage of the system obtained by current calculation, the output of the transformation function is transmitted to the coordinated frequency modulation units such as wind, light, water, fire storage and the like in a signal form,H vi andD vi is a virtual rotor parameter, Δf(s) Is the system frequency variation deltafThe laplace transform function of;
the overall transfer function is expressed as:
in the formula (I), the compound is shown in the specification,R i andK int respectively, a virtual droop constant and a virtual integral gain;
based on the virtual inertial transfer function, the frequency outer loop is designed as follows:
in the formula (I), the compound is shown in the specification,for the description of the frequency variation of the system,is the nominal frequency of the system and is,J p andD p virtual inertia and damping coefficient, respectively, of the systemP R For wind, light, water and fire storage of active power regulating quantity, deltaP unc For uncontrolled disturbances inside and outside the system, ΔPThe system active power output quantity is obtained;
on the basis, a state feedback robust controller is further designedM=KxTo ensure the system to be disturbed by the outsideiStability under the condition and maximally reducing disturbance to system regulated output functionyThe closed loop stability control of the frequency is realized, and the corresponding closed loop system is as follows:
in the formula (I), the compound is shown in the specification,y(k) Is composed ofkThe output function of the time of day system,Cin order to output the matrix for the state,D ω in order to perturb the output matrix,Kin order to obtain the gain of the gain,x(k) Is composed ofkThe state variable input by the system at the moment,Din order to control the output matrix,i(k) Is composed ofkDisturbance variables of the time system;
robust gain matrixKThe design method is as follows:
the robust performance constraint inequality of the system is:
wherein, the first and the second end of the pipe are connected with each other,Ain the form of a matrix of parameters,WandXis a positive definite matrix to be solved, T represents the matrix transposition,Gis a matrix of parameters, and is,a parameter matrix is designed for the controller,γ 1 for the controller transfer function minimum rejection ratio,Iis a matrix of the units,Zin order for the controller to evaluate the output matrix,C 2 designing a parameter matrix for the controller;
positive definite matrix if optimal solution existsWAndXthe system is gradually stabilized and the gain parameter is controlled to beK=X(W) -1 I.e. byu(t)=X(W) -1 xFor co-frequency modulationA controller; the linear matrix inequality is used for solving the control parameters, so that the solving complexity is reduced, and the optimal control parameters are ensured.
5. The wind, light, water, fire and storage combined secondary frequency modulation method based on real-time inertia estimation as claimed in claim 1, wherein:
step 4, providing an interconnected power system real-time inertia estimation method based on multivariate random forest regression, and estimating the real-time inertia of a power system by utilizing PMU data and environmental information;
training a multivariate regression model for system inertia estimation using available inertial data, load curves, features extracted in environmental frequency measurements at different locations, and weather data; for applications with large amounts of training data, system inertia is estimated using multivariate random forest regression, MRFR, as a machine learning model;
MRFR is a set of regression trees trained by guided sampling and random feature selection, and it mainly comprises the following steps:
1) From a training sample setSIn the random drawingmA sample point is obtained to obtain a new oneS 1 -S n A training subset;
2) Training a CART regression tree by using training subsets, wherein in the training process, the segmentation rule of each node is that all the characteristics are randomly selectedkA feature is then derived fromkSelecting the optimal cutting point from the characteristics to divide left and right subtrees;
3) Obtaining a plurality of CART regression tree models through the step 2), wherein the final prediction result of each CART regression tree is the mean value from the sample point to the leaf node;
4) The prediction result of the multivariate random forest is the average value of the prediction results of all CART regression trees;
the use of MRFR includes both offline training, where the MRFR is trained using available offline data, and online applications, where the trained MRFR will receive online measurements and extracted features and use them to estimate the total inertia of the power system.
6. The wind, light, water, fire and storage combined secondary frequency modulation method based on real-time inertia estimation as claimed in claim 1, wherein:
step 5, establishing an input-output relation between a gain parameter of the controller and a corresponding target of real-time inertia and a given frequency of the interconnected power system based on a deep neural network method;
after the real-time inertia is obtained through estimation, a key problem is that the virtual inertia required by an area is accurately calculated under the condition of giving the real-time inertia and the frequency response target of an interconnected power system, and because the relation between the real-time working condition of the system and the lowest point of the frequency response is nonlinear, the relation among the total inertia, the area virtual inertia and the lowest point of the system frequency of the interconnected system is accurately modeled by using a machine learning method such as a deep neural network method;
input data of the neural network are the total inertia of the interconnected system and the lowest point target of the system frequency, and are converted into a two-dimensional mapping and pooling layer in the convolutional layer, and finally, a complete connection layer comprises an output control signal and a regional virtual inertia regulation value; the convolution layer comprises a plurality of convolution kernels, each element corresponds to a weight coefficient and a deviation value, the size of each element is determined by the convolution kernels, and the convolution layer can be used for extracting the characteristics of normalized input data and is described as follows:
in the formula (I), the compound is shown in the specification,O n j, the output of the convolutional layer is shown,fin order to map the layers of the image,U i n,-1 represents a parameterized spatial filter, the specific actions of which are automatically learned from the data during the training of the network,M n ij, represents the input of the convolutional layer(s),b n j, in order to make the real number bias term,represents a cross product;
the function of the pooling layer is to select features and filter layers of information from the convolved output data, which is expressed as follows:
in the formula (I), the compound is shown in the specification,Q n+1,j the water is output from the pond-forming layer,gin order to pool the mapping, the mapping is performed,k n+1,j andr n+1,j respectively representing the multiplicative deviation and the additive deviation,representing a pooling function;
the filter layer can thus be expressed as:
in the formula (I), the compound is shown in the specification,p、qrespectively representing the length and width of the filter layer; the pooling operation is to reduce the output of the upper layers, which makes the transformation of the predicted input data more robust,J i,j is the output of the filtering layer, and is the output of the filtering layer,S p,q in order to filter the parameters of the process,b i+p ,j+q--11 is a real bias term;
finally, the complete connection layer can expand the feature map into vectors and output values of each complete connection layer; passing to the output requires defining a penalty function to accomplish this classification; the fully connected layer performs a non-linear combination, and this process is described as follows:
in the formula (I), the compound is shown in the specification,Y n representing the output of the fully connected layer(s),fa non-linear function representing a fully connected layer,w n andp n respectively representing the weight and deviation of the fully connected layer;
for a given system, a machine learning model can be trained by using simulation data or historical data, and the trained machine learning model is an adaptive model and realizes accurate mapping of input and output; and estimating a virtual inertia set value of the target with the lowest point of the given frequency by using information such as a pre-training model and system inertia.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211226765.8A CN115296309B (en) | 2022-10-09 | 2022-10-09 | Wind, light, water, fire and storage combined secondary frequency modulation method based on real-time inertia estimation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211226765.8A CN115296309B (en) | 2022-10-09 | 2022-10-09 | Wind, light, water, fire and storage combined secondary frequency modulation method based on real-time inertia estimation |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115296309A true CN115296309A (en) | 2022-11-04 |
CN115296309B CN115296309B (en) | 2023-02-14 |
Family
ID=83834199
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211226765.8A Active CN115296309B (en) | 2022-10-09 | 2022-10-09 | Wind, light, water, fire and storage combined secondary frequency modulation method based on real-time inertia estimation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115296309B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116667389A (en) * | 2023-07-25 | 2023-08-29 | 昆明理工大学 | Novel wind-storage combined prediction correction control method for inertia lifting of power system |
Citations (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2019205626A1 (en) * | 2018-04-23 | 2019-10-31 | 华北电力科学研究院有限责任公司 | Coordinated frequency modulation device for wind power storage |
CN110879917A (en) * | 2019-11-08 | 2020-03-13 | 北京交通大学 | Electric power system transient stability self-adaptive evaluation method based on transfer learning |
CN110932629A (en) * | 2019-12-20 | 2020-03-27 | 华中科技大学 | Inertia estimation method based on single neuron network and self-adaptive adjustment strategy thereof |
CN110970933A (en) * | 2019-11-23 | 2020-04-07 | 国网辽宁省电力有限公司电力科学研究院 | Virtual inertia compensation method of light-storage combined power generation system based on active support control |
CN111080597A (en) * | 2019-12-12 | 2020-04-28 | 西南交通大学 | Track fastener defect identification algorithm based on deep learning |
US20200144946A1 (en) * | 2018-07-03 | 2020-05-07 | Shanghai Jiaotong University | Adaptive Inertia Control Method of IIDG Based on VSG |
CN111740432A (en) * | 2020-06-23 | 2020-10-02 | 上海电力大学 | New energy power system load frequency control method based on hybrid energy storage adjustment |
CN111864813A (en) * | 2020-06-23 | 2020-10-30 | 国网辽宁省电力有限公司电力科学研究院 | Wind/thermal power combined frequency control method based on virtual weight coefficient |
CN111900742A (en) * | 2020-07-02 | 2020-11-06 | 武汉大学 | Frequency modulation method of wind storage system based on double-layer cooperative control |
CN112152233A (en) * | 2020-09-18 | 2020-12-29 | 国电南瑞科技股份有限公司 | Energy storage power station frequency modulation control method and device considering charge state and communication delay |
CN112636398A (en) * | 2020-12-23 | 2021-04-09 | 东南大学 | Wind-fire-storage combined secondary frequency modulation method |
CN112701706A (en) * | 2020-12-01 | 2021-04-23 | 国网辽宁省电力有限公司阜新供电公司 | Method and system for analyzing secondary frequency modulation characteristics of battery energy storage power station participating in power grid |
WO2021164112A1 (en) * | 2020-02-18 | 2021-08-26 | 山东大学 | Frequency control method and system during using wind farm as black-start power source by means of optimal configuration of energy storage |
CN113904386A (en) * | 2021-10-25 | 2022-01-07 | 国网宁夏电力有限公司 | Photovoltaic frequency modulation control parameter optimization method considering equivalent inertia and damping requirements |
CN114069711A (en) * | 2021-11-25 | 2022-02-18 | 江苏科技大学 | Virtual inertia control system for offshore wind power |
CN114784854A (en) * | 2022-04-07 | 2022-07-22 | 中国电力科学研究院有限公司 | Wind storage combined power station participating frequency modulation optimization control method, system, equipment and medium |
CN115021336A (en) * | 2022-06-28 | 2022-09-06 | 河海大学 | Optimization method for high-proportion new energy power system inertia deficiency scene |
CN115149580A (en) * | 2022-08-29 | 2022-10-04 | 国网江西省电力有限公司电力科学研究院 | Wind, light, water, fire and storage combined secondary frequency modulation method considering uncertainty delay |
-
2022
- 2022-10-09 CN CN202211226765.8A patent/CN115296309B/en active Active
Patent Citations (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2019205626A1 (en) * | 2018-04-23 | 2019-10-31 | 华北电力科学研究院有限责任公司 | Coordinated frequency modulation device for wind power storage |
US20200144946A1 (en) * | 2018-07-03 | 2020-05-07 | Shanghai Jiaotong University | Adaptive Inertia Control Method of IIDG Based on VSG |
CN110879917A (en) * | 2019-11-08 | 2020-03-13 | 北京交通大学 | Electric power system transient stability self-adaptive evaluation method based on transfer learning |
CN110970933A (en) * | 2019-11-23 | 2020-04-07 | 国网辽宁省电力有限公司电力科学研究院 | Virtual inertia compensation method of light-storage combined power generation system based on active support control |
CN111080597A (en) * | 2019-12-12 | 2020-04-28 | 西南交通大学 | Track fastener defect identification algorithm based on deep learning |
CN110932629A (en) * | 2019-12-20 | 2020-03-27 | 华中科技大学 | Inertia estimation method based on single neuron network and self-adaptive adjustment strategy thereof |
WO2021164112A1 (en) * | 2020-02-18 | 2021-08-26 | 山东大学 | Frequency control method and system during using wind farm as black-start power source by means of optimal configuration of energy storage |
CN111864813A (en) * | 2020-06-23 | 2020-10-30 | 国网辽宁省电力有限公司电力科学研究院 | Wind/thermal power combined frequency control method based on virtual weight coefficient |
CN111740432A (en) * | 2020-06-23 | 2020-10-02 | 上海电力大学 | New energy power system load frequency control method based on hybrid energy storage adjustment |
CN111900742A (en) * | 2020-07-02 | 2020-11-06 | 武汉大学 | Frequency modulation method of wind storage system based on double-layer cooperative control |
CN112152233A (en) * | 2020-09-18 | 2020-12-29 | 国电南瑞科技股份有限公司 | Energy storage power station frequency modulation control method and device considering charge state and communication delay |
CN112701706A (en) * | 2020-12-01 | 2021-04-23 | 国网辽宁省电力有限公司阜新供电公司 | Method and system for analyzing secondary frequency modulation characteristics of battery energy storage power station participating in power grid |
CN112636398A (en) * | 2020-12-23 | 2021-04-09 | 东南大学 | Wind-fire-storage combined secondary frequency modulation method |
CN113904386A (en) * | 2021-10-25 | 2022-01-07 | 国网宁夏电力有限公司 | Photovoltaic frequency modulation control parameter optimization method considering equivalent inertia and damping requirements |
CN114069711A (en) * | 2021-11-25 | 2022-02-18 | 江苏科技大学 | Virtual inertia control system for offshore wind power |
CN114784854A (en) * | 2022-04-07 | 2022-07-22 | 中国电力科学研究院有限公司 | Wind storage combined power station participating frequency modulation optimization control method, system, equipment and medium |
CN115021336A (en) * | 2022-06-28 | 2022-09-06 | 河海大学 | Optimization method for high-proportion new energy power system inertia deficiency scene |
CN115149580A (en) * | 2022-08-29 | 2022-10-04 | 国网江西省电力有限公司电力科学研究院 | Wind, light, water, fire and storage combined secondary frequency modulation method considering uncertainty delay |
Non-Patent Citations (3)
Title |
---|
POUYA BABAHAJIANI等: "Intelligent Demand Response Contribution in Frequency Control of Multi-Area Power Systems", 《IEEE TRANSACTIONS ON SMART GRID》 * |
张从越等: "分布式光伏集群鲁棒虚拟同步化控制方法", 《中国电机工程学报》 * |
张程铭等: "考虑频率安全的电力系统规划与运行优化研究综述与展望", 《电网技术》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116667389A (en) * | 2023-07-25 | 2023-08-29 | 昆明理工大学 | Novel wind-storage combined prediction correction control method for inertia lifting of power system |
CN116667389B (en) * | 2023-07-25 | 2023-10-13 | 昆明理工大学 | Novel wind-storage combined prediction correction control method for inertia lifting of power system |
Also Published As
Publication number | Publication date |
---|---|
CN115296309B (en) | 2023-02-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Khooban et al. | A novel deep reinforcement learning controller based type-II fuzzy system: Frequency regulation in microgrids | |
CN115296308B (en) | Robust cooperative frequency modulation method considering energy storage charge state and adaptive inertia level | |
Pandey et al. | A literature survey on load–frequency control for conventional and distribution generation power systems | |
Shakibjoo et al. | Load frequency control for multi-area power systems: A new type-2 fuzzy approach based on Levenberg–Marquardt algorithm | |
Aravindh et al. | Design of observer-based non-fragile load frequency control for power systems with electric vehicles | |
Golshannavaz et al. | A two-stage robust-intelligent controller design for efficient LFC based on Kharitonov theorem and fuzzy logic | |
CN113054688B (en) | Renewable energy cluster output distributed coordination control method considering uncertainty | |
CN115296309B (en) | Wind, light, water, fire and storage combined secondary frequency modulation method based on real-time inertia estimation | |
CN114597969B (en) | Power distribution network double-layer optimization method considering intelligent soft switch and virtual power plant technology | |
Jiao et al. | Decentralized voltage control of wind farm based on gradient projection method | |
Elnaggar et al. | L1 adaptive Fuzzy control of wind energy conversion systems via variable structure adaptation for all wind speed regions | |
CN115149580A (en) | Wind, light, water, fire and storage combined secondary frequency modulation method considering uncertainty delay | |
Lasheen et al. | Adaptive control design and implementation for collective pitch in wind energy conversion systems | |
Malmir et al. | Controlling megawatt class WECS by ANFIS network trained with modified genetic algorithm | |
CN115795992A (en) | Park energy Internet online scheduling method based on virtual deduction of operation situation | |
CN115689375A (en) | Virtual power plant operation control method, device, equipment and medium | |
Kazemi et al. | Adaptive fractional‐order control of power system frequency in the presence of wind turbine | |
Lakshmi et al. | Optimal robust PID-PSS design for melioration of power system stability using search and rescue algorithm | |
CN113394770A (en) | Interconnected micro-grid group frequency complete distribution type optimization control method and system | |
CN116154791B (en) | Power angle stable control method, system and terminal for cooperative multiple controllable resources | |
Kayalvizhi et al. | Frequency control of microgrid with wind perturbations using levy walks with spider monkey optimization algorithm | |
CN108988399B (en) | Energy storage fast frequency modulation method based on active imbalance distance | |
CN115622131A (en) | Micro-grid frequency robust optimal H with energy storage 2 /H ∞ Controller design method | |
CN115102228A (en) | Multi-target coordination frequency optimization method and device for wind power plant containing flywheel energy storage | |
Dapoz et al. | Mixed regression clustering techniques for the medium term prediction of reactive power within transmission grids |
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 |