CN117318021A - Electric power system dynamic state estimation method and system based on long-term and short-term memory network - Google Patents
Electric power system dynamic state estimation method and system based on long-term and short-term memory network Download PDFInfo
- Publication number
- CN117318021A CN117318021A CN202311165743.XA CN202311165743A CN117318021A CN 117318021 A CN117318021 A CN 117318021A CN 202311165743 A CN202311165743 A CN 202311165743A CN 117318021 A CN117318021 A CN 117318021A
- Authority
- CN
- China
- Prior art keywords
- phase angle
- time
- voltage amplitude
- voltage phase
- power system
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 108
- 230000007787 long-term memory Effects 0.000 title claims abstract description 27
- 230000006403 short-term memory Effects 0.000 title abstract description 14
- 239000011159 matrix material Substances 0.000 claims abstract description 70
- 238000001914 filtration Methods 0.000 claims abstract description 36
- 238000012937 correction Methods 0.000 claims abstract description 24
- 230000015654 memory Effects 0.000 claims abstract description 24
- 238000005259 measurement Methods 0.000 claims description 28
- 230000008569 process Effects 0.000 claims description 25
- 238000012935 Averaging Methods 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 5
- 238000000354 decomposition reaction Methods 0.000 claims description 5
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000008859 change Effects 0.000 description 6
- 238000012549 training Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 238000012360 testing method Methods 0.000 description 5
- 230000005540 biological transmission Effects 0.000 description 4
- 230000001902 propagating effect Effects 0.000 description 3
- 239000000243 solution Substances 0.000 description 3
- 230000007704 transition Effects 0.000 description 3
- 238000003491 array Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000000835 fiber Substances 0.000 description 2
- 238000009499 grossing Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- RYGMFSIKBFXOCR-UHFFFAOYSA-N Copper Chemical compound [Cu] RYGMFSIKBFXOCR-UHFFFAOYSA-N 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 210000004027 cell Anatomy 0.000 description 1
- 229910052802 copper Inorganic materials 0.000 description 1
- 239000010949 copper Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000013213 extrapolation Methods 0.000 description 1
- 238000002347 injection Methods 0.000 description 1
- 239000007924 injection Substances 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012067 mathematical method Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
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
- H02J3/004—Generation forecast, e.g. methods or systems for forecasting future energy generation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
-
- 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]
Landscapes
- Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Software Systems (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Power Engineering (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
A method and a system for estimating dynamic state of an electric power system based on a long-term and short-term memory network comprise the following steps: the method comprises the steps of measuring the active power, reactive power, voltage amplitude and voltage phase angle of each node to form a quantity, and forming a state quantity by the voltage amplitude and the voltage phase angle; respectively constructing a voltage amplitude prediction model and a voltage phase angle prediction model based on the long-short term memory network; determining volume points based on a third-order-spherical radial volume rule, and taking the volume points as input values of a voltage amplitude prediction model and a voltage phase angle prediction model to obtain a voltage amplitude prediction value and a voltage phase angle prediction value; and carrying out self-adaptive volume Kalman filtering on the voltage amplitude predicted value and the voltage phase angle predicted value, wherein a time-varying noise correction factor is utilized to carry out time-varying correction on a system noise covariance matrix, and average values are respectively obtained on the filtered voltage amplitude predicted value and the filtered voltage phase angle predicted value to obtain a dynamic state estimation result of the power system. The invention accurately estimates the running state of the power system.
Description
Technical Field
The invention belongs to the field of dynamic state estimation of power systems, and particularly relates to a method and a system for estimating dynamic state of a power system based on a long-term and short-term memory network.
Background
In the context of new power systems, the scale and complexity of the power grid is continually increasing. Because of measurement errors in the directly measured power system operation data, the directly measured data cannot provide effective data support for power system operation analysis. The dynamic state estimation technology of the power system can filter measurement errors in measurement data and estimate state variables of the power system. The dynamic state estimation algorithm of the power system needs to establish a state equation to predict the state quantity. Because of the high nonlinearity degree of the power system, a state equation with clear physical meaning is difficult to establish, an accurate state transition matrix cannot be obtained, the effect of a prediction step is greatly reduced, and the filtering precision of a dynamic state estimation method is finally influenced. The filtering precision is an important index for measuring the goodness of the dynamic estimation method of the power system.
In the prior art, as the change of the state quantity is the result of the change of the node injection power, the prediction step in most power system dynamic state estimation methods adopts a method of directly carrying out linear extrapolation on the state variable by adopting a two-parameter exponential smoothing method, and the method establishes a state transition matrix through a mathematical equation, so that the state quantity is predicted by establishing the state equation, the calculated quantity is small, the prediction result can be obtained in a shorter time when the dimension of the operation data of the power system is not high, and the prediction precision can be higher when the dimension is smaller. Under the background of a novel power system, the scale and complexity of a power grid are continuously improved after novel factors such as new energy sources, electric vehicles and the like are accessed, and the power system has the characteristics of high dimensionality and strong nonlinearity. The traditional two-parameter exponential smoothing prediction method is not based on an actual power grid model, and can not ensure higher prediction precision in a high-dimensional power system only according to the process of approximating state transition by a mathematical equation, so that the filtering precision of dynamic state estimation is affected.
Disclosure of Invention
In order to solve the defects in the prior art, the invention provides a method and a system for estimating the dynamic state of a power system based on a long-short-period memory network.
The invention adopts the following technical scheme.
The invention provides a power system dynamic state estimation method based on a long-term and short-term memory network, which comprises the following steps:
the method comprises the steps of measuring the active power, reactive power, voltage amplitude and voltage phase angle of each node to form a quantity, and forming a state quantity by the voltage amplitude and the voltage phase angle;
respectively constructing a voltage amplitude prediction model and a voltage phase angle prediction model based on the long-short term memory network; determining a volume point based on a third-order-spherical radial volume rule, and taking the volume point as an input value of a voltage amplitude prediction model and a voltage phase angle prediction model to obtain a voltage amplitude prediction value and a voltage phase angle prediction value;
performing volume Kalman filtering on the voltage amplitude predicted value and the voltage phase angle predicted value, wherein time-varying noise correction factors are utilized to perform time-varying correction on a system noise covariance matrix;
and respectively averaging the filtered voltage amplitude predicted value and the voltage phase angle predicted value to obtain a dynamic state estimation result of the power system.
Preferably, the network structure of the voltage amplitude prediction model and the voltage phase angle prediction model are the same.
Preferably, determining the volume point based on the third-order-spherical radial volume rule comprises:
and (3) performing Cholesky decomposition on the error covariance matrix at the moment k-1 to obtain:
wherein S is k-1 Representing an error covariance matrix L k-1 Square root matrix of (a);
the volume point is determined by the following relation:
in the method, in the process of the invention,
representing the state variable corresponding to the i-th volume point at the time of k-1,
ξ i represent S k-1 The corresponding matrix of coefficients is used to determine,
representing the state variable estimate at time k-1,
i is an n-order identity matrix,
2n is the number of volume points.
Inputting the volume point into a voltage amplitude prediction model and a voltage phase angle prediction model to obtain a state quantity prediction value corresponding to the ith volume point at the k moment, wherein the method specifically comprises the following steps of:
and 2n volume points, wherein each volume point corresponds to a state quantity predicted value, and the state quantity predicted value at the k moment is obtained by averaging the state quantity predicted values corresponding to all the volume points.
Preferably, the error covariance matrix predictor is calculated with the following relation:
in the method, in the process of the invention,
2n is the number of volume points,
the state quantity predicted value corresponding to the ith volume point at the k moment is expressed,
the state quantity predicted value at the time k is indicated.
A predicted value representing the k-time error covariance matrix,
M k-1 a system noise covariance matrix at time k-1 is shown.
Preferably, the time-varying noise correction factor satisfies the following relation:
in the method, in the process of the invention,
is a time-varying noise correction factor at time k-1,
beta is denoted as forgetting factor.
Preferably, the time-varying corrected system noise covariance matrix is as follows:
in the method, in the process of the invention,
M k represented as a system noise covariance matrix at time k,
K k denoted as the kalman gain at time k,
L k represented as an error covariance matrix at time k,
ε k denoted as the residual at time k.
Preferably, performing the bulk kalman filtering on the voltage magnitude prediction value and the voltage phase angle prediction value includes: the root mean square error is introduced as the filtering precision, and the calculation formula is as follows:
in the method, in the process of the invention,
is the true value of the state quantity of the jth node at the moment k,
the state quantity predicted value of the j-th node at the k moment,
when the filtering precision is satisfied, the filtering is ended.
Preferably, the measurement error autocovariance matrix C is calculated in the following relation k And a measurement error cross covariance matrix D k ：
In the method, in the process of the invention,
the predicted value is measured for the amount at time k,
representing the amount of measurement corresponding to the ith volume point at time k,
R k to measure the noise covariance matrix.
Preferably, the Kalman gain K is calculated in the following relation k ：
K k ＝C k D k -1
Calculating to obtain a k moment state quantity and an error covariance matrix by using a Kalman gain:
in the method, in the process of the invention,
z k for the measurement of the quantity at time k,
the state quantity at time k.
The invention also provides a power system dynamic state estimation system based on the long-term memory network, which comprises: the system comprises a state quantity acquisition module, a state quantity prediction module, a state quantity filtering module and a dynamic state estimation module;
the state quantity acquisition module is used for measuring the active power, reactive power, voltage amplitude and voltage phase angle constitution quantity of each node, and constructing a state quantity by the voltage amplitude and the voltage phase angle;
the state quantity prediction module is used for respectively constructing a voltage amplitude prediction model and a voltage phase angle prediction model based on the long-short-term memory network; determining a volume point based on a third-order-spherical radial volume rule, and taking the volume point as an input value of a voltage amplitude prediction model and a voltage phase angle prediction model to obtain a voltage amplitude prediction value and a voltage phase angle prediction value;
a state quantity filtering module for performing adaptive volume Kalman filtering on the voltage amplitude predicted value and the voltage phase angle predicted value, wherein the time-varying noise correction factor is utilized to perform time-varying correction on the system noise covariance matrix,
and the dynamic state estimation module is used for respectively averaging the filtered voltage amplitude predicted value and the voltage phase angle predicted value to obtain a dynamic state estimation result of the power system.
The invention also provides a terminal, which comprises a processor and a storage medium; the storage medium is used for storing instructions; the processor is configured to operate in accordance with the instructions to perform the steps of the method.
The invention also proposes a computer-readable storage medium, on which a computer program is stored, which program, when being executed by a processor, implements the steps of the method.
Compared with the prior art, the method has the beneficial effects that the capacity of processing high-dimensional and strong-nonlinearity data of the long-term memory network is fully exerted, the nonlinearity and high-dimensional characteristics of the power system are captured from the historical operation data of the power system, and a more accurate power system dynamic state prediction model, namely a voltage amplitude prediction model and a voltage phase angle prediction model, is established according to the characteristics.
According to the invention, the volume points are determined based on the third-order-spherical radial volume rule, the volume points are respectively used as input values of the voltage amplitude prediction model and the voltage phase angle prediction model, the trained electric power system state prediction model based on the long-short-term memory network is used for predicting the running state of the electric power system, the voltage amplitude prediction value and the voltage phase angle prediction value are obtained, and the accuracy of the prediction step is improved by establishing the prediction model which is more in line with the actual running condition of the electric power system, so that the filtering accuracy is further improved.
The invention also introduces time-varying noise correction factors into the traditional volume Kalman filtering (Cubature Kalman Filter, CKF) method to carry out time-varying correction on the system noise covariance matrix, realizes the function of self-adaptive volume Kalman filtering on the voltage amplitude predicted value and the voltage phase angle predicted value, and improves the accuracy of dynamic state estimation.
Drawings
FIG. 1 is a flow chart of a method for estimating dynamic state of an electric power system based on a long-term and short-term memory network;
FIG. 2 is a graph showing a load factor change of a region according to an embodiment of the present invention;
FIG. 3 is a graph of test set node 4 voltage magnitude fit results in an embodiment of the invention; node 4 is a node in the standard IEEE30 node system;
FIG. 4 is a graph of test set node 4 voltage phase angle fit results in an embodiment of the invention; node 4 is a node in the standard IEEE30 node system;
FIG. 5 is a graph showing the magnitude of the voltage at node 3 versus CKF in an embodiment of the invention; node 3 is a node in the standard IEEE30 node system;
FIG. 6 is a graph comparing the voltage phase angle at node 2 with the CKF method according to an embodiment of the present invention; node 2 is a node in the standard IEEE30 node system;
FIG. 7 is a graph comparing the root mean square error index of the voltage amplitude of IEEE30 node with the index of CKF method in accordance with an embodiment of the present invention;
FIG. 8 is a graph comparing the RMS error index of the IEEE30 node voltage phase angle with the index of the CKF method in accordance with an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. The embodiments described herein are merely some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art without inventive faculty, are within the scope of the invention, based on the spirit of the invention.
The method for estimating the dynamic state of the power system based on the long-term and short-term memory network, as shown in figure 1, comprises the following steps:
step 1, obtaining the number and the type of nodes of a power system and the configuration positions of various measuring devices. The active power, reactive power, voltage amplitude and voltage phase angle of each node are used for forming the quantity measurement, and the voltage amplitude and the voltage phase angle are used for forming the state quantity.
Specifically, the nodes include PQ nodes, PV nodes, and balancing nodes. Types of metrology devices include SCADA and PMU.
In a non-limiting preferred embodiment, the quantity is measured z k ＝[P k ,Q k ,V k ,θ k ] T Including the active power P of each node at time k k Reactive power Q k Amplitude V of voltage k Phase angle θ of voltage k State quantity x k ＝[V k ,θ k ] T The voltage amplitude V of each node at k time is included k And voltage phase angleθ k 。
And 2, respectively constructing a voltage amplitude prediction model and a voltage phase angle prediction model with identical network structures based on the long-term and short-term memory network.
Specifically, a power system state prediction model based on a long-term and short-term memory network is constructed to predict the state quantity of the power system, the input quantity of the prediction model is the node voltage amplitude and phase angle at the moment k-1, and the output quantity is the predicted value of the node voltage amplitude and phase angle at the moment k.
The long-period memory network controls the flow of information through a gating mechanism, thereby realizing the capture of information memory and time sequence dependency relationship during long-sequence training, the input gate controls new information input, the forget gate controls whether old information needs to be forgotten or not, the output gate controls the output of current information, and the internal state determines the change between the current state and the state at the previous moment. The long-term and short-term memory network satisfies the following relation:
in the method, in the process of the invention,
σ represents the Sigmoid function,
phi represents the tanh function and,
f t indicating the state of the left-behind door,
in t indicating the state of the input door,
g t representing the state of the input node,
o t indicating the status of the output door and,
S t representing the state of the state cell,
h t indicating the state of the output unit,
M fx representing input x in a forget gate state t The corresponding weight is used to determine the weight,
M fh indicating the last moment output h in the state of forgetting to gate t-1 The corresponding weight is used to determine the weight,
M inx representing input x in input gate state t Corresponding toThe weight of the material to be weighed,
M inh indicating the output h at the last moment in the state of the input gate t-1 The corresponding weight is used to determine the weight,
M gx representing input x in input node state t The corresponding weight is used to determine the weight,
M gh indicating the output h at the last moment in the state of the input node t-1 The corresponding weight is used to determine the weight,
M ox representing input x in output gate state t The corresponding weight is used to determine the weight,
M oh indicating the output h at the last moment in the output gate state t-1 The corresponding weight is used to determine the weight,
b f representing a deviation term in the forgotten gate state,
b in representing the bias term in the state of the input gate,
b g representing the deviation term in the state of the input node,
b o representing the deviation term in the output gate state.
Therefore, the capacity of the long-term and short-term memory network for processing high-dimensional and strong-nonlinearity data is fully exerted, the nonlinearity and high-dimensional characteristics of the power system are captured from the historical operation data of the power system, a more accurate dynamic state prediction model of the power system is built according to the characteristics, and a voltage amplitude prediction model and a voltage phase angle prediction model which are identical in two network structures are built.
Specifically, a training set is constructed by utilizing historical operation data of the power system, node voltage amplitude data and phase angle data are divided into two training sets, two long-short-period memory networks are constructed, and the two long-period memory networks are respectively used for predicting node voltage amplitude at the time of k-1 and node voltage phase angle at the time of k-1. The two long-term and short-term memory network models are independent of each other and do not affect each other. The two long-short-term memory network models are different in training sets, so that prediction of different types of data is achieved, and a power system state prediction model is constructed by means of single-step prediction of the long-short-term memory network.
Step 3, carrying out dynamic state estimation based on the voltage amplitude prediction model and the voltage phase angle prediction model, wherein the method comprises the following steps: determining a volume point based on a third-order-spherical radial volume rule, and taking the volume point as an input value of a voltage amplitude prediction model and a voltage phase angle prediction model to obtain a voltage amplitude prediction value and a voltage phase angle prediction value; and performing volume Kalman filtering on the voltage amplitude predicted value and the voltage phase angle predicted value, wherein a time-varying noise correction factor is utilized to perform time-varying correction on the system noise covariance matrix. And respectively averaging the filtered voltage amplitude predicted value and the voltage phase angle predicted value to obtain a dynamic state estimation result of the power system.
The single third-order-spherical radial volume rule is to select volume points according to prior mean and covariance through the volume rule, then transfer the volume points through a nonlinear function, and then weight the volume points transferred by the nonlinear function to approximate state posterior mean and covariance. The method comprises three stages of prediction, updating and estimation, wherein in the prediction stage, volume points are calculated, volume point state quantity is predicted, target state quantity is predicted and a prediction covariance matrix is calculated, in the updating stage, volume points are calculated by using the prediction covariance matrix, volume point quantity measurement is predicted, target quantity measurement is predicted, measurement covariance matrix is calculated, state measurement covariance is calculated, kalman gain is calculated by using the state measurement covariance and the volume points, state estimation is performed by using the target state quantity and the target quantity obtained by prediction in the estimation stage, and covariance estimation is performed by using the prediction covariance matrix, kalman gain and the volume points.
The dynamic state estimation method provided by the invention comprises prediction and filtering. The state prediction model of the electric power system constructed by the long-short-term memory network is utilized, the state quantity prediction value is determined by combining a third-order-spherical radial volume rule, and the state estimation value which is closer to a true value is obtained by filtering by a mathematical method by combining the state quantity prediction value, the measurement equation and the like, so that the effect of state estimation is realized.
Specifically, step 3 includes:
and 3.1, setting an initial value of an error covariance matrix, an initial value of a system noise covariance matrix, an initial value of a measurement noise covariance matrix, an initial value of voltage amplitude and phase angle of each node.
Step 3.2, performing Cholesky decomposition on the error covariance matrix at the moment k-1, and calculating volume points;
specifically, the error covariance matrix L at time k-1 is expressed as follows k-1 Cholesky decomposition was performed to give:
wherein S is k-1 Represents L k-1 Is a square root matrix of (a).
The volume point is determined by the following relation:
in the method, in the process of the invention,
representing the state variable corresponding to the i-th volume point at the time of k-1,
ξ i represent S k-1 The corresponding matrix of coefficients is used to determine,
representing the state variable estimate at time k-1,
i is an n-order identity matrix,
2n is the number of volume points.
Step 3.3, the volume pointInputting a voltage amplitude prediction model and a voltage phase angle prediction model to obtain a state quantity pre-set corresponding to the ith volume point at the k momentMeasuring:
in the method, in the process of the invention,
the state quantity predicted value corresponding to the ith volume point at the k moment is expressed,
g () represents a long-short-term memory network.
Specifically, the volume point is input into a voltage amplitude prediction model and a voltage phase angle prediction model to obtain a state quantity prediction value corresponding to the ith volume point at the k moment, and the method specifically comprises the following steps: 2n volume points, each volume point corresponds to a state quantity predicted value, and the state quantity predicted value at the moment k is obtained by averaging all the state quantity predicted values, as follows:
step 3.4, calculating an error covariance matrix prediction value according to the following relation:
in the method, in the process of the invention,
a predicted value representing the k-time error covariance matrix,
M k-1 representing the system noise covariance matrix at time k-1,
the state quantity predicted value at the time k is indicated.
And carrying out time-varying correction on the system noise covariance matrix by using a time-varying noise correction factor, carrying out self-adaptive volume Kalman filtering on the state quantity predicted value at the next moment, and updating the system noise covariance matrix in real time to improve the filtering precision, so that the obtained state estimated value of the power system is closer to the result of a true value.
Specifically, the time-varying noise correction factor is given by the following relation:
the updated system noise covariance matrix is shown as follows:
in the method, in the process of the invention,
is a time-varying noise correction factor at time k-1,
beta is denoted as forgetting factor, and can adapt to the situation of large noise variation when beta is valued larger, and in a non-limiting preferred embodiment beta is valued at 0.95.
M k Represented as a system noise covariance matrix at time k,
K k denoted as the kalman gain at time k,
L k represented as an error covariance matrix at time k,
denoted as the residual of time k,
as a predicted value of the state quantity at time k-1 to time k,
is the i-th volume point generated.
In order to compare the effectiveness of the power system dynamic state estimation method provided by the invention, performing volume Kalman filtering on the voltage amplitude predicted value and the voltage phase angle predicted value comprises the following steps: the root mean square error is introduced as the filtering precision, and the calculation formula is as follows:
in the method, in the process of the invention,
is the true value of the state variable of the jth node at time k,
the state variable predicted value of the j-th node at the k moment.
The smaller the root mean square error value is, the closer the state quantity predicted value is to the state quantity true value, namely the better the filtering precision is, and when the filtering precision requirement is met, the filtering is ended.
Step 3.5, predicting the error covariance matrixCholesky decomposition was performed to give:
in the method, in the process of the invention,representation->Is a square root matrix of (a).
Step 3.6, calculating a new volume point according to the following relation:
step 3.7, propagating the new volume point by the following relation:
in the method, in the process of the invention,
representing the amount of measurement corresponding to the i-th volume point,
h () represents a measurement equation.
The measurement equation is as follows:
in the method, in the process of the invention,
G ij for the mutual conductance between nodes i, j,
B ij for the mutual susceptance between nodes i, j,
θ ij is the voltage phase angle difference between nodes i, j.
Acquiring the number and the type of nodes of the power system and the configuration positions of various measuring devices, and calculating a node admittance matrix G ij 、B ij 。
Step 3.8, calculating the measurement prediction value according to the following relation
Step 3.9, calculating a measurement error auto-covariance matrix C according to the following relation k And a measurement error cross covariance matrix D k ：
In the method, in the process of the invention,
R k to measure the noise covariance matrix.
Step 3.10, calculating the Kalman gain K according to the following relation k ：
K k ＝C k D k -1
Step 3.11, calculating the state quantity at the k momentError covariance matrix L k ：
Step 3.12, setting k to k+1.
Step 3.13 repeating steps 3.2 to 3.12 until k reaches k max Wherein k is max Is the maximum of the number of dynamic state estimation times.
The test case used in the present invention is a standard IEEE30 node system. Parameters of the standard IEEE30 node system can be obtained through MATPOWER toolkit.
In order to simulate the real running condition of the power grid, the load of each node changes according to a load coefficient change curve of a certain area shown in fig. 2. State estimation is performed once every 15min in a dayI.e. 96 state estimations are made a day. The real value of the state variable is calculated according to the tide, the measurement is calculated by the measurement equation and the Gaussian noise is added, namely, the result obtained by calculating the measurement equation is added and distributed as N (0, theta) 2 ) Random noise of (a) is provided.
Two sets of experiments were set up: the dynamic state estimation is performed by the method of the invention, the volume kalman filter (Cubature Kalmen Filter, CKF), respectively. Wherein θ=0.001; the initial value of the error covariance matrix and the system noise covariance matrix is L 0 ＝M 0 ＝θ 2 I n×n I is an n-order identity matrix; historical data of a power system state prediction model based on a long-short-term memory network is obtained through CKF dynamic state estimation based on two-parameter exponential smooth prediction under a typical daily load change curve. Based on 9500 groups of historical data, the first 7600 groups are selected for training, and the second 1900 groups are selected for testing. The related parameters of the long-term and short-term memory network are set as follows: the layer number of the long-period memory network is 2; the number of neurons of each layer of long-term and short-term memory network is 120; the feature dimension is 60; the maximum number of training rounds was 200. Under the standard IEEE30 node system, the fitting result of the test set node 4 is shown in fig. 3 and 4. As can be seen from fig. 3 and 4, the trained model can better predict the state quantity of the power system.
Under the standard IEEE30 node system, the voltage amplitude and CKF of the node 3 under the method of the present invention are shown in fig. 5, and the voltage phase angle and CKF of the node 2 under the method of the present invention are shown in fig. 6. It can be seen from fig. 5 and 6 that the method of the present invention is closer to the true value and the estimation accuracy is better than CKF, regardless of the voltage amplitude or phase angle.
Under the standard IEEE30 node system, the comparison of the root mean square error index of the node voltage amplitude and CKF in the method of the invention is shown in FIG. 7. The comparison of the root mean square error index of the node voltage phase angle and the CKF in the method of the present invention is shown in FIG. 8. As can be seen from fig. 7 and 8, the root mean square error index of the node voltage amplitude and the phase angle is smaller than CKF in the method of the invention, which indicates that the estimation accuracy is better.
Table 1 shows the mean value of the root mean square error at each time in the method of the present invention compared to CKF. As can be seen from table 1, the method of the present invention is used for power system dynamic state estimation, and the mean value of root mean square error is smaller in the method of the present invention than in CKF. Therefore, the method has better estimation precision when the dynamic state of the power system is estimated, and the running state of the power system can be estimated better.
TABLE 1 mean value of root mean square error at various moments in the method of the present invention compared to CKF
The invention also provides a power system dynamic state estimation system based on the long-term memory network, which comprises: the system comprises a state quantity acquisition module, a state quantity prediction module, a state quantity filtering module and a dynamic state estimation module;
the state quantity acquisition module is used for measuring the active power, reactive power, voltage amplitude and voltage phase angle constitution quantity of each node, and constructing a state quantity by the voltage amplitude and the voltage phase angle;
the state quantity prediction module is used for respectively constructing a voltage amplitude prediction model and a voltage phase angle prediction model based on the long-short-term memory network; determining a volume point based on a third-order-spherical radial volume rule, and taking the volume point as an input value of a voltage amplitude prediction model and a voltage phase angle prediction model to obtain a voltage amplitude prediction value and a voltage phase angle prediction value;
a state quantity filtering module for performing adaptive volume Kalman filtering on the voltage amplitude predicted value and the voltage phase angle predicted value, wherein the time-varying noise correction factor is utilized to perform time-varying correction on the system noise covariance matrix,
and the dynamic state estimation module is used for respectively averaging the filtered voltage amplitude predicted value and the voltage phase angle predicted value to obtain a dynamic state estimation result of the power system.
The present disclosure may be a system, method, and/or computer program product. The computer program product may include a computer readable storage medium having computer readable program instructions embodied thereon for causing a processor to implement aspects of the present disclosure.
The computer readable storage medium may be a tangible device that can hold and store instructions for use by an instruction execution device. The computer readable storage medium may be, for example, but not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium would include the following: portable computer disks, hard disks, random Access Memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static Random Access Memory (SRAM), portable compact disk read-only memory (CD-ROM), digital Versatile Disks (DVD), memory sticks, floppy disks, mechanical coding devices, punch cards or in-groove structures such as punch cards or grooves having instructions stored thereon, and any suitable combination of the foregoing. Computer-readable storage media, as used herein, are not to be construed as transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., optical pulses through fiber optic cables), or electrical signals transmitted through wires.
The computer readable program instructions described herein may be downloaded from a computer readable storage medium to a respective computing/processing device or to an external computer or external storage device over a network, such as the internet, a local area network, a wide area network, and/or a wireless network. The network may include copper transmission cables, fiber optic transmissions, wireless transmissions, routers, firewalls, switches, gateway computers and/or edge servers. The network interface card or network interface in each computing/processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium in the respective computing/processing device.
Computer program instructions for performing the operations of the present disclosure can be assembly instructions, instruction Set Architecture (ISA) instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, c++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The computer readable program instructions may be executed entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computer (for example, through the Internet using an Internet service provider). In some embodiments, aspects of the present disclosure are implemented by personalizing electronic circuitry, such as programmable logic circuitry, field Programmable Gate Arrays (FPGAs), or Programmable Logic Arrays (PLAs), with state information of computer readable program instructions, which can execute the computer readable program instructions.
Finally, it should be noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the above embodiments, it should be understood by those skilled in the art that: modifications and equivalents may be made to the specific embodiments of the invention without departing from the spirit and scope of the invention, which is intended to be covered by the claims.
Claims (13)
1. The utility model provides a power system dynamic state estimation method based on long-term memory network, which is characterized by comprising the following steps:
the method comprises the steps of measuring the active power, reactive power, voltage amplitude and voltage phase angle of each node to form a quantity, and forming a state quantity by the voltage amplitude and the voltage phase angle;
respectively constructing a voltage amplitude prediction model and a voltage phase angle prediction model based on the long-short term memory network; determining a volume point based on a third-order-spherical radial volume rule, and taking the volume point as an input value of a voltage amplitude prediction model and a voltage phase angle prediction model to obtain a voltage amplitude prediction value and a voltage phase angle prediction value;
performing volume Kalman filtering on the voltage amplitude predicted value and the voltage phase angle predicted value, wherein time-varying noise correction factors are utilized to perform time-varying correction on a system noise covariance matrix;
and respectively averaging the filtered voltage amplitude predicted value and the voltage phase angle predicted value to obtain a dynamic state estimation result of the power system.
2. The method for estimating dynamic state of electric power system based on long-term memory network according to claim 1, wherein,
the network structure of the voltage amplitude prediction model and the network structure of the voltage phase angle prediction model are the same.
3. The method for estimating dynamic state of electric power system based on long-term memory network according to claim 1, wherein,
determining the volume point based on the third-order-sphere radial volume rule includes:
and (3) performing Cholesky decomposition on the error covariance matrix at the moment k-1 to obtain:
wherein S is k-1 Representing an error covariance matrix L k-1 Square root matrix of (a);
the volume point is determined by the following relation:
in the method, in the process of the invention,
representing the state variable corresponding to the i-th volume point at the time of k-1,
ξ i represent S k-1 The corresponding matrix of coefficients is used to determine,
representing the state variable estimate at time k-1,
i is an n-order identity matrix,
2n is the number of volume points.
4. The method for estimating dynamic state of power system based on long-term memory network according to claim 3, wherein,
inputting the volume point into a voltage amplitude prediction model and a voltage phase angle prediction model to obtain a state quantity prediction value corresponding to the ith volume point at the k moment, wherein the method specifically comprises the following steps of:
and 2n volume points, wherein each volume point corresponds to a state quantity predicted value, and the state quantity predicted value at the k moment is obtained by averaging the state quantity predicted values corresponding to all the volume points.
5. The method for estimating dynamic state of power system based on long-term memory network as claimed in claim 4, wherein,
calculating an error covariance matrix prediction value according to the following relation:
in the method, in the process of the invention,
2n is the number of volume points,
the state quantity predicted value corresponding to the ith volume point at the k moment is expressed,
the state quantity predicted value at the time k is indicated.
A predicted value representing the k-time error covariance matrix,
M k-1 a system noise covariance matrix at time k-1 is shown.
6. The method for estimating dynamic state of electric power system based on long-term memory network according to claim 1, wherein,
the time-varying noise correction factor satisfies the following relationship:
in the method, in the process of the invention,
is a time-varying noise correction factor at time k-1,
beta is denoted as forgetting factor.
7. The method for estimating dynamic state of power system based on long-term memory network as claimed in claim 6, wherein,
the time-varying corrected system noise covariance matrix is as follows:
in the method, in the process of the invention,
M k represented as a system noise covariance matrix at time k,
K k denoted as the kalman gain at time k,
L k represented as an error covariance matrix at time k,
ε k denoted as the residual at time k.
8. The method for estimating dynamic state of power system based on long-term memory network as claimed in claim 7, wherein,
performing bulk kalman filtering on the voltage magnitude prediction value and the voltage phase angle prediction value includes: the root mean square error is introduced as the filtering precision, and the calculation formula is as follows:
in the method, in the process of the invention,
is the true value of the state quantity of the jth node at the moment k,
the state quantity predicted value of the j-th node at the k moment,
when the filtering precision is satisfied, the filtering is ended.
9. The method for estimating dynamic state of power system based on long-term memory network as claimed in claim 7, wherein,
calculating a measurement error autocovariance matrix C in the following relation k And a measurement error cross covariance matrix D k ：
In the method, in the process of the invention,
the predicted value is measured for the amount at time k,
representing the amount of measurement corresponding to the ith volume point at time k,
R k to measure the noise covariance matrix.
10. The method for estimating dynamic state of power system based on long-term memory network as claimed in claim 9, wherein,
the Kalman gain K is calculated according to the following relation k ：
K k ＝C k D k -1
Calculating to obtain a k moment state quantity and an error covariance matrix by using a Kalman gain:
in the method, in the process of the invention,
z k for the measurement of the quantity at time k,
the state quantity at time k.
11. An electric power system dynamic state estimation system based on a long-short-term memory network, which is used for realizing the method of any one of claims 1 to 10,
the system comprises: the system comprises a state quantity acquisition module, a state quantity prediction module, a state quantity filtering module and a dynamic state estimation module;
the state quantity acquisition module is used for measuring the active power, reactive power, voltage amplitude and voltage phase angle constitution quantity of each node, and constructing a state quantity by the voltage amplitude and the voltage phase angle;
the state quantity prediction module is used for respectively constructing a voltage amplitude prediction model and a voltage phase angle prediction model based on the long-short-term memory network; determining a volume point based on a third-order-spherical radial volume rule, and taking the volume point as an input value of a voltage amplitude prediction model and a voltage phase angle prediction model to obtain a voltage amplitude prediction value and a voltage phase angle prediction value;
a state quantity filtering module for performing adaptive volume Kalman filtering on the voltage amplitude predicted value and the voltage phase angle predicted value, wherein the time-varying noise correction factor is utilized to perform time-varying correction on the system noise covariance matrix,
and the dynamic state estimation module is used for respectively averaging the filtered voltage amplitude predicted value and the voltage phase angle predicted value to obtain a dynamic state estimation result of the power system.
12. A terminal comprising a processor and a storage medium; the method is characterized in that:
the storage medium is used for storing instructions;
the processor being operative according to the instructions to perform the steps of the method of any one of claims 1-10.
13. Computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the steps of the method according to any of claims 1-10.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311165743.XA CN117318021A (en) | 2023-09-11 | 2023-09-11 | Electric power system dynamic state estimation method and system based on long-term and short-term memory network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311165743.XA CN117318021A (en) | 2023-09-11 | 2023-09-11 | Electric power system dynamic state estimation method and system based on long-term and short-term memory network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117318021A true CN117318021A (en) | 2023-12-29 |
Family
ID=89259418
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311165743.XA Pending CN117318021A (en) | 2023-09-11 | 2023-09-11 | Electric power system dynamic state estimation method and system based on long-term and short-term memory network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117318021A (en) |
-
2023
- 2023-09-11 CN CN202311165743.XA patent/CN117318021A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Xiao et al. | Accurate state-of-charge estimation approach for lithium-ion batteries by gated recurrent unit with ensemble optimizer | |
| Yu et al. | Model parameter identification for lithium batteries using the coevolutionary particle swarm optimization method | |
| CN108089135B (en) | Extreme learning model-based battery state prediction system and implementation method thereof | |
| CN108155648A (en) | Method for estimating state based on the infinite Extended Kalman filter of adaptive H | |
| CN112488396A (en) | Wavelet transform-based electric power load prediction method of Holt-Winters and LSTM combined model | |
| Wei et al. | Lithium-ion battery modeling and state of charge estimation | |
| WO2023184824A1 (en) | Method and system for estimating state of charge of battery cluster, electronic device, and storage medium | |
| CN113671381A (en) | Lithium ion power battery estimation method based on time convolution network | |
| CN115221798B (en) | Battery thermal process space-time modeling prediction method, system, equipment and medium | |
| CN113190969B (en) | Lithium battery model parameter identification method based on information evaluation mechanism | |
| CN113608121A (en) | Lithium battery SOC estimation method based on fuzzy fractional order unscented Kalman filtering | |
| CN110554320A (en) | SOC estimation method of lithium ion battery | |
| CN115932586A (en) | Method, device, equipment and medium for estimating state of charge of battery on line | |
| CN116754959A (en) | SOC estimation method based on improved GWO optimized forgetting factor on-line parameter identification | |
| CN115407205A (en) | Multi-time scale SOC and SOH collaborative estimation method considering temperature influence | |
| CN114239796A (en) | Power system state estimation method based on extended Kalman filtering | |
| CN117420447A (en) | Lithium battery SOC estimation method and system considering noise deviation compensation and electronic device | |
| CN115878963B (en) | Method, system, terminal and medium for predicting metering error of capacitive voltage transformer | |
| CN113093014A (en) | Online collaborative estimation method and system for SOH and SOC based on impedance parameters | |
| CN116449219A (en) | Lithium battery model parameter identification method for improving VFFRLS | |
| CN117318021A (en) | Electric power system dynamic state estimation method and system based on long-term and short-term memory network | |
| CN116054780A (en) | Particle filtering method based on state track clustering similarity and application | |
| 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 | |
| CN114924191A (en) | Method and device for estimating battery capacity on line | |
| Zhu et al. | State estimation based on improved cubature Kalman filter algorithm |
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 |