CN-121980491-A - Virtual power plant parameter self-learning method

CN121980491ACN 121980491 ACN121980491 ACN 121980491ACN-121980491-A

Abstract

The invention relates to the technical field of virtual power plants, in particular to a virtual power plant parameter self-learning method which comprises the steps of collecting multi-source measurement data of virtual power plant equipment, wherein the multi-source measurement data comprise a plurality of virtual power plant equipment parameters, conducting standardization processing on the multi-source measurement data to obtain measurement vectors, conducting preliminary estimation on the measurement vectors by adopting a Kalman filtering algorithm to obtain state vectors, conducting parameter identification on the measurement vectors by adopting a recursive least square method to obtain parameter estimation vectors, obtaining nominal parameter vectors of the virtual power plant equipment, conducting weighted fusion on the nominal parameter vectors, the state vectors and the parameter estimation vectors to obtain fusion parameter vectors, calculating comprehensive confidence according to the parameter estimation vectors and the measurement vectors, judging the effectiveness of the fusion parameter vectors according to the comprehensive confidence, obtaining judgment results, and using parameters or triggering alarm according to the judgment results, so that the problem that the calculation accuracy of aggregation capacity is low due to the fact that an existing system uses static nominal parameters is solved.

Inventors

HUANG CHUZHI
ZHUANG SHAOYANG
Wang Zita
HUANG ZONGWEI
CAI SHANSHAN
CHEN WEIPENG

Assignees

泉州亿兴电力工程建设有限公司鲤城自动化分公司

Dates

Publication Date: 20260505
Application Date: 20251223

Claims (10)

1. The virtual power plant parameter self-learning method is characterized by comprising the following steps of: collecting multi-source measurement data of virtual power plant equipment, wherein the multi-source measurement data comprises a plurality of virtual power plant equipment parameters, and performing standardization processing on the multi-source measurement data to obtain measurement vectors; Performing preliminary estimation on the measurement vector by adopting a Kalman filtering algorithm to obtain a state vector; Carrying out parameter identification on the measurement vector by adopting a recursive least square method to obtain a parameter estimation vector; the nominal parameter vector of the virtual power plant equipment is obtained, and the nominal parameter vector, the state vector and the parameter estimation vector are subjected to weighted fusion to obtain a fusion parameter vector; calculating comprehensive confidence coefficient according to the parameter estimation vector and the measurement vector; and judging the validity of the fusion parameter vector according to the comprehensive confidence coefficient, obtaining a judging result, and using parameters or triggering an alarm according to the judging result.
2. The method for self-learning parameters of a virtual power plant according to claim 1, wherein the preliminary estimation of the measurement vector by using a kalman filter algorithm is performed to obtain a state vector, and specifically comprises: Establishing a Kalman filtering system based on virtual power plant equipment parameters, wherein the Kalman filtering system comprises a state equation and an observation equation, the state equation is used for describing the change of the virtual power plant equipment parameters with time, the observation equation is used for describing the relation between multi-source measurement data of the virtual power plant equipment and the state of the virtual power plant equipment, and initializing a state vector, a first error covariance matrix, a process noise covariance matrix and an observation noise covariance matrix, wherein the state vector is defined as: ; Wherein, the The state vector is represented as a function of the state vector, The i-th parameter is indicated as such, N represents the number of virtual power plant parameters; predicting a state vector at the current moment according to a state prediction formula, wherein the state prediction formula is shown as follows: ; wherein k is a time mark, Representing an a priori state estimate of the current time, Representing the posterior state estimate at the last instant, a representing the state transition matrix, B representing the control input matrix, A control input representing a current time; Predicting the first error covariance matrix according to an error covariance prediction formula, wherein the error covariance prediction formula is shown as follows: ; Wherein, the Representing an a priori first error covariance matrix at the current instant, A first error covariance matrix representing the last moment, T representing a transpose matrix, and Q representing a process noise covariance matrix; the observation equation is shown as follows: ; Wherein, the Representing an output vector of a Kalman filtering system, and taking the measurement vector as the output vector of the Kalman filtering system; Representing the state vector at the current moment, C representing the observation matrix of the kalman filter system, Representing measurement noise of the Kalman filtering system; And calculating Kalman gain according to the prior first error covariance matrix, wherein the calculation formula is shown as follows: ; Wherein, the The gain of kalman is indicated as such, Representing a measurement noise covariance matrix; Correcting the prior state estimation at the current moment according to the measurement vector and the Kalman gain to obtain posterior state estimation at the current moment, namely a state vector at the current moment, so as to update the state vector, wherein the calculation formula is shown as follows: ; Correcting the prior first error covariance matrix at the current moment according to the Kalman gain to obtain a first error covariance matrix at the current moment, so as to update the first error covariance matrix, wherein the calculation formula is as follows: ; Wherein I represents an identity matrix.
3. The method for self-learning parameters of a virtual power plant according to claim 2, wherein the step of performing parameter identification on the measurement vector by using a recursive least square method to obtain a parameter estimation vector comprises the following steps: a linear parameter model is constructed as shown in the following formula: ; Wherein, the Representing an output vector of the linear parameter model, and taking the measurement vector as the output vector of the linear parameter model; representing regression vectors, defined as ; Representing a model parameter vector to be identified; An observation noise representing a linear parametric model; parameter identification is carried out by adopting a forgetting factor recursive least square method, and an identification equation is shown as follows: ; ; Wherein, the Representing a parameter estimation vector; representing a second error covariance matrix; The forgetting factor is represented, and the value range is (0.95, 0.99).
4. A method for self-learning parameters of a virtual power plant according to claim 3, wherein the obtaining a nominal parameter vector of the virtual power plant, and performing weighted fusion on the nominal parameter vector, the state vector and the parameter estimation vector to obtain a fused parameter vector, specifically comprises: Obtaining a nominal parameter vector of the virtual power plant equipment; the ith parameter of the nominal parameter vector satisfies , An ith parameter representing a nominal parameter vector, Representation of Is the mean value, Is a normal distribution of variance, the ith parameter of the state vector is expressed as Obeying the expected value of Variance is Is expressed as The ith parameter of the parameter estimation vector is expressed as Obeying the expected value of Variance is Is expressed as ; Fusing the nominal parameter vector H, the state vector and the parameter estimation vector according to a Bayesian theory to obtain posterior distribution of the fused parameter vector: ; and calculates the ith parameter of the fusion parameter vector as shown in the following formula: ; Wherein, the The i parameter of the fusion parameter vector is represented, j is the mark of the information source, and the marks of the information source corresponding to the nominal parameter vector, the state vector and the parameter estimation vector are respectively 1,2 and 3; a weight factor representing a jth information source; indicating the expected value corresponding to the jth information source, Representing the variance corresponding to the jth information source.
5. The method for self-learning parameters of a virtual power plant according to claim 1, wherein the calculating the integrated confidence level according to the parameter estimation vector and the measurement vector specifically comprises: and calculating parameter convergence evaluation according to the parameter estimation vector, wherein the calculation formula is as follows: ; Wherein, the Parameter convergence evaluation of an i-th parameter representing a parameter estimation vector; calculating a parameter stability score according to the parameter convergence evaluation, wherein the calculation formula is as follows: ; Wherein, the Representing a parameter stability score; calculating a multi-source measurement data quality score according to the measurement vector, wherein the calculation formula is as follows: ; ; Wherein, the Representing the multi-source measured data quality score, exp represents an exponential function based on a natural constant e, Representing a residual error between the measurement vector and the state vector; And calculating comprehensive confidence according to the parameter stability score and the multi-source measurement data quality score, wherein the calculation formula is as follows: ; Wherein, the The integrated confidence level is represented as a function of the integrated confidence level, Representing the adjustable parameter.
6. The method for self-learning parameters of a virtual power plant according to claim 1, wherein the step of judging the validity of the fused parameter vector according to the comprehensive confidence level to obtain a judgment result, and using parameters or triggering an alarm according to the judgment result specifically comprises: setting a judgment threshold, triggering parameters to reinitiate and/or trigger alarm signals if the comprehensive confidence coefficient is smaller than the judgment threshold, otherwise marking the parameters in the fusion parameter vector to be valid, and using the parameters corresponding to the comprehensive confidence coefficient.
7. A virtual power plant parameter self-learning device, comprising: the data acquisition module is configured to acquire multi-source measurement data of the virtual power plant equipment, wherein the multi-source measurement data comprises a plurality of virtual power plant equipment parameters, and the multi-source measurement data is subjected to standardization processing to obtain measurement vectors; The Kalman filtering module is configured to perform preliminary estimation on the measurement vector by adopting a Kalman filtering algorithm to obtain a state vector; The parameter identification module is configured to conduct parameter identification on the measurement vector by adopting a recursive least square method to obtain a parameter estimation vector; the fusion module is configured to acquire a nominal parameter vector of the virtual power plant equipment, and perform weighted fusion on the nominal parameter vector, the state vector and the parameter estimation vector to acquire a fusion parameter vector; a quality assessment module configured to calculate a comprehensive confidence level from the parameter estimation vector and the measurement vector; The judging module is configured to judge the validity of the fusion parameter vector according to the comprehensive confidence coefficient, obtain a judging result, and use parameters or trigger an alarm according to the judging result.
8. An electronic device, comprising: One or more processors; A memory for storing one or more programs, When executed by the one or more processors, causes the one or more processors to implement the method of any of claims 1-6.
9. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the method according to any of claims 1-6.
10. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any of claims 1-6.

Description

Virtual power plant parameter self-learning method Technical Field The invention relates to the technical field of virtual power plants, in particular to a virtual power plant parameter self-learning method. Background Along with the large-scale grid connection of new energy, the fluctuation and uncertainty of a power system are obviously increased, and higher requirements are put on the adjusting capability of the system. Virtual power plants ((VirtualPowerPlant, VPP)) can provide important regulatory support for power systems by aggregating distributed power, energy storage, adjustable load, and other resources. In VPP operation, the electrical parameters of the distributed power sources (wind, photovoltaic, energy storage) change as the device decays, the environment changes. The existing system uses static nominal parameters to cause low calculation precision of the aggregate capacity, error is 10-25%, actual response capability is inconsistent with expected response capability, and VPP scheduling failure or potential safety hazard exists. Disclosure of Invention The application aims to provide a virtual power plant parameter self-learning method aiming at the technical problems. In a first aspect, the invention provides a method for self-learning parameters of a virtual power plant, comprising the following steps: Collecting multi-source measurement data of virtual power plant equipment, wherein the multi-source measurement data comprises a plurality of virtual power plant equipment parameters, and performing standardization processing on the multi-source measurement data to obtain measurement vectors; Carrying out preliminary estimation on the measurement vector by adopting a Kalman filtering algorithm to obtain a state vector; Carrying out parameter identification on the measurement vector by adopting a recursive least square method to obtain a parameter estimation vector; The method comprises the steps of obtaining a nominal parameter vector of virtual power plant equipment, and carrying out weighted fusion on the nominal parameter vector, a state vector and a parameter estimation vector to obtain a fusion parameter vector; calculating comprehensive confidence coefficient according to the parameter estimation vector and the measurement vector; and judging the validity of the fusion parameter vector according to the comprehensive confidence coefficient, obtaining a judging result, and using parameters or triggering an alarm according to the judging result. Preferably, a Kalman filtering algorithm is adopted to perform preliminary estimation on the measurement vector to obtain a state vector, and the method specifically comprises the following steps: Establishing a Kalman filtering system based on virtual power plant equipment parameters, wherein the Kalman filtering system comprises a state equation and an observation equation, the state equation is used for describing the change of the virtual power plant equipment parameters along with time, the observation equation is used for describing the relation between multi-source measurement data of the virtual power plant equipment and the state of the virtual power plant equipment, and initializing a state vector, a first error covariance matrix, a process noise covariance matrix and an observation noise covariance matrix, wherein the state vector is defined as: ; Wherein, the The state vector is represented as a function of the state vector,The i-th parameter is indicated as such,N represents the number of virtual power plant parameters; predicting a state vector at the current moment according to a state prediction formula, wherein the state prediction formula is shown as follows: ; wherein k is a time mark, Representing an a priori state estimate of the current time,Representing the posterior state estimate at the last instant, a representing the state transition matrix, B representing the control input matrix,A control input representing a current time; Predicting the first error covariance matrix according to an error covariance prediction formula, wherein the error covariance prediction formula is shown as follows: ; Wherein, the Representing an a priori first error covariance matrix at the current instant,A first error covariance matrix representing the last moment, T representing a transpose matrix, and Q representing a process noise covariance matrix; the observation equation is shown as follows: ; Wherein, the Representing the output vector of the Kalman filtering system, and taking the measurement vector as the output vector of the Kalman filtering system; Representing the state vector at the current moment, C representing the observation matrix of the kalman filter system, Representing measurement noise of the Kalman filtering system; And calculating Kalman gain according to the prior first error covariance matrix, wherein the calculation formula is shown as follows: ; Wherein, the The gain of kalman is indicated as such,Representing a me