CN115758884A - Digital twin construction method for distributed resources of virtual power plant - Google Patents
Digital twin construction method for distributed resources of virtual power plant Download PDFInfo
- Publication number
- CN115758884A CN115758884A CN202211444526.XA CN202211444526A CN115758884A CN 115758884 A CN115758884 A CN 115758884A CN 202211444526 A CN202211444526 A CN 202211444526A CN 115758884 A CN115758884 A CN 115758884A
- Authority
- CN
- China
- Prior art keywords
- power plant
- digital twin
- model
- simulation module
- output
- 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
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Abstract
The invention provides a digital twin construction method of distributed resources of a virtual power plant, which comprises the following steps: s1, constructing a digital twin body simulation module model of each power plant in distributed resources; s2, collecting input variables of the simulation module model to form an input vector set; s3, collecting output variables of the simulation module model to form an output vector set; s4, collecting model parameters of the simulation module model to form a model parameter set; s5, collecting data to form a sample data set, and determining an optimization objective function; the digital twin simulation module model of the photovoltaic power plant and the wind power plant adopts various input variables, the historical data of the related input variables is easy to collect, accurate prediction of the near day (the current day and the next few days) can be performed, the accurate prediction of the power plant output by using the digital twin simulation module model after the model is established is facilitated, and the optimized scheduling of the electric power and the optimized distribution of the electric power resources are facilitated on the basis of the accurate prediction of the power plant output.
Description
Technical Field
The invention relates to the technical field of virtual power plants, in particular to a digital twin construction method for distributed resources of a virtual power plant.
Background
The virtual power plant is a power supply coordination management system which realizes the aggregation and coordination optimization of DER (distributed generation), an energy storage system, controllable loads, electric vehicles and the like through an advanced information communication technology and a software system and is used as a special power plant to participate in the operation of a power market and a power grid. The core of the virtual plant concept can be summarized as "communication" and "aggregation". The key technologies of the virtual power plant mainly comprise a coordination control technology, an intelligent metering technology and an information communication technology. The most attractive function of the virtual power plant is to enable the aggregation of DER to participate in the operation of the power market and the auxiliary service market, and provide management and auxiliary services for the power distribution network and the power transmission network.
The digital twin refers to the full utilization of data such as physical models, sensor updating, operation history and the like, the integration of multidisciplinary, multi-physical quantity, multi-scale and multi-probability simulation processes and the completion of mapping in a virtual space so as to reflect the full life cycle process of corresponding entity equipment;
the virtual power plant has wide resource distribution range, the power plant output is inaccurate in prediction in actual use, the optimal scheduling of power and the optimal distribution of power resources are influenced, and meanwhile, the individual optimization and the overall optimization of the virtual power plant are inconsistent with each other along with the continuous operation of the power plant;
therefore, a digital twin construction method of the distributed resources of the virtual power plant is provided.
Disclosure of Invention
In view of the above, embodiments of the present invention are intended to provide a method for constructing a digital twin of distributed resources of a virtual power plant, so as to solve or alleviate technical problems in the prior art, and to provide at least one useful choice.
The technical scheme of the embodiment of the invention is realized as follows: a digital twin construction method of distributed resources of a virtual power plant comprises the following steps:
s1, constructing a digital twin body simulation module model of each power plant in distributed resources;
s2, collecting input variables of the simulation module model to form an input vector set;
s3, collecting output variables of the simulation module model to form an output vector set;
s4, collecting model parameters of the simulation module model to form a model parameter set;
s5, collecting data to form a sample data set, and determining an optimization objective function;
s6, determining the value range of the model parameters, and performing parameter optimization to obtain optimized parameter values;
and S7, performing simulation calculation and output prediction by using the module model, and updating the optimized sample data.
Further preferably: in the step S1, when the digital twin simulation module model is constructed, the method includes the following steps:
s11, selecting one power plant in distributed resources of the virtual power plant, selecting the type of a digital twin simulation module model of the power plant, determining an input variable and an output variable of the digital twin simulation module model of the power plant, and collecting historical data of the input variable and the output variable;
s12, dividing the collected historical data into training sample data and optimization sample data of the power plant digital twin simulation module model;
s13, training the power plant digital twin simulation module model by using training sample data of the power plant digital twin simulation module model;
s14, obtaining a model parameter reference value of the power plant digital twin simulation module model after training is finished;
s15, repeating the steps S11-S14 until the construction of the digital twin simulation module models of all the power plants in the distributed resources of the virtual power plants is completed.
Further preferred is: in step S2, the input vector set is [ D ] 11 ,D 12 ,…,D 1m1 ,…,D 2m2 ,D 31 ,…,D k1 ,D k2 ,…,D kmk ]；
Wherein k is the number of digital twin simulation module models of the power plant, [ D 11 ,D 12 ,…,D 1m1 ]For m1 input variables, [ D ] of the 1 st power plant digital twin simulation module model k1 ,D k2 ,…,D kmk ]Mk input variables of the kth power plant digital twin simulation module model; d 2m2 The number of the m2 th input variables of the 2 nd power plant digital twin simulation module model; d 31 The number of the 1 st input variables of the 3 rd power plant digital twin simulation module model; m1, m2, \ 8230, mk are respectively 1 st, 2 nd, \8230, and the number of input variables of the kth power plant digital twin simulation module model.
Further preferably: in the step S3, the output vector set is [ C ] 11 ,C 12 ,…,C 1i ,…,C 2i ,C 31 ,…,C k1 ,C k2 ,…,C ki ]；
Wherein [ C 11 ,C 12 ,…,C 1i ]I output variables, C, for the 1 st plant digital twin simulation module model 2i For the ith output variable, C, of the 2 nd power plant digital twin simulation module model 31 For the 1 st output variable of the 3 rd plant digital twin simulation module model, [ C [ ] k1 ,C k2 ,…,C ki ]I output variables of the digital twin simulation module model of the kth power plant.
Further preferably: in step S4, the model parameter set is [ F 11 ,F 12 ,…,F 1n1 ,…,F 2n2 ,F 31 ,…,F k1 ,F k2 ,…,F knk ]；
Wherein, [ F ] 11 ,F 12 ,…,F 1n1 ]N1 model parameters for the 1 st plant digital twin simulation module model，F 2n2 For the n2 model parameter, F, of the 2 nd power plant digital twin simulation module model 31 For the 1 st model parameter, [ F ] of the 3 rd plant digital twin simulation module model k1 ,F k2 ,…,F knk ]Nk model parameters of a kth power plant digital twin simulation module model; n1, n2, \ 8230, nk are respectively 1 st, 2 nd, \8230, and the number of model parameters of the kth power plant digital twin simulation module model.
Further preferred is: in step S5, the optimization objective is the minimization of the absolute value of the model weighted relative error and Q, and the optimization objective function is:
wherein E is the weighted total output, having
Or any of the equivalent writes:
wherein, p is the optimization sample data of the 1 st to w th groups in the optimization sample data set; r stands for the 1 st to k th power plant digital twin simulation module model, δ r The power plant output weight coefficient corresponding to the r power plant; q is the 1 st to the ith output of the power plant digital twin simulation module model; e prq The error between the qth output of the ith power plant digital twin simulation module model of the pth group of optimization sample data and the corresponding actual output value of the optimization sample data (i.e., the qth actual output value of the ith power plant of the pth group of optimization sample data), O prq Optimizing the qth actual output value of the r power plant of the sample data for the pth group;
the mean value of the qth actual output value of the qth power plant in the w groups of optimized sample data; mu.s q Is the output variable weight coefficient.
For the power plant output weight coefficient delta r And the output variable weight coefficient mu q Respectively have
P r For the historical average output active power of the r-th distributed plant, a factor δ of the plant output weight r The method comprises the following steps:
further preferably: in the step S6, the value range of each model parameter in the model parameter set is determined; the value range of the model parameters in the model parameter set is between 1 time and lambda time of the lambda fraction of the model parameter reference value; λ is 2 or more and 10 or less;
and after the value range of each model parameter in the model parameter set is determined, the method further comprises the following steps:
performing iterative optimization calculation on the model parameters in the model parameter set by adopting an optimization algorithm; and after the optimization calculation is finished, taking the optimal solution of the model parameter set as a new model parameter reference value.
Further preferred is: in the step S7, when the digital twin simulation module models of each power plant are used for performing simulation calculation and output prediction:
continuously collecting sample data of input variables and output variables of the digital twin simulation module model of each power plant;
when the data volume of the continuously acquired sample data reaches or exceeds 20% of the data volume of the optimized sample data set, replacing the data in the optimized sample data set by the newly acquired sample data by adopting a rolling method;
and repeating the steps S6-S7 to continuously integrate and optimize the model parameters of the digital twin simulation module of the power plant in each distributed resource.
The invention also provides a digital twin construction system of the distributed resources of the virtual power plant, which comprises the following steps:
the training module is used for constructing a digital twin body simulation module model of each power plant in the distributed resources;
the acquisition module is used for acquiring an input vector set formed by input variables of the simulation module model, an output vector set formed by output variables of the simulation module model and a model parameter set formed by model parameters of the simulation module model;
the target optimization module is used for collecting data to form a sample data set, determining an optimization objective function, and determining the value range of a model parameter to optimize the parameter to obtain an optimization parameter value;
and the prediction module is used for performing simulation calculation and output prediction by using the module model and updating the optimized sample data.
The invention also provides a computer device comprising a processor, a memory coupled to the processor, and program instructions stored in the memory, which when executed by the processor, cause the processor to perform the steps of the method of digital twinning of virtual power plant distributed resources as described in any of the above.
Due to the adoption of the technical scheme, the embodiment of the invention has the following advantages:
1. the digital twin simulation module model of the photovoltaic power plant and the wind power plant adopts various input variables, the historical data of the related input variables is easy to collect, accurate prediction of the near day (the current day and the next few days) can be performed, the accurate prediction of the power plant output by using the digital twin simulation module model after the model is established is facilitated, and the optimized scheduling of the electric power and the optimized distribution of the electric power resources are facilitated on the basis of the accurate prediction of the power plant output.
2. According to the invention, on the basis of independent modeling by using training data, the power plants in the virtual power plant uniformly optimize the overall parameters of all distributed power plant digital twin simulation module models by using the integrated optimization sample data, collect and accumulate new historical data in the process of simulation, verification and prediction by using the power plant models, roll and replace part of old optimization sample data, and continuously and intermittently optimize the overall parameters of all distributed power plant digital twin simulation module models, so that the distributed power plant individuals are always in a satisfactory running state, and the individual optimization and the overall optimization compatibility of the virtual power plant are ensured to be consistent.
The foregoing summary is provided for the purpose of description only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of the present invention will be readily apparent by reference to the drawings and following detailed description.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the embodiments or technical descriptions will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a flow chart of the method of the present invention.
Detailed Description
In the following, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments may be modified in various different ways, all without departing from the spirit or scope of the present invention. Accordingly, the drawings and description are to be regarded as illustrative in nature, and not as restrictive.
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
Example one
As shown in fig. 1, an embodiment of the present invention provides a digital twin construction method for distributed resources of a virtual power plant, including the following steps:
s1, constructing a digital twin body simulation module model of each power plant in distributed resources;
when the digital twin simulation module model is constructed, the method comprises the following steps:
s11, selecting one power plant in distributed resources of the virtual power plant, selecting the type of a digital twin simulation module model of the power plant, determining input variables and output variables of the digital twin simulation module model of the power plant and collecting historical data of the input variables and the output variables;
s12, dividing the collected historical data into training sample data and optimization sample data of the power plant digital twin simulation module model;
s13, training the power plant digital twin simulation module model by using training sample data of the power plant digital twin simulation module model;
s14, obtaining a model parameter reference value of the power plant digital twin simulation module model after training is finished;
s15, repeating the steps S11-S14 until the construction of the digital twin simulation module models of all the power plants in the distributed resources of the virtual power plants is completed;
s2, collecting input variables of the simulation module model to form an input vector set;
the input vector set is [ D ] 11 ,D 12 ,…,D 1m1 ,…,D 2m2 ,D 31 ,…,D k1 ,D k2 ,…,D kmk ]；
Wherein k is the number of digital twin simulation module models of the power plant, [ D 11 ,D 12 ,…,D 1m1 ]For m1 input variables, [ D ] of the 1 st power plant digital twin simulation module model k1 ,D k2 ,…,D kmk ]Mk input variables of the kth power plant digital twin simulation module model; d 2m2 The number of the (m 2) th input variables of the 2 nd power plant digital twin simulation module model; d 31 The number of the 1 st input variables of the 3 rd power plant digital twin simulation module model; m1, m2, \ 8230, mk are respectively 1 st, 2 nd, \8230, and the number of input variables of the kth power plant digital twin simulation module model;
s3, collecting output variables of the simulation module model to form an output vector set;
set of output vectors as [ C 11 ,C 12 ,…,C 1i ,…,C 2i ,C 31 ,…,C k1 ,C k2 ,…,C ki ]；
Wherein [ C 11 ,C 12 ,…,C 1i ]For the i output variables of the 1 st power plant digital twin simulation module model, C 2i For the ith output variable, C, of the 2 nd power plant digital twin simulation module model 31 For the 1 st output variable of the 3 rd plant digital twin simulation module model, [ C [ ] k1 ,C k2 ,…,C ki ]I output variables of the digital twin simulation module model of the kth power plant;
s4, collecting model parameters of the simulation module model to form a model parameter set;
the model parameter set is [ F ] 11 ,F 12 ,…,F 1n1 ,…,F 2n2 ,F 31 ,…,F k1 ,F k2 ,…,F knk ]；
Wherein, [ F ] 11 ,F 12 ,…,F 1n1 ]N1 model parameters, F, for the 1 st plant digital twin simulation module model 2n2 For the n2 model parameter, F, of the 2 nd power plant digital twin simulation module model 31 For the 1 st model parameter of the 3 rd plant digital twin simulation module model, [ F ] k1 ,F k2 ,…,F knk ]Nk model parameters of a kth power plant digital twin simulation module model; n1, n2, \ 8230, nk are respectively 1 st, 2 nd, \8230, and the number of model parameters of the kth power plant digital twin simulation module model;
s5, collecting data to form a sample data set, and determining an optimization objective function;
the optimization objective is the minimization of the model weighted relative error absolute value and Q, and the optimization objective function is:
wherein E is a weighted total output having
Or any of the equivalent writes:
wherein, p is the 1 st to w th groups of optimization sample data in the optimization sample data set; r represents the 1 st to kth power plant digital twin simulation module models, δ r A power plant output weight coefficient corresponding to the r-th power plant; q is the 1 st to the ith output of the power plant digital twin simulation module model; e prq The error between the qth output of the ith power plant digital twin simulation module model of the pth group of optimization sample data and the corresponding actual output value of the optimization sample data (i.e., the qth actual output value of the ith power plant of the pth group of optimization sample data), O prq Optimizing the qth actual output value of the qth power plant of the sample data for the pth group;
the mean value of the qth actual output value of the qth power plant in the w groups of optimized sample data; mu.s q Is the output variable weight coefficient.
For the power plant output weight coefficient delta r And the output variable weight coefficient mu q Respectively have
P r For the historical average output active power of the r-th distributed plant, a factor δ of the plant output weight r The method comprises the following steps:
s6, determining the value range of the model parameter, and performing parameter optimization to obtain an optimized parameter value;
determining the value range of each model parameter in the model parameter set; the value range of the model parameters in the model parameter set is between 1 time and lambda time of the lambda fraction of the model parameter reference value; λ is 2 or more and 10 or less;
and after the value range of each model parameter in the model parameter set is determined, the method further comprises the following steps:
performing iterative optimization calculation on the model parameters in the model parameter set by adopting an optimization algorithm; after the optimization calculation is finished, taking the optimal solution of the model parameter set as a new model parameter reference value
S7, performing simulation calculation and output prediction by using the module model, and updating optimized sample data;
when the digital twin simulation module models of each power plant are adopted for simulation calculation and output prediction:
continuously collecting sample data of input variables and output variables of the digital twin simulation module model of each power plant;
when the data volume of the continuously acquired sample data reaches or exceeds 20% of the data volume of the optimized sample data set, replacing the data in the optimized sample data set by the newly acquired sample data by adopting a rolling method;
repeating the steps S6-S7 to continuously integrate and optimize the model parameters of the power plant digital twin simulation module in each distributed resource;
the optimization algorithm adopts a genetic algorithm, or a particle swarm algorithm, or a wolf colony algorithm. The end condition of the optimization is that the sum Q of the weighted absolute errors of the models is smaller than a set error threshold epsilon, or the optimization calculation reaches a set iteration number.
i is greater than or equal to 1; the output variables of each power plant digital twin simulation module model at least comprise active power output; the active power output included in the output variable is daily average active power output; and the weight coefficient of an output variable corresponding to daily average active power output is not lower than 0.6.
Input variables of the photovoltaic power plant digital twin simulation module model comprise the month of day, the highest weather temperature of day, the average weather temperature of day, the weather state of day and the construction age of the photovoltaic power plant.
Input variables of the wind power plant digital twin simulation module model comprise daily average wind speed, daily maximum wind speed and the month (season) of the day.
The invention also provides a digital twin construction system of the distributed resources of the virtual power plant, which comprises the following steps:
the training module is used for constructing a digital twin body simulation module model of each power plant in the distributed resources;
the acquisition module is used for acquiring an input variable of the simulation module model to form an input vector set, acquiring an output variable of the simulation module model to form an output vector set and acquiring a model parameter set of the simulation module model;
the target optimization module is used for collecting data to form a sample data set, determining an optimization objective function, and determining the value range of a model parameter to optimize the parameter to obtain an optimization parameter value;
and the prediction module is used for performing simulation calculation and output prediction by using the module model and updating the optimized sample data.
The invention also provides a computer device comprising a processor, a memory coupled to the processor, the memory having stored therein program instructions which, when executed by the processor, cause the processor to perform the steps of the method of digital twinning of virtual power plant distributed resources as defined in any one of the preceding claims.
Example two
The present invention also provides an embodiment practiced according to an embodiment method: constructing a power plant digital twin simulation module model: specifically, a BP neural network is adopted to construct a photovoltaic power plant digital twin simulation module model.
The input variable of the photovoltaic power plant digital twin simulation module model is the month M of the day 1 Day maximum climate temperature T 1 Day weather condition A 1 . Wherein, the month of the dayM 1 Equal to the month value, e.g. data day at 3 months, then M 1 Equal to 3; day weather State A 1 Including 5 states of sunny, cloudy, rainy, snowy, etc., which are expressed by numbers 1, 2, 3, 4, and 5, respectively. The output variable of the digital twin simulation module model of the photovoltaic power plant is daily average active power output P G1 Sunday average reactive power output Q G1 . The second simulation model of the embodiment is used for predicting the month M of the day 1 For the calculated actual data of the day to be predicted, the daily maximum climate temperature T 1 Day weather condition A 1 The predicted day data to be predicted according to weather forecast and the like is adopted.
In a second embodiment, a three-layer BP network with hidden nodes of 4 is selected, wherein 3 × 4 weight coefficients of 12 weight coefficients from the input layer to the hidden nodes are set as V G1 To V G12 The threshold value of 4 hidden layer nodes is theta G1 To theta G4 (ii) a 4 multiplied by 2 between the hidden node and the output node and with 8 weight coefficients V G13 To V G20 And the threshold value of 2 output nodes is theta G5 、θ G6 . At the moment, the two embodiments of the power plant digital twin simulation module model have 26 model parameters which are weight coefficients V respectively G1 To V G20 Threshold value theta G1 To theta G6 . The excitation functions of the output layer and the hidden layer both adopt unipolar Sigmoid functions.
The historical data of the second embodiment of the power plant digital twin simulation module model is the month M of the day of at least 2 years 1 Daily maximum climate temperature T 1 Day weather condition A 1 And data, and corresponding daily average active power output and daily average reactive power output actual value data. Some of them (between 30% and 70%) are randomly selected, for example, half of the data are randomly selected as training sample data, and the rest half of the data are left as optimization sample data. Training a neural network by using selected training sample data and adopting a gradient descent method or optimization algorithms such as particle swarm optimization, genetic algorithm and the like, wherein the training target is that the average error of the system is minimum (the sum of squares of the average errors is minimum); setting weight coefficient V of neural network after training G1 To V G20 Respectively has a value of V gG1 To V gG20 Threshold θ of neural network after training is finished G1 To theta G6 Are each theta gG1 To theta gG6 Then V is gG1 To V gG20 And theta gG1 To theta gG6 The model parameter reference value of the second embodiment of the digital twin simulation module model of the power plant is shown. In addition, the average value of the actual daily average active power output in the training sample data of the second embodiment is calculated and is used as the historical average output active power P of the second embodiment of the power plant digital twin simulation module model 1 (ii) a Or calculating the average value of the actual daily average active power output in all historical data (including training sample data and optimization sample data) of the second embodiment, and taking the average value as the historical average output active power P of the second embodiment of the digital twin simulation module model of the power plant 1 。
EXAMPLE III
The present invention also provides an embodiment practiced according to an embodiment method: a RBF neural network is adopted to construct a photovoltaic power plant digital twin simulation module model, and an input variable is a month M of a day 2 Day-to-day average climate temperature T Q2 Day weather State A 2 Photovoltaic power plant construction life S 2 . Month of day M 2 Meaning of (1) and the month of day M of example two 1 Same, day weather State A 2 Meaning of (2) and the daily weather Condition A of example two 1 The same; day average climate temperature T Q2 The average value of the temperature (unified as centigrade) measured at the photovoltaic cell of the photovoltaic power plant in the daytime interval in 1 day is adopted, the daytime interval is determined according to common knowledge, the measurement interval is more than or equal to 30 minutes and less than or equal to 2 hours, for example, the average value of the temperature measured every 1 hour between 5 am and 7 am (namely the daytime interval) in 1 day is taken as the daytime average climate temperature T Q2 (ii) a Photovoltaic power plant construction life S 2 The value is 1 in the 1 st year of photovoltaic power plant construction, and the value is 2 in the 2 nd year, and so on. The output variable of the digital twin simulation module model of the photovoltaic power plant is daily average active power output P G2 Sunday average reactive power output Q G2 . The third simulation module model of the embodiment is used for predicting the month M of the day 2 Photovoltaic power plant construction life S 2 The average day climate temperature T is the actual data of the estimated day to be predicted Q2 Day weather State A 2 The predicted day data to be predicted according to weather forecast and the like is adopted.
The RBF neural network of the third embodiment adopts 5 hidden nodes with the function functions being Gaussian functions, 5 hidden nodes and 4 input M 2 、T Q2 、A 2 、S 2 The corresponding central points are respectively [ Z 11 ，Z 12 ，Z 13 ，Z 14 ]、[Z 21 ，Z 22 ，Z 23 ，Z 24 ]、[Z 31 ，Z 32 ，Z 33 ，Z 34 ]、[Z 41 ，Z 42 ，Z 43 ，Z 44 ]、[Z 51 ，Z 52 ，Z 53 ，Z 54 ](ii) a The normalization constants (base widths) of the 5 hidden nodes are respectively | φ Z1 ，φ z2 ，φ z3 ，φ z4 ，φ z5 L. the method is used for the preparation of the medicament. 5 hidden nodes and 2 output nodes P G2 、Q G2 Are respectively [ Y ] as the connection weight coefficients 11 ，Y 12 ，Y 13 ，Y 14 ，Y 15 ]、[Y 21 ，Y 22 ，Y 23 ，Y 24 ，Y 25 ]. At the moment, the three embodiments of the power plant digital twin simulation module model have 35 model parameters in common.
Example three the historical data is the month M of the day of at least 2 years 2 Day-to-day average climate temperature T Q2 Day weather State A 2 Photovoltaic power plant construction life S 2 And data, and corresponding daily average active power output and daily average reactive power output actual value data. Some of them (between 30% and 70%) are randomly selected, for example, half of the data are randomly selected as training sample data, and the rest half of the data are left as optimization sample data. Using the selected training sample data, and solving a central point parameter and a standardized constant (base width) parameter of the Gaussian function by using a clustering algorithm; and solving the connection weight coefficient between 5 hidden nodes and 2 output nodes by adopting a least square algorithm in an identification theory. Model parameter base obtained after trainingThe standard value is that 5 central points are respectively [ Z ] g11 ，Z g12 ，Z g13 ，Z g14 ]、[Z g21 ，Z g22 ，Z g23 ，Z g24 ]、[Z g31 ，Z g32 ，Z g33 ，Z g34 ]、[Z g41 ，Z g42 ，Z g43 ，Z g44 ]、[Z g51 ，Z g52 ，Z g53 ，Z g54 ]The normalization constants (base widths) of the 5 hidden nodes are [ phi ] respectively gZ1 ，φ gZ2 ，φ gZ3 ，φ gZ4 ，φ gZ5 ]The connection weight coefficients between 5 hidden nodes and 2 output nodes are respectively [ Y g11 ，Y g12 ，Y g13 ，Y g14 ，Y g15 ]、[Y g21 ，Y g22 ，Y g23 ，Y g24 ，Y g25 ]. Similarly, calculating the average value of the actual daily average active power output in the training sample data of the third embodiment, and taking the average value as the historical average output active power P of the digital twin simulation module model of the power plant 2 (ii) a Or, calculating an average value of actual daily average active power output in all historical data (including training sample data and optimization sample data) of the third embodiment, and taking the average value as historical average output active power P of the third embodiment of the digital twin simulation module of the power plant 2 。
In the second embodiment, a BP neural network is adopted to construct a power plant digital twin simulation module model, and input variables can be changed, or input variables can be added/deleted, or input variables can be changed and added/deleted. For example, the daily maximum climate temperature is changed to the daily average climate temperature, or the daily maximum climate temperature and the daily average climate temperature are included in the input at the same time; and simultaneously or independently increasing the construction life of the input variable photovoltaic power plant. Input variable of the second embodiment is designed to increase the building life S of the photovoltaic power plant 1 And the average day climate temperature, i.e. the input variable is the month M of the day 1 Daily maximum climate temperature T 1 Average day climate temperature T Q1 Day weather State A 1 Photovoltaic power plant construction life S 1 The output variable is daily average active power output P G1 Average daily workRate output Q G1 Selecting a three-layer BP network with hidden layer node number of 6, wherein 30 weight coefficients of 5 multiplied by 6 between the input layer and the hidden layer nodes are set as V H1 To V H30 The threshold value of 6 hidden layer nodes is theta H1 To theta H6 (ii) a The total 12 weight coefficients from the hidden node to the output node are V in the range of 6 multiplied by 2 H31 To V H42 And the threshold value of 2 output nodes is theta H7 To theta H8 . At the moment, the power plant digital twin simulation module models have 50 model parameters in total.
In the third embodiment, the RBF neural network is adopted to construct the photovoltaic power plant digital twin simulation module model, and input variables can be changed, or input variables are added/deleted, or input variables are changed and added/deleted. For example, the day average climate temperature is changed to the day maximum climate temperature; the input variable photovoltaic plant construction age is deleted simultaneously or individually. The three-input variable of the embodiment is changed into the month M of the day 2 Day maximum climate temperature T 2 Day weather State A 2 Photovoltaic power plant construction life S 2 (ii) a The output variable is daily average active power output P G2 Sunday average reactive power output Q G2 .6 hidden nodes with function functions being Gaussian functions are adopted, and the 6 hidden nodes and the input M are 2 、T 2 、A 2 、S 2 The corresponding central points are respectively [ B 11 ，B 12 ，B 13 ，B 14 ]、[B 21 ，B 22 ，B 23 ，B 24 ]、[B 31 ，B 32 ，B 33 ，B 34 ]、[B 41 ，B 42 ，B 43 ，B 44 ]、[B 51 ，B 52 ，B 53 ，B 54 ]、[B 61 ，B 62 ，B 63 ，B 64 ](ii) a The normalization constants (base widths) of the 6 hidden nodes are respectively | φ B1 ，φ B2 ，φ B3 ，φ B4 ，φ B5 ，φ B6 L, |;6 hidden nodes and 2 output nodes P G2 、Q G2 The connection weight coefficients between are [ YB 11 ，Y B12 ，Y B13 ，Y B14 ，Y B15 ，Y B16 ]、[Y B21 ，Y B22 ，Y B23 ，Y B24 ，Y B25 ，Y B26 ]. At the moment, the three embodiments of the plant digital twin simulation module model have 42 model parameters in common.
Example four
The present invention also provides an embodiment practiced according to a method of the embodiment: and (3) constructing a power plant digital twin simulation module model, specifically constructing a wind power plant digital twin simulation module model by adopting a BP neural network. The input variable of the wind power plant digital twin simulation module model is the daily average wind speed S F1 Daily maximum wind speed S F2 The season S F3 Wherein the daily average wind speed S F1 The average value of the real-time wind speed in one day can be an average value of a plurality of sampling values or an integral average value of continuous sampling; season S F3 The values of the data days in spring, summer, autumn and winter are 1, 2, 3 and 4 respectively. The output variable of the digital twin simulation module model of the wind power plant is daily average active power output P G3 Daily average reactive power output Q G3 . The four simulation module models of the embodiment are used for predicting the season S of the day F3 For the actual data of the estimated day to be predicted, the daily average wind speed S F1 Day maximum wind speed S F2 The predicted day data to be predicted according to weather forecast and the like is adopted.
In the fourth embodiment, a three-layer BP network with hidden nodes of 5 is selected, wherein 15 weight coefficients of 3 × 5 between the input layer and the hidden nodes are set as V F1 To V F15 The threshold value of 5 hidden layer nodes is theta F1 To theta F5 (ii) a The total 10 weight coefficients from the hidden node to the output node are 5 multiplied by 2 and are V F16 To V F25 And the threshold value of 2 output nodes is theta F6 、θ F7 . At the moment, the four embodiments of the digital twin simulation module model of the power plant have 32 model parameters which are weight coefficients V F1 To V F25 Threshold value theta F1 To theta F7 . The excitation functions of the output layer and the hidden layer adopt unipolar Sigmoid functions.
The historical data of the fourth embodiment of the digital twin simulation module model of the power plant is at least 2 years oldMean wind speed S F1 Day maximum wind speed S F2 The season S F3 And data, and corresponding daily average active power output and daily average reactive power output actual value data. Some of them (between 30% and 70%) are randomly selected, for example, half of the data are randomly selected as training sample data, and the rest half of the data are left as optimization sample data. Training a neural network by using selected training sample data and adopting a gradient descent method or optimization algorithms such as particle swarm optimization, genetic algorithm and the like, wherein the training target is that the average error of the system is minimum (the sum of squares of the average errors is minimum); setting weight coefficient V of neural network after training F1 To V F25 Respectively has a value of V gF1 To V gF25 Threshold θ of neural network after training is finished F1 To theta F7 Are respectively theta gF1 To theta gF7 Then V is gF1 To V gF25 And theta gF1 To theta gF7 The model parameter benchmark value is the model parameter benchmark value of the fourth embodiment of the digital twin simulation module model of the power plant. In addition, the average value of the actual daily average active power output in the training sample data of the fourth embodiment is calculated and is used as the historical average output active power P of the digital twin simulation module model of the power plant 3 (ii) a Or, calculating an average value of actual daily average active power output in all historical data (including training sample data and optimization sample data) in the fourth embodiment, and taking the average value as historical average output active power P of the fourth embodiment of the digital twin simulation module of the power plant 3 。
EXAMPLE five
A RBF neural network is adopted to construct a wind power plant digital twin simulation module model, and the input variable is daily average wind speed S F1 Daily maximum wind speed S F2 Month of day M 4 The output variable is daily average active power output P G4 Daily average reactive power output Q G4 . Example five simulation Module model for predicting the month M of the day 4 For the actual data of the estimated day to be predicted, the daily average wind speed S F1 Daily maximum wind speed S F2 The predicted day data to be predicted according to weather forecast and the like is adopted.
In the fifth embodiment, the RBF neural network uses 6 hidden nodes whose function functions are gaussian functions, and the 6 hidden nodes and the input S F1 、S F2 、M 4 The corresponding central points are respectively [ R ] 11 ，R 12 ，R 13 ]、[R 21 ，R 22 ，R 23 ]、[R 31 ，R 32 ，R 33 ]、[R 41 ，R 42 ，R 43 ]、[R 51 ，R 52 ，R 53 ]、[R 61 ，R 62 ，R 63 ](ii) a The normalization constants (base widths) of the 6 hidden nodes are [ phi ] R1 ，φ R2 ，φ R3 ，φ R4 ，φ R5 ，φ R6 ].6 hidden nodes and 2 output nodes P G4 、Q G4 、P M4 Are respectively [ YR ] as the connection weight coefficient between 11 ，YR 12 ，YR 13 ，YR 14 ，YR 15 ，YR 16 ]、[YR 21 ，YR 22 ，YR 23 ，YR 24 ，YR 25 ，YR 26 ]. At the moment, the five embodiments of the power plant digital twin simulation module model have 36 model parameters in common.
Example five historical data is daily average wind speed S for at least 2 years F1 Daily maximum wind speed S F2 The month of day M 4 Data, and corresponding daily average active power output and daily average reactive power output. Some of them (between 30% and 70%) are randomly selected, for example, half of the data are randomly selected as training sample data, and the rest half of the data are left as optimization sample data. Using the selected training sample data, and solving parameters of 6 central points and a standardized constant (basic width) of a Gaussian function by using a clustering algorithm; and solving the connection weight coefficients between the 6 hidden nodes and the 2 output nodes by adopting a least square algorithm in an identification theory. The model parameter reference value obtained after training is that 6 central points are respectively R g11 ，R g12 ，R g13 ]、[R g21 ，R g22 ，R g23 ]、[R g31 ，R g32 ，R g33 ]、[R g41 ，R g42 ，R g43 ]、[R g51 ，R g52 ，R g53 ]、[R g61 ，R g62 ，R g63 ]The normalization constants (base widths) of the 6 hidden layer nodes are [ phi ] gR1 ，φ gR2 ，φ gR3 ，φ gR4 ，φ gR5 ，φ gR6 ]The connection weight coefficients between 6 hidden nodes and 2 output nodes are [ Y ] respectively gR11 ，Y gR12 ，Y gR13 ，Y gR14 ，Y gR15 ，Y gR16 ]、[Y gR21 ，Y gR22 ，Y gR23 ，Y gR24 ，Y gR25 ，Y gR26 ]. Similarly, the average value of the actual daily average active power output in the training sample data of the fifth embodiment is calculated and is used as the historical average output active power P of the digital twin simulation module model of the power plant 4 (ii) a Or calculating the average value of the actual daily average active power output in all the historical data of the fifth embodiment, and taking the average value as the historical average output active power P of the fifth embodiment of the digital twin simulation module model of the power plant 4 。
And (4) conclusion:
the output variables of examples two through five all included daily average active power output and daily average reactive power output. Output variables can be added/deleted simultaneously while the daily average active power output is maintained, for example, daily maximum active power output is all added, or daily average reactive power output is deleted simultaneously, etc.
The method for constructing the digital twin simulation module models of the photovoltaic power plant and the wind power plant by respectively adopting the BP neural network and the RBF neural network can also be used for modeling the digital twin simulation module models of the hydraulic power plant and the thermal power plant. The digital twin simulation module models of the photovoltaic power plant, the wind power plant, the hydraulic power plant, the thermal power plant and the like can also adopt other neural networks, for example, a DRNN diagonal recurrent neural network, a CMAC cerebellar neural network model and the like are adopted to model the digital twin simulation module models of the photovoltaic power plant and the wind power plant.
EXAMPLE six
The invention also provides a model parameter embodiment of the digital twin simulation module model of each power plant in the integrated optimization virtual power plant distributed resources.
Model parameters of digital twin simulation module models of each power plant in the distributed resources are integrated and optimized, specifically, 3 distributed power plants in the second embodiment, the third embodiment and the fourth embodiment are provided for the virtual power plant, and k is equal to 3.
And (1) integrating input variables of all power plant digital twin simulation module models in distributed resources to form an input vector set. Example two the input variable is the month M of the day 1 Daily maximum climate temperature T 1 Day weather condition A 1 A total of 3 input variables, i.e. m1 equals 3; example three the input variable is the month M of the day 2 Average day climate temperature T Q2 Day weather State A 2 Photovoltaic power plant construction life S 2 A total of 4 input variables (m 2 equals 4); the input variable for the fourth example is the daily average wind speed S F1 Daily maximum wind speed S F2 The season S F3 There are 3 input variables (m 3 equals 3). The set of input vectors formed by the input variables of all the power plant digital twin simulation module models in the distributed resources is set as [ M 1 ，T 1 ，A 1 ，M 2 ，T Q2 ，A 2 ，S 2 ，S F1 ，S F2 ，S F3 ]。
And (2) integrating output variables in the output variables of all the power plant digital twin simulation module models to form an output vector set. The output variable of the second embodiment is daily average active power output P G1 Sunday average reactive power output Q G1 The output variable of the third embodiment is the daily average active power output P G2 Sum-day average reactive power output Q G2 The output variable of the fourth embodiment is daily average active power output P G3 Sum-day average reactive power output Q G3 Therefore, the output variables of each power plant digital twin simulation module model are 2, i is equal to 2; an output vector set formed by integrating output variables of all power plant digital twin simulation module models is P G1 ，Q G1 ，P G2 ，Q G2 ，P G3 ，Q G3 ]。
And (3) integrating the model parameters of all the power plant digital twin simulation module models to form a model parameter set. In the second embodiment, the 26 model parameters (n 1 equals 26) are weight coefficients V G1 To V G20 Threshold value theta G1 To theta G6 (ii) a The 35 model parameters (n 2 equals 35) of the third embodiment are respectively the central points [ Z ] 11 ，Z 12 ，Z 13 ，Z 14 ]、[Z 21 ，Z 22 ，Z 23 ，Z 24 ]、[Z 31 ，Z 32 ，Z 33 ，Z 34 ]、[Z 41 ，Z 42 ，Z 43 ，Z 44 ]、[Z 51 ，Z 52 ，Z 53 ，Z 54 ]Normalized constant (base width) [ phi ] Z1 ，φ Z2 ，φ Z3 ，φ z4 ，φ z5 ]Connecting the weight coefficients [ Y 11 ，Y 12 ，Y 13 ，Y 14 ，Y 15 ]、[Y 21 ，Y 22 ，Y 23 ，Y 24 ，Y 25 ](ii) a The 32 model parameters (n 3 equals 32) of the fourth embodiment are the weight coefficients V F1 To V F25 Threshold value θ F1 To theta F7° The model parameter set formed by integrating the model parameters of all the power plant digital twin simulation module models is
[V G1 ,…,V G20 ,θ G1 ,…,θ G6 ,Z 11 ,…Z 54 ,φ Z1 ,…,φ z5 ,Y 11 ,…,Y 25 ,V F1 ,…,V F25 ,θ F1 8230and; the number of model parameters to be optimized is 93.
And step four, integrating the optimized sample data of all the power plant digital twin simulation module models in the virtual power plant distributed resources to form an optimized sample data set. In the sixth embodiment, the optimized sample data in the second to fourth embodiments are collected to form an optimized sample data set. The total amount of historical data of the power plant digital twin simulation module model is not necessarily the same, but the number of data sets (namely the days of the daily data) for optimizing sample data is reserved
Should be the same. For example, if the historical data of 800 days, 1000 days and 1100 days are collected in the second, third and fourth embodiments respectively, 400 days (400 groups) of historical data can be reserved as the optimization sample data, and the remaining historical data of 400 days, 600 days and 700 days are the modeling training sample data of the digital twin simulation module model in the second, third and fourth embodiments respectively; or historical data of 450 days are all reserved as optimization sample data, and the rest historical data of 350 days, 550 days and 650 days are respectively the modeling training sample data of the digital twin simulation module model of the second embodiment, the third embodiment and the fourth embodiment.
Step fifthly, determining an optimized objective function. If 400 days of historical data are selected as optimization sample data in the second, third and fourth embodiments, that is, 400 groups of optimization sample data are total, and w is equal to 400; example six the optimization objective function is
Wherein the weighted total output E is
Or is written equivalently
Namely that
E＝400μ 1 (δ 1 ·P W1 +δ 2 ·P W2 +δ 3 ·P W3 )+400μ 2 (δ 1 ·Q W1 +δ 2 ′Q W2 +δ 3 ·Q W3 ) (4):
Wherein, the first and the second end of the pipe are connected with each other,
is P W1 Average value of actual daily average active power output in 400 sets of optimized sample data of example 1;
is Q W1 The average value of the actual daily average reactive power output in the 400 sets of optimized sample data of embodiment 1;
is P W2 Optimizing the average value of the actual daily average active power output in the sample data for 400 groups of the third embodiment;
is Q W2 Optimizing the average value of the actual daily average reactive power output in the sample data for 400 groups of the third embodiment;
is P W3 Average value of actual daily average active power output in 400 sets of optimized sample data of example 3;
is Q W3 The average of the actual daily average reactive power output in the sample data was optimized for the 400 sets of example 3. r is equal to 1, 2 and 3 respectively represent the power plant digital twin simulation module models of the second, third and fourth embodiments, in the formulas (1), (2), (3) and (4),
the output variables of the second, third and fourth embodiments are daily average active power output and daily average reactive power output, and the daily average active power output is considered to be larger in resource allocation and scheduling of the virtual power plantIn the output variable weight coefficient, the output variable weight coefficient corresponding to daily average active power output is not lower than 0.6; in the sixth embodiment, the output variable weight coefficient μ corresponding to the daily average active power output is taken 1 0.7, output variable weight coefficient mu corresponding to daily average reactive power output 2 Is 0.3.
Sixthly, determining the value range of each model parameter in the model parameter set. The value rule of each model parameter in the model parameter set is that the value range of a certain model parameter is between 1 and lambda times of the lambda of the model parameter reference value. In the sixth embodiment, if λ is equal to 5, the value range of each model parameter in the model parameter set is 0.2 times (1 in 5) to 5 times of the reference value of the model parameter; for example, V G1 Is [0.2V ] gG1 5V gG1 ]，V G2 Is [0.2V ] gG2 5V gG2 ]，θ G1 Is [0.2 theta ] gG1 5θ gG1 ]，Z 11 Is [0.2Z ] g11 5Z g11 ]，θ F7 Is [0.2 theta ] gF7 5θ gF7 ](ii) a And so on.
And step-and-step, performing iterative optimization calculation on the parameters in the model parameter set by adopting an optimization algorithm. And the sixth embodiment adopts a particle swarm optimization algorithm to carry out optimization, wherein the end condition of the optimization is that the sum Q of the weighted relative errors of the models is smaller than a set error threshold epsilon, or the number of iterations for which the optimization iteration reaches the set number is optimized. Epsilon ranges from 0.01 to 0.1, in the sixth embodiment, epsilon ranges from 0.05, and the set number of iterations is 1500. The optimization algorithm may also employ a genetic algorithm, an ant colony algorithm, a wolf colony algorithm, or the like.
And after the optimization calculation is finished, taking the optimal solution of the model parameter set as a new model parameter reference value.
And step one, simulating and calculating and predicting output by the aid of the digital twin simulation module models of each power plant of the virtual power plant according to the new model parameter reference value. The output prediction refers to calculating an output value by adopting a power plant digital twin simulation module model according to input data of a calculated and predicted power plant digital twin simulation module model (namely, according to weather forecast prediction)And performing power dispatching operation by using the output value as a predicted value. For example, according to the estimated month M of the day of the second day 1 Day maximum climate temperature T 1 Day weather State A 1 And (3) calculating the daily average active power output and the daily average reactive power output of the power plant on the next day by adopting the digital twin simulation module model of the power plant in the embodiment 1.
Continuously collecting sample data of input variables and output variables of each power plant digital twin simulation module model while carrying out simulation calculation and output prediction by adopting each power plant digital twin simulation module model; when the data volume of the continuously acquired sample data reaches or exceeds 20% of the data volume of the optimized sample data set, replacing the data in the optimized sample data set by the newly acquired sample data by adopting a rolling method, namely replacing the oldest optimized sample data with the same number of groups in the optimized sample data set by using a plurality of groups of latest sample data; for example, the optimization sample data set of the sixth embodiment has 400 groups (400 days) of sample data, and when the sample data volume of input variables and output variables of each power plant digital twin simulation module model is continuously acquired and reaches or exceeds 80 groups (80 days), the oldest 80 groups (or more than 80 groups) of optimization sample data in the optimization sample data set are replaced by the newest 80 groups (or more than 80 groups) of sample data. And sixthly, continuously integrating and optimizing the model parameters of the power plant digital twin simulation module in each distributed resource.
The digital twin is a technical means of creating a virtual entity of a physical entity in a digital mode, and simulating, verifying, predicting and controlling the whole life cycle process of the physical entity by means of historical data, real-time data, algorithm models and the like.
The virtual power plant adopts a hardware system architecture of autonomous dispersion, cloud-edge cooperation and virtual-real fusion, edge computing service and controlled distributed resources are regarded as an atomic node, and a gateway carries an autonomous operation process to locally control the controlled distributed resources, so that the atomic node has controllability; the gateways realize dynamic construction of logical relationship through subscription/publication of messages, so that the atomic nodes have autonomous coordination. The virtual entity or the virtual gateway can be used as an equivalent individual to participate in communication with the physical entity or the gateway, so that the virtual-real integration is realized on a coordination mechanism.
The virtual power plant adopts a connection-aggregation-application three-layer software system structure, the connection layer is mainly realized by edge computing service carried by a gateway, intelligent perception of distributed resources is realized by a plurality of intelligent modules, and simulation and interaction of the distributed resources are realized by a digital twin module. And the aggregation layer builds an autonomous distributed architecture through the middleware server, and provides configuration and forwarding of messages for information interaction of the edge computing nodes. The application layer is combined with various application scenes of the virtual power plant, business logic and guiding rules in the response process are customized, and self-optimization-seeking operation of mass distributed resources of the virtual power plant is achieved.
The power plants in the distributed resources of the virtual power plants are used for producing and outputting electric energy, the capacity of the electric energy produced by each distributed power plant and output is obtained and accurately predicted through various ways, and the premise that the virtual power plants carry out aggregation scheduling and intelligent power distribution is provided. On the basis of the existing hardware system architecture and software system structure of the virtual power plant, the distributed power plant digital twin simulation module model for power output simulation and prediction has high quality, and is the key for the virtual power plant to accurately perform relevant verification, prediction and control.
The digital twin simulation module model of the photovoltaic power plant and the wind power plant adopts input variables such as the month of the day, the highest weather temperature of the day, the average weather temperature of the day, the weather state of the day, the construction age of the photovoltaic power plant and the like, the digital twin simulation module model of the wind power plant adopts input variables such as the average wind speed of the day, the maximum wind speed of the day, the month (season) of the day and the like, historical data of relevant input variables are easy to collect, accurate prediction of the last day (the current day and the following days) can be performed, accurate prediction of power plant output by using the digital twin simulation module model after the model is built is facilitated, and optimal scheduling of power and optimal distribution of power resources are facilitated on the basis of accurate prediction of power plant output. On the basis that a power plant in a virtual power plant is independently modeled by using training data, the overall parameters of all distributed power plant digital twin simulation module models are uniformly optimized by using the integrated optimization sample data, new historical data is collected and accumulated in the process of simulation, verification and prediction by using the power plant model, part of old optimization sample data is replaced in a rolling mode, the overall parameters of all distributed power plant digital twin simulation module models are continuously and intermittently optimized, the distributed power plant individuals are ensured to be always in a satisfactory running state, and the individual trend is ensured to be compatible with the overall trend of the virtual power plant.
The above description is only for the specific embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive various changes or substitutions within the technical scope of the present invention, and these should be covered by the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the appended claims.
Claims (10)
1. A digital twin construction method of distributed resources of a virtual power plant is characterized by comprising the following steps:
s1, constructing a digital twin body simulation module model of each power plant in distributed resources;
s2, collecting input variables of the simulation module model to form an input vector set;
s3, collecting output variables of the simulation module model to form an output vector set;
s4, collecting model parameters of the simulation module model to form a model parameter set;
s5, collecting data to form a sample data set, and determining an optimization objective function;
s6, determining the value range of the model parameters, and performing parameter optimization to obtain optimized parameter values;
and S7, performing simulation calculation and output prediction by using the module model, and updating the optimized sample data.
2. The digital twin construction method of distributed resources of a virtual power plant according to claim 1, characterized in that: in the step S1, when the digital twin simulation module model is constructed, the method includes the following steps:
s11, selecting one power plant in distributed resources of the virtual power plant, selecting the type of a digital twin simulation module model of the power plant, determining input variables and output variables of the digital twin simulation module model of the power plant and collecting historical data of the input variables and the output variables;
s12, dividing the collected historical data into training sample data and optimization sample data of the power plant digital twin simulation module model;
s13, training the power plant digital twin simulation module model by using training sample data of the power plant digital twin simulation module model;
s14, obtaining a model parameter reference value of the power plant digital twin simulation module model after training is finished;
and S15, repeating the steps S11-S14 until the construction of the digital twin simulation module models of all the power plants in the distributed resources of the virtual power plant is completed.
3. The digital twin construction method of distributed resources of a virtual power plant according to claim 1, characterized in that: in the step S2, the input vector set is [ D ] 11 ,D 12 ,…,D 1m1 ,…,D 2m2 ,D 31 ,…,D k1 ,D k2 ,…,D kmk ]；
Wherein k is the number of the digital twin simulation module models of the power plant, [ D ] 11 ,D 12 ,…,D 1m1 ]For m1 input variables, [ D ] of the 1 st power plant digital twin simulation module model k1 ,D k2 ,…,D kmk ]Mk input variables of the kth power plant digital twin simulation module model; d 2m2 The number of the m2 th input variables of the 2 nd power plant digital twin simulation module model; d 31 The number of the 1 st input variables of the 3 rd power plant digital twin simulation module model; m1, m2, \ 8230, mk are respectively 1 st, 2 nd, \8230, and the number of input variables of the kth power plant digital twin simulation module model.
4. The digital twin construction method of the distributed resources of the virtual power plant according to claim 1, characterized in that: the steps areIn step S3, the set of output vectors is [ C ] 11 ,C 12 ,…,C 1i ,…,C 2i ,C 31 ,…,C k1 ,C k2 ,…,C ki ]；
Wherein [ C 11 ,C 12 ,…,C 1i ]For the i output variables of the 1 st power plant digital twin simulation module model, C 2i For the ith output variable, C, of the 2 nd power plant digital twin simulation module model 31 For the 1 st output variable of the 3 rd plant digital twin simulation module model, [ C [ ] k1 ,C k2 ,…,C ki ]I output variables of the digital twin simulation module model of the kth power plant.
5. The digital twin construction method of the distributed resources of the virtual power plant according to claim 1, characterized in that: in step S4, the model parameter set is [ F 11 ,F 12 ,…,F 1n1 ,…,F 2n2 ,F 31 ,…,F k1 ,F k2 ,…,F knk ]；
Wherein, [ F ] 11 ,F 12 ,…,F 1n1 ]N1 model parameters, F, for the 1 st power plant digital twin simulation module model 2n2 For the n2 model parameter, F, of the 2 nd power plant digital twin simulation module model 31 For the 1 st model parameter, [ F ] of the 3 rd plant digital twin simulation module model k1 ,F k2 ,…,F knk ]Nk model parameters of the kth power plant digital twin simulation module model; n1, n2, \ 8230, nk are respectively 1 st, 2 nd, \8230, and the number of model parameters of the kth power plant digital twin simulation module model.
6. The digital twin construction method of distributed resources of a virtual power plant according to claim 1, characterized in that: in step S5, the optimization objective is the minimization of the absolute value of the model weighted relative error and Q, and the optimization objective function is:
wherein E is a weighted total output having
Or any of the equivalent writes:
wherein, p is the optimization sample data of the 1 st to w th groups in the optimization sample data set; r stands for the 1 st to k th power plant digital twin simulation module model, δ r The power plant output weight coefficient corresponding to the r power plant; q is the 1 st to the ith output of the power plant digital twin simulation module model; e prq The difference between the qth output of the model of the digital twin simulation module for the mth power plant of the optimized sample data for the pth group and the corresponding actual output value of the optimized sample data (i.e., the qth actual output value of the mth power plant of the optimized sample data for the pth group), O prq Optimizing the qth actual output value of the qth power plant of the sample data for the pth group;
optimizing the mean value of the qth actual output value of the r power plant in the sample data for the w groups; mu.s q Is the output variable weight coefficient.
For the power plant output weight coefficient delta r And an output variable weight coefficient mu q Respectively have
P r For the historical average output active power of the r-th distributed power plant, a power plant output weight coefficient delta r The method comprises the following steps:
7. the digital twin construction method of distributed resources of a virtual power plant according to claim 1, characterized in that: in the step S6, the value range of each model parameter in the model parameter set is determined; the value range of the model parameters in the model parameter set is between 1 time and lambda time of the lambda fraction of the model parameter reference value; λ is 2 or more and 10 or less;
and after the value range of each model parameter in the model parameter set is determined, the method further comprises the following steps:
performing iterative optimization calculation on the model parameters in the model parameter set by adopting an optimization algorithm; and after the optimization calculation is finished, taking the optimal solution of the model parameter set as a new model parameter reference value.
8. The digital twin construction method of distributed resources of a virtual power plant according to claim 1, characterized in that: in the step S7, when the digital twin simulation module models of each power plant are used for simulation calculation and output prediction:
continuously collecting sample data of input variables and output variables of the digital twin simulation module model of each power plant;
when the data volume of the continuously acquired sample data reaches or exceeds 20% of the data volume of the optimized sample data set, replacing the data in the optimized sample data set by the newly acquired sample data by adopting a rolling method;
and repeating the steps S6-S7 to continuously integrate and optimize the model parameters of the digital twin simulation module of the power plant in each distributed resource.
9. A digital twin construction system of a distributed resource of a virtual power plant according to any of claims 1-8, characterized by comprising:
the training module is used for constructing a digital twin body simulation module model of each power plant in the distributed resources;
the acquisition module is used for acquiring an input vector set formed by input variables of the simulation module model, an output vector set formed by output variables of the simulation module model and a model parameter set formed by model parameters of the simulation module model;
the target optimization module is used for collecting data to form a sample data set, determining an optimization objective function, and determining a value range of a model parameter to perform parameter optimization to obtain an optimization parameter value;
and the prediction module is used for performing simulation calculation and output prediction by using the module model and updating the optimized sample data.
10. A computer device, characterized in that the computer device comprises a processor, a memory coupled to the processor, in which memory program instructions are stored which, when executed by the processor, cause the processor to carry out the steps of the method of digital twinning construction of virtual power plant distributed resources according to any of the claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211444526.XA CN115758884A (en) | 2022-11-18 | 2022-11-18 | Digital twin construction method for distributed resources of virtual power plant |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211444526.XA CN115758884A (en) | 2022-11-18 | 2022-11-18 | Digital twin construction method for distributed resources of virtual power plant |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115758884A true CN115758884A (en) | 2023-03-07 |
Family
ID=85373197
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211444526.XA Pending CN115758884A (en) | 2022-11-18 | 2022-11-18 | Digital twin construction method for distributed resources of virtual power plant |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115758884A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116702523A (en) * | 2023-08-08 | 2023-09-05 | 北京中电普华信息技术有限公司 | Simulation method for power resource regulation, electronic equipment and computer medium |
| CN117236152A (en) * | 2023-11-10 | 2023-12-15 | 国网浙江省电力有限公司宁波供电公司 | Twin simulation method and system for new energy power grid |
-
2022
- 2022-11-18 CN CN202211444526.XA patent/CN115758884A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116702523A (en) * | 2023-08-08 | 2023-09-05 | 北京中电普华信息技术有限公司 | Simulation method for power resource regulation, electronic equipment and computer medium |
| CN116702523B (en) * | 2023-08-08 | 2023-10-27 | 北京中电普华信息技术有限公司 | Simulation method for power resource regulation, electronic equipment and computer medium |
| CN117236152A (en) * | 2023-11-10 | 2023-12-15 | 国网浙江省电力有限公司宁波供电公司 | Twin simulation method and system for new energy power grid |
| CN117236152B (en) * | 2023-11-10 | 2024-04-09 | 国网浙江省电力有限公司宁波供电公司 | Twin simulation method and system for new energy power grid |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115758884A (en) | Digital twin construction method for distributed resources of virtual power plant | |
| CN109492815B (en) | Energy storage power station site selection and volume fixing optimization method for power grid under market mechanism | |
| CN112614009B (en) | Power grid energy management method and system based on deep expectation Q-learning | |
| CN110518580B (en) | Active power distribution network operation optimization method considering micro-grid active optimization | |
| CN106886603B (en) | Hierarchical distributed system architecture and method for demand response resource combination optimization | |
| CN111709109B (en) | Photovoltaic absorption capacity calculation method and device considering source-load time sequence correlation | |
| CN107067190A (en) | The micro-capacitance sensor power trade method learnt based on deeply | |
| CN107403256A (en) | One kind considers the probabilistic photovoltaic microgrid battery energy storage collocation method of demand response | |
| CN105826944A (en) | Method and system for predicting power of microgrid group | |
| CN111612244B (en) | QRA-LSTM-based method for predicting nonparametric probability of photovoltaic power before day | |
| CN108964124A (en) | Increment power distribution network is fallen into a trap and the grid-connected Optimal Configuration Method of the distributed generation resource of Respondence to the Price of Electric Power | |
| CN110137981B (en) | Distributed energy storage aggregator AGC method based on consistency algorithm | |
| CN109829599B (en) | Cluster division method and device for power distribution network based on high-proportion renewable energy | |
| CN112821432A (en) | Double-layer multi-position configuration method of energy storage system under wind and light access | |
| CN114202229B (en) | Determining method of energy management strategy of micro-grid based on deep reinforcement learning | |
| CN112101626A (en) | Distributed photovoltaic power generation power prediction method and system | |
| CN114971090A (en) | Electric heating load prediction method, system, equipment and medium | |
| CN111259315B (en) | Decentralized scheduling method of multi-subject coordinated pricing mode | |
| CN110555562A (en) | layered energy management method for comprehensive energy system in ubiquitous power Internet of things | |
| CN113644682B (en) | Multi-zone collaborative management and control method and device for high-permeability active power distribution network and terminal equipment | |
| CN108205110A (en) | The detection method and device of generator of wind generating set failure | |
| CN115498694A (en) | Multi-microgrid group planning method and device and electronic equipment | |
| CN110766224B (en) | Optimal configuration method and device for capacity of photovoltaic-heat storage device | |
| CN114091767A (en) | Method and system for robust random planning of data-driven comprehensive energy system | |
| CN111682539A (en) | Partitioning method and system for power grid with distributed power supply based on simplified comprehensive scene |
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 |