CN114493050A

CN114493050A - Multi-dimensional fusion new energy power parallel prediction method and device

Info

Publication number: CN114493050A
Application number: CN202210397812.9A
Authority: CN
Inventors: 李卓环; 李鹏; 马溪原; 陈元峰; 陈炎森; 程凯; 周悦; 包涛; 周长城; 张子昊
Original assignee: Southern Power Grid Digital Grid Research Institute Co Ltd
Current assignee: Southern Power Grid Digital Grid Research Institute Co Ltd
Priority date: 2022-04-08
Filing date: 2022-04-08
Publication date: 2022-05-13
Anticipated expiration: 2042-04-08
Also published as: CN114493050B

Abstract

The application relates to a multi-dimensional fusion new energy power parallel prediction method and device. The method comprises the following steps: acquiring a target power generation physical model and a target data knowledge model corresponding to the target new energy station; determining a target power generation physical model and a target data knowledge model, and respectively fusing according to a first fusion mode and a second fusion mode to obtain a first fusion model and a second fusion model; acquiring data to be processed corresponding to the target new energy station, wherein the data to be processed comprises station data, weather data and power generation historical operation data of the target new energy station; inputting the data to be processed into a first fusion model to obtain a first prediction result, and inputting the data to be processed into a second fusion model to obtain a second prediction result; and inputting the first prediction result and the second prediction result into a parallel output learning device for processing to obtain a power prediction result of the target new energy station. By adopting the method, the generated power prediction precision of the new energy station can be improved.

Description

Multi-dimensional fusion new energy power parallel prediction method and device

Technical Field

The present application relates to the field of new energy power generation technologies, and in particular, to a multi-dimensional fusion new energy power parallel prediction method, apparatus, computer device, storage medium, and computer program product.

Background

With the development of new power technologies, new energy power generation technologies have emerged. The new energy source comprises renewable energy sources generally, and comprises solar energy, biomass energy, wind energy, geothermal energy, wave energy, ocean current energy, tidal energy, hydrogen energy and the like. The new energy power generation is a process of converting the new energy into electric energy by using a power generation power device to realize power generation. At present, the main new energy power generation is to utilize a wind power generation station and a photovoltaic power generation station to generate power.

The new energy power generation brings huge impact and challenge to the safety and stability of a power grid, so that the operation risk of a power system is difficult to control. In order to improve the risk response capability of the power system, the generated power prediction accuracy of the new energy station adapting to different scenes, terrains, landforms and climatic features needs to be improved.

At present, a plurality of artificial intelligence-based methods are proposed in the field of power generation power prediction of new energy stations, but the methods train a machine learning model by acquiring massive sample data so as to excavate the incidence relation between input and output from the sample data. The generated power prediction precision of the new energy station obtained by the method mainly depends on the accuracy and the quantity of the sample data, and when the accuracy of the sample data is not high or the quantity of the sample data is small, the generated power prediction precision of the new energy station is difficult to guarantee, and the generated power of the new energy station under different scenes, terrains, landforms and climatic characteristics is difficult to predict.

Disclosure of Invention

In view of the above, it is necessary to provide a multi-dimensional fusion new energy power parallel prediction method, apparatus, computer device, computer readable storage medium and computer program product capable of improving the generated power prediction accuracy of a new energy station.

In a first aspect, the application provides a multi-dimensional fusion new energy power parallel prediction method. The method comprises the following steps:

acquiring a target power generation physical model and a target data knowledge model corresponding to the target new energy station;

determining a target power generation physical model and a target data knowledge model, and respectively fusing according to a first fusion mode and a second fusion mode to obtain a first fusion model and a second fusion model;

acquiring data to be processed corresponding to the target new energy station, wherein the data to be processed comprises station data, weather data and power generation historical operation data of the target new energy station;

inputting the data to be processed into a first fusion model to obtain a first prediction result, and inputting the data to be processed into a second fusion model to obtain a second prediction result;

and inputting the first prediction result and the second prediction result into a parallel output learning device for processing to obtain a power prediction result of the target new energy station.

In one embodiment, the step of constructing the target data knowledge model comprises: determining candidate new energy stations matched with the target new energy station, and acquiring candidate knowledge models of the candidate new energy stations; performing migration learning based on the candidate knowledge model of the candidate new energy station and the expert knowledge base to construct a target knowledge model corresponding to the target new energy station; constructing a target data model based on station data, weather data and power generation historical operation data of the target new energy station and each candidate new energy station; and fusing the target knowledge model and the target data model to obtain the target data knowledge model.

In one embodiment, determining candidate new energy sites that match the target new energy site includes: calculating similar distances between a target new energy station and other new energy stations according to station data of the new energy stations; and determining the new energy stations with similar distances meeting the close range condition as candidate new energy stations.

In one embodiment, in the case that the target new energy station belongs to a wind power station, the target power generation physical model comprises a fan model and a wind power plant model considering wake effect; the wind power generation system comprises a wind power generation system, a wind power generation system and a wind power generation system, wherein the wind power generation system comprises a wind power generation system and a wind power generation system, and the wind power generation system comprises a wind power generation system and a wind power generation system; under the condition that the target new energy station belongs to a photovoltaic power generation station, the target power generation physical model comprises a photovoltaic single machine model and a photovoltaic power station model; the photovoltaic single-machine model comprises at least one of a silicon crystal plate model, a photovoltaic plate model and a power irradiance model, and the photovoltaic power station model comprises at least one of a photovoltaic array model and a photovoltaic cluster model.

In one embodiment, the first fusion mode includes a learner fusion mode, and the inputting the data to be processed into the first fusion model to obtain the first prediction result includes: inputting data to be processed of the target new energy station into a target power generation physical model to obtain a physical prediction result; and inputting the data to be processed and the physical prediction result of the target new energy station into a target data knowledge model to obtain a first prediction result.

In one embodiment, the second fusion mode includes a weighted fusion mode, and the inputting of the data to be processed into the second fusion model to obtain the second prediction result includes: acquiring a correction fusion coefficient and an error vector corresponding to the target data knowledge model; inputting data to be processed into a target power generation physical model to obtain a physical prediction result; inputting data to be processed into a target data knowledge model to obtain a data knowledge prediction result; and performing weighted fusion processing on the physical prediction result and the data knowledge prediction result based on the corrected fusion coefficient and the error vector to obtain a second prediction result.

In one embodiment, the obtaining step of the modified fusion coefficient includes: acquiring training data corresponding to the target new energy station and a historical power label; inputting training data into a target power generation physical model to obtain a physical prediction result of an optimization stage; inputting training data into a target data knowledge model to obtain a data knowledge prediction result in an optimization stage; determining a weighted prediction result according to the corrected fusion variable, the physical prediction result of the optimization stage, the data knowledge prediction result of the optimization stage and the error vector corresponding to the target data knowledge model; and solving the correction fusion variable in an iterative calculation mode towards the direction of minimizing the difference between the historical power label and the weighted prediction result to obtain a correction fusion coefficient.

In a second aspect, the application further provides a multi-dimensional fusion new energy power parallel prediction device. The device comprises:

the acquisition module is used for acquiring a target power generation physical model and a target data knowledge model corresponding to the target new energy station;

the acquisition module is also used for determining a first fusion model and a second fusion model which are obtained by fusing the target power generation physical model and the target data knowledge model respectively according to a first fusion mode and a second fusion mode;

the acquisition module is further used for acquiring data to be processed corresponding to the target new energy station, wherein the data to be processed comprises station data, weather data and power generation historical operation data of the target new energy station;

the prediction module is used for inputting the data to be processed into the first fusion model to obtain a first prediction result and inputting the data to be processed into the second fusion model to obtain a second prediction result;

and the output module is used for inputting the first prediction result and the second prediction result into the parallel output learning device for processing to obtain a power prediction result of the target new energy station.

In a third aspect, the present application also provides a computer device. The computer device comprises a memory storing a computer program and a processor implementing the following steps when executing the computer program:

acquiring data to be processed corresponding to the target new energy station, wherein the data to be processed comprises station data, weather data and historical power generation operation data of the target new energy station;

In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of:

In a fifth aspect, the present application further provides a computer program product. The computer program product comprising a computer program which when executed by a processor performs the steps of:

The multi-dimension fused new energy power parallel prediction method, the device, the computer equipment, the storage medium and the computer program product, the target data knowledge model is obtained by fusing the target data model and the target knowledge model, then the target data knowledge model and the target power generation physical model are fused, so that the multidimensional fusion of data dimension, physical dimension and knowledge dimension is realized, further realizes the power prediction of the new energy station driven by data-physics-knowledge combination according to multi-dimensional fusion, the method has the advantages that massive knowledge can be extracted through the fusion target knowledge model, the generated power is predicted based on the fused target data knowledge model, and compared with the generated power prediction which is performed only by depending on a limited number of sample data, the generated power prediction accuracy of the target new energy station can be improved. Furthermore, the interpretability of the generated power prediction is expanded by fusing the target generated physical model, so that the predicted behaviors of the first fusion model and the second fusion model are easier to understand by users. In addition, the first prediction result and the second prediction result are processed and then output through the parallel output learner, and compared with the method of directly outputting the first prediction result and the second prediction result, the method has the advantages that: the first prediction result and the second prediction result are mutually corrected, prediction errors can be reduced, and therefore the generated power prediction accuracy of the target new energy station can be further improved.

Drawings

Fig. 1 is an application environment diagram of a new energy power parallel prediction method based on multi-dimensional fusion in one embodiment;

FIG. 2 is a schematic flow chart illustrating a multi-dimensional merged new energy power parallel prediction method according to an embodiment;

FIG. 3 is a flowchart illustrating the steps of constructing a knowledge model of target data in one embodiment;

FIG. 4 is a block diagram of the structure of a knowledge model of target data in one embodiment;

FIG. 5 is a schematic flow chart illustrating a method for multi-dimensional merged parallel prediction of new energy power in another embodiment;

FIG. 6 is a block diagram of an embodiment of a multidimensional-fusion new energy power parallel prediction device;

FIG. 7 is a diagram illustrating an internal structure of a computer device according to an embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

The multi-dimensional fusion new energy power parallel prediction method provided by the embodiment of the application can be applied to the application environment shown in fig. 1. Wherein the terminal 102 communicates with the server 104 via a network. The data storage system may store data that the server 104 needs to process. The data storage system may be integrated on the server 104, or may be located on the cloud or other network server. The terminal 102 may independently execute the multi-dimensional integrated new energy power parallel prediction method provided by the embodiment of the present application, and the terminal 102 and the server 104 may also cooperatively execute the multi-dimensional integrated new energy power parallel prediction method provided by the embodiment of the present application.

When the terminal 102 independently executes the multi-dimensional fusion new energy power parallel prediction method, the terminal 102 acquires a target power generation physical model and a target data knowledge model corresponding to a target new energy station; determining a target power generation physical model and a target data knowledge model, and respectively fusing according to a first fusion mode and a second fusion mode to obtain a first fusion model and a second fusion model; acquiring data to be processed corresponding to the target new energy station, wherein the data to be processed comprises station data, weather data and power generation historical operation data of the target new energy station; inputting the data to be processed into a first fusion model to obtain a first prediction result, and inputting the data to be processed into a second fusion model to obtain a second prediction result; and inputting the first prediction result and the second prediction result into a parallel output learning device for processing to obtain a power prediction result of the target new energy station.

When the terminal 102 and the server 104 cooperatively execute the multi-dimensional fusion new energy power parallel prediction method, the terminal 102 acquires a target power generation physical model and a target data knowledge model corresponding to a target new energy station; determining a target power generation physical model and a target data knowledge model, and respectively fusing according to a first fusion mode and a second fusion mode to obtain a first fusion model and a second fusion model; the method comprises the steps of obtaining data to be processed corresponding to a target new energy station, wherein the data to be processed comprises station data, weather data and power generation historical operation data of the target new energy station, and sending the data to be processed, a first fusion model and a second fusion model to a server 104. The server 104 inputs the data to be processed into the first fusion model to obtain a first prediction result, and inputs the data to be processed into the second fusion model to obtain a second prediction result; and inputting the first prediction result and the second prediction result into a parallel output learning device for processing to obtain a power prediction result of the target new energy station.

The terminal 102 may be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, internet of things devices and portable wearable devices, and the internet of things devices may be smart speakers, smart televisions, smart air conditioners, smart car-mounted devices, and the like. The portable wearable device can be a smart watch, a smart bracelet, a head-mounted device, and the like. The server 104 may be implemented as a stand-alone server or as a server cluster comprised of multiple servers.

In one embodiment, as shown in fig. 2, a multi-dimensional merged new energy power parallel prediction method is provided, and the method may be executed by a terminal or a server alone, or may be executed by the terminal and the server in cooperation. The embodiment of the present application is described by taking the application of the method to the terminal in fig. 1 as an example, and includes the following steps:

step 202, a target power generation physical model and a target data knowledge model corresponding to the target new energy station are obtained.

And the target new energy station is a new energy station for predicting the generated power. The new energy station is a wind power generation station or all equipment below a grid-connected point of a photovoltaic power generation station which is connected into a power system in a centralized manner, and comprises a transformer, a bus, a circuit, a converter, an energy storage unit, a wind turbine generator, a photovoltaic power generation system, reactive power regulation equipment, auxiliary equipment and the like.

The target power generation physical model is used for predicting the power generation power of the target new energy station, reflects the physical definition of physical association relation existing among data in the target new energy station and the existing engineering model, covers objects including wind power generation stations, photovoltaic power generation stations and various devices in an electric power system, and covers the range including theories of physical laws, meteorological characteristics and the like. When the target new energy station belongs to a wind power generation station, the target power generation physical model is a power generation physical model corresponding to the wind power generation station and comprises at least one of a fan model and a wind power plant model considering wake effect; when the target new energy station belongs to the photovoltaic power generation station, the target power generation physical model is a power generation physical model corresponding to the photovoltaic power generation station and comprises at least one of a photovoltaic single-machine model and a photovoltaic power station model.

The target data knowledge model can also be called a target data-knowledge model, is a model obtained based on fusion of data dimensions and knowledge dimensions, is a machine learning model obtained by fusing and model training the data model and the knowledge model corresponding to the target new energy station and is used for predicting the generating power of the target new energy station.

Specifically, the terminal obtains a target power generation physical model and a target data knowledge model corresponding to the target new energy station.

And 204, determining a first fusion model and a second fusion model obtained by fusing the target power generation physical model and the target data knowledge model respectively according to the first fusion mode and the second fusion mode.

The first fusion mode and the second fusion mode are two different fusion modes and are used for fusing the target power generation physical model and the target data knowledge model. Correspondingly, the first fusion model and the second fusion model are two different models, are both models obtained based on multi-dimensional fusion of data dimension, physical dimension and knowledge dimension, both comprise a data model, a power generation physical model and a knowledge model of the new energy station, and are all used for predicting the power generation power of the target new energy station.

Specifically, the terminal fuses a target power generation physical model and a target data knowledge model according to a first fusion mode to obtain a first fusion model; and fusing the target power generation physical model and the target data knowledge model according to a second fusion mode to obtain a second fusion model.

And step 206, acquiring data to be processed corresponding to the target new energy station, wherein the data to be processed comprises station data, weather data and power generation historical operation data of the target new energy station.

The data to be processed is a basis for predicting the generated power of the target new energy station and serves as input of the first fusion model and the second fusion model. Station data of the target new energy station are used for representing the attributes of the target new energy station, and the attributes can include terrain, landform, longitude and latitude and the like; the weather data of the target new energy station are used for representing the meteorological conditions of the target new energy station, and can comprise temperature, humidity, precipitation, air pressure, wind direction, wind speed and the like; the historical power generation operation data of the target new energy field station is historical power generation power values which are collected by the power system and stored in the database, and can include historical power generation power values of a centralized wind power plant and a centralized photovoltaic power station which are collected by the power grid dispatching automation system, and also can include historical power generation power values of a distributed wind power plant and a distributed photovoltaic power station which are collected by the power grid distribution automation system or the power grid metering automation system.

Specifically, the terminal acquires data to be processed corresponding to the target new energy station, wherein the data to be processed comprises station data, weather data and power generation historical operation data of the target new energy station.

And 208, inputting the data to be processed into the first fusion model to obtain a first prediction result, and inputting the data to be processed into the second fusion model to obtain a second prediction result.

And the first prediction result is a predicted power generation power value of the target new energy station, which is obtained by performing power prediction based on the first fusion model, and is used as the output of the first fusion model. The second prediction result is a predicted power generation power value of the target new energy station, which is obtained by performing power prediction based on the second fusion model, and is used as the output of the second fusion model.

Specifically, the terminal inputs the data to be processed into the first fusion model to obtain a first prediction result, and inputs the data to be processed into the second fusion model to obtain a second prediction result.

And step 210, inputting the first prediction result and the second prediction result into a parallel output learning device for processing to obtain a power prediction result of the target new energy station.

The parallel output learning device is a mathematical model for weighted summation or a machine learning model which is obtained through model training and used for predicting the generated power of the new energy station.

The power prediction result is a predicted power generation power value of the target new energy station.

Specifically, in an embodiment, the parallel output learner is a mathematical model for weighted summation, the terminal inputs the first prediction result and the second prediction result into the parallel output learner, and the parallel output learner performs weighted summation on the first prediction result and the second prediction result to obtain the power prediction result of the target new energy station, at this time, the power prediction result is a weighted sum of the first prediction result and the second prediction result, where the weights of the first prediction result and the second prediction result may be adjusted according to actual situations, which is not limited in this embodiment.

In another embodiment, the parallel output learner is a machine learning model obtained through model training and used for generating power prediction of the new energy station. The training mode of the parallel output learning device comprises the following steps: the method comprises the steps that a terminal obtains a sample set, wherein the sample set can be station data, weather data and power generation historical operation data of a plurality of new energy stations, training data and real power data are included in the sample set, the training data are input into a first fusion model to obtain a first sample prediction result, and the training data are input into a second fusion model to obtain a second sample prediction result; constructing a parallel output learning device to be trained, inputting the first sample prediction result and the second sample prediction result into the parallel output learning device to be trained for processing, and obtaining a sample power prediction result; and training the parallel output learner to be trained based on the difference between the sample power prediction result and the corresponding real power data until a training stopping condition is reached (for example, a loss function corresponding to the parallel output learner to be trained obtains a minimum value), and obtaining the trained parallel output learner. At this time, inputting the first prediction result and the second prediction result into a parallel output learning device which is trained to be processed, and obtaining a power prediction result of the target new energy station, wherein the power prediction result comprises: and the terminal inputs the first prediction result and the second prediction result into the parallel output learning device, and the parallel output learning device processes the first prediction result and the second prediction result to obtain a power prediction result of the target new energy station.

According to the multi-dimension-fused new energy power parallel prediction method, the data dimension, the physical dimension and the knowledge dimension are fused in a multi-dimension mode by fusing a target data model, a target knowledge model and a target physical model, and then the power prediction of the new energy station driven by data-physics-knowledge in a combined mode is achieved according to the multi-dimension fusion, wherein massive knowledge can be extracted through fusing the target knowledge model, the power generation power prediction is conducted based on the fused target data knowledge model, and compared with the power generation power prediction conducted only depending on a limited number of sample data, the purpose of improving the power generation power prediction accuracy of the target new energy station can be achieved. Furthermore, the interpretability of the generated power prediction is expanded by fusing the target generated physical model, so that the predicted behaviors of the first fusion model and the second fusion model are easier to understand by users. In addition, the first prediction result and the second prediction result are processed and then output through the parallel output learner, and compared with the method of directly outputting the first prediction result and the second prediction result, the method has the advantages that: the first prediction result and the second prediction result are mutually corrected, prediction errors can be reduced, and therefore the generated power prediction accuracy of the target new energy station can be further improved.

In one embodiment, as shown in FIG. 3, the construction step of the target data knowledge model comprises:

step 302, determining candidate new energy stations matched with the target new energy station, and acquiring candidate knowledge models of the candidate new energy stations.

The candidate new energy field stations are similar to or matched with the target new energy field station in the aspects of terrain, landform, longitude and latitude, meteorological conditions and the like, and the candidate new energy field stations comprise two or more than two. The candidate knowledge model is a knowledge model of the candidate new energy station, and comprises field knowledge experience related to station data, weather data, historical power generation operation data, a power generation power prediction model, historical prediction effect of the power generation power prediction model and the like of the candidate new energy station, and the field knowledge experience can be a function, a mathematical model or a machine learning model in form.

Specifically, the terminal determines candidate new energy stations matched with the target new energy station from the plurality of new energy stations and obtains candidate knowledge models of the candidate new energy stations.

And step 304, performing transfer learning based on the candidate knowledge model of the candidate new energy station and the expert knowledge base to construct a target knowledge model corresponding to the target new energy station.

The expert knowledge base is related to the new energy station and mainly comprises the following three layers: judgment rules based on massive expert experience, control rules for problem solving, and data for explaining the state, facts, and concepts of a problem, as well as current conditions and general knowledge, etc. The migration learning is a machine learning method, namely, a model developed for a task A is taken as an initial point and is reused in the process of developing the model for a task B. The target knowledge model is a knowledge model for power generation power prediction of the target new energy yard.

Specifically, the terminal integrates a candidate knowledge model of the candidate new energy station with an expert knowledge base through an integrated learning framework to obtain a target knowledge model corresponding to the target new energy station. For example, the ensemble learning framework may be bagging learning, the input of the ensemble learning framework is the candidate knowledge model of the candidate new energy station and knowledge in the expert knowledge base, and the output of the ensemble learning framework is the target knowledge model corresponding to the target new energy station, that is, the knowledge model obtained through migration learning.

And step 306, constructing a target data model based on the station data, the weather data and the power generation historical operation data of the target new energy station and each candidate new energy station.

The target data model is used for predicting the generated power of the target new energy station.

Specifically, the terminal stores station data, weather data and power generation historical operation data of the target new energy station and each candidate new energy station in a unified database in a unified data format, a data template and data granularity to obtain a target data model.

And 308, fusing the target knowledge model and the target data model to obtain a target data knowledge model.

As shown in fig. 4, the target data knowledge model is composed of a data module, an environment module, a knowledge module, and a learning module. Wherein the data module corresponds to a target data model for providing an input data setx(ii) a The environment module comprises statistical rules and input-output mapping rules, and the input data setxInput vector inxAfter inputting the environment module, returning the output vectorŷ ₀Output vectorŷ ₀Is an approximate data set obtained by mapping the environment moduleŷ ₀The input-output mapping rule depends on the unknown conditional probability; knowledge module provides knowledge setskFor formalizing and embedding knowledge experience in the domain into the learning process; the learning module is used for input vectorxProcessing and learning to return approximate quantitiesŷFor approximating the actual true output vectory。

Specifically, the terminal constructs a data knowledge model to be trained based on a target knowledge model, a target data model and an expected generalized risk formula (1); and training the data knowledge model to be trained until the loss function obtains the minimum value, and stopping training to obtain the trained target data knowledge model.

(1)

In equation (1):

learning parameters for the target data-knowledge model;

setting the risk to be generalized as a training objective function;F(x,y) Global learning space for characterizing problems for true but unknown joint probability distributions: (x,y) A probability distribution of (a);Las a loss function, the effect of which is to compute a set of pass-through inputsxApproximation of learning output

With actual true value datasetyThe difference between them;kis a set of knowledge that is to be gathered,k _iis the first in the knowledge setiAn element;

is directed to knowledgek _iThe hyper-parameter of (c);

e represents the E norm as a knowledge constraint operator;

is a constraint threshold;

is a knowledge ofk _iFor input data setxTo (1)iThe type of the constraint is that the constraint,iis a positive integer.

In the embodiment, the target knowledge model is obtained through migration learning by obtaining the candidate knowledge model of the candidate new energy field station matched with the target new energy field station, and the target knowledge model increases knowledge in the candidate knowledge model of the candidate new energy field station and knowledge in an expert knowledge base related to the new energy field station on the basis of the existing knowledge model of the target new energy field station, so that the problem of low prediction accuracy of the generated power of the new energy field station due to insufficient sample data is solved, and the purpose of improving the prediction accuracy of the generated power of the target new energy field station is achieved. In addition, a target data model is constructed through data of the target new energy station and data of all candidate new energy stations, and the target data model is not constructed through data of a plurality of new energy stations including the target new energy station and the candidate new energy stations, so that the purpose of reducing the calculation amount in the fusion process can be achieved, and the system cost is saved. In addition, the fusion of massive data and massive knowledge is realized by fusing the target knowledge model and the target data model, and the purpose of constructing the target data knowledge model driven by data-knowledge cooperation is achieved.

Wherein, the similarity distance is a weighted distance, and the calculation formula of the similarity distance is as follows:

(2)

in equation (2): weight ofw _n≥0，x _iAndx _jstation data of the target new energy station and other new energy stations respectively,pis a positive integer greater than 1.

The close-range condition is used to indicate that the similar distance between the target new energy source station and the other new energy source stations is small, for example, the close-range condition is that the difference value from the minimum similar distance does not exceed a distance threshold, or the close-range condition is that the difference value from the minimum similar distance is arranged in the first N bits, where N is a positive integer.

Specifically, the terminal acquires station data of a plurality of new energy stations including a target new energy station from a database; calculating similar distances between a target new energy station and other new energy stations according to station data of the new energy stations; and determining the new energy stations with similar distances meeting the close range condition as candidate new energy stations.

In this embodiment, by calculating the similar distances between the target new energy station and other new energy stations, the new energy station whose similar distance satisfies the short-distance condition is determined as a candidate new energy station, and the purpose of determining the candidate new energy station matched with the target new energy station can be achieved.

In one embodiment, in the case that the target new energy station belongs to a wind power station, the target power generation physical model comprises a wind turbine model and a wind power plant model considering wake effect; the wind power generation system comprises a wind power generation system, a wind power generation system and a wind power generation system, wherein the wind power generation system comprises a wind power generation system and a wind power generation system, and the wind power generation system comprises a wind power generation system and a wind power generation system; under the condition that the target new energy station belongs to a photovoltaic power generation station, the target power generation physical model comprises a photovoltaic single machine model and a photovoltaic power station model; the photovoltaic single-machine model comprises at least one of a silicon crystal plate model, a photovoltaic plate model and a power irradiance model, and the photovoltaic power station model comprises at least one of a photovoltaic array model and a photovoltaic cluster model.

In this embodiment, by determining corresponding target power generation physical models for two different types of target new energy stations for wind power generation and photovoltaic power generation, the generated power is predicted based on the target power generation physical models corresponding to the target new energy stations of a specific type, instead of predicting the generated power based on the target power generation physical models corresponding to the target new energy stations of all types, and the purpose of improving the generated power prediction efficiency of the target new energy stations can be achieved. Moreover, interpretability of power generation power prediction can be expanded by fusing a fan model comprising at least one of a circuit model, a magnetic circuit model, a mechanical model and a power wind speed model for fan power generation, a wind power plant model comprising at least one of a wake effect model and a wind power cluster model, a photovoltaic single machine model comprising at least one of a silicon crystal plate model, a photovoltaic plate model and a power irradiance model, and a photovoltaic power plant model comprising at least one of a photovoltaic array model and a photovoltaic cluster model, and theoretical and engineering models corresponding to various devices in a power generation plant station, a photovoltaic power generation plant station and a power system are based on, so that prediction behaviors of the first fusion model and the second fusion model can be more easily understood by users.

The model formula of the fusion mode of the learner is as follows:

（3）

in equation (3):kis a time tag;fandhmapping functions of input data and prediction results in the target power generation physical model and the target data knowledge model are respectively reflected;x _k+1is composed ofkA first prediction result of the target new energy station at the +1 moment;x ´ _k+1is composed ofkPredicting the target power generation physical model at the +1 moment to obtain a physical prediction result;ucarrying out random error vector in the prediction process for the target data knowledge model; x_kAnd Y_kIdentical, vectors consisting of input data, where X_kIs an input to a physical model of the target power generation, Y_kIs input to the target data knowledge model.

And the physical prediction result is a characteristic vector consisting of predicted power generation power values of the target new energy station, which is obtained by performing power prediction based on the target power generation physical model. The first prediction result is a characteristic vector formed by predicted power generation power values of the target new energy station, which is obtained by performing power prediction based on the first fusion model.

Specifically, the terminal forms a vector X by using the data to be processed of the target new energy station_kInputting the data into a target power generation physical model to obtain a physical prediction resultx´ _k+1(ii) a Vector Y formed by data to be processed of target new energy station_kAnd physical prediction resultsx´ _k+1Inputting a target data knowledge model to obtain a first prediction resultx _k+1。

In this embodiment, the purpose of obtaining the first prediction result can be achieved by predicting the generated power based on the first fusion model obtained by the learner fusion method.

The model formula of the weighted fusion mode is as follows:

(4)

in equation (4):ato modify the fusion coefficient, takea∈[0,1]。X_kAnd Y_kIdentical, are vectors of input data, where X_kIs an input to a physical model of the target power generation, Y_kIs input to the target data knowledge model.kIn the form of a time stamp,fandhmapping functions of input data and prediction results in the target power generation physical model and the target data knowledge model are respectively reflected;x _k+1is composed ofkA second prediction result of the target new energy station at the moment + 1;f(X_k) Is composed ofkPredicting a target power generation physical model at a moment to obtain a physical prediction result;h (Y_k) Is composed ofkPredicting a data knowledge prediction result obtained by predicting a target data knowledge model at a moment;uand carrying out random error vector in the prediction process for the target data knowledge model.

And the physical prediction result is a characteristic vector consisting of predicted power generation power values of the target new energy station, which is obtained by performing power prediction based on the target power generation physical model. The data knowledge prediction result is a characteristic vector formed by predicted power generation power values of the target new energy station, which is obtained by performing power prediction based on the target data knowledge model. The second prediction result is a characteristic vector formed by predicted power generation power values of the target new energy station, which is obtained by performing power prediction based on the second fusion model.

Specifically, the terminal obtains a correction fusion coefficient and an error vector corresponding to the target data knowledge model, wherein the correction fusion coefficient can be obtained in advance based on actual demand setting or training based on training data; vector X consisting of data to be processed of target new energy station_kInputting a target power generation physical model to obtain a physical prediction resultf(X_k) (ii) a Vector Y formed by data to be processed of target new energy station_kInputting a target data knowledge model to obtain a data knowledge prediction resulth (Y_k) (ii) a Based on modified fusion coefficientsaAnd error vectoruAnd performing weighted fusion processing on the physical prediction result and the data knowledge prediction result to obtain a second prediction result.

In this embodiment, the purpose of obtaining the second prediction result can be achieved by performing the generation power prediction based on the second fusion model obtained in the weighted fusion manner.

In one embodiment, the specific determining step of the modified fusion coefficient includes: acquiring training data corresponding to the target new energy station and a historical power label; inputting training data into a target power generation physical model to obtain a physical prediction result of an optimization stage; inputting training data into a target data knowledge model to obtain a data knowledge prediction result in an optimization stage; determining a weighted prediction result according to the corrected fusion variable, the physical prediction result of the optimization stage, the data knowledge prediction result of the optimization stage and the error vector corresponding to the target data knowledge model; and solving the correction fusion variable in an iterative calculation mode towards the direction of minimizing the difference between the historical power label and the weighted prediction result to obtain a correction fusion coefficient.

Specifically, the terminal may obtain the optimal modified fusion coefficient in advance through an iterative training mode. The terminal may obtain a set of training samples, which may include training data and historical power tags. The training data can be station data, weather data and power generation historical operation data of a plurality of new energy stations, or station data, weather data and power generation historical operation data of the target new energy station. The historical power tag is real power data, and can be obtained based on historical operation data of power generation. The terminal can input the training data into the target power generation physical model to obtain a physical prediction result of an optimization stage; and inputting the training data into the target data knowledge model to obtain a data knowledge prediction result in an optimization stage. And determining a weighted prediction result according to the corrected fusion variable, the physical prediction result of the optimization stage, the data knowledge prediction result of the optimization stage and the error vector corresponding to the target data knowledge model.

The terminal can calculate the difference between the historical power label and the weighted prediction result according to the following formula (5), and then solves the correction fusion variable in an iterative calculation mode towards the direction of minimizing the difference between the historical power label and the weighted prediction result to obtain a correction fusion coefficient.

Wherein the content of the first and second substances,

for historical power tags (i.e. true power information),athe variable is characterized, corrected and fused in the training stage, and the variable is solved after the training is finishedaIs a specific value. WhereinX _{k training}Is the training data input into the target power generation physical model, isY _{k training}Is the training data input to the target data knowledge model.kIn the form of a time stamp,fandhmapping functions of input data and prediction results in the target power generation physical model and the target data knowledge model are respectively reflected;f(X _{k training}) Is composed ofkPredicting a target power generation physical model at a moment to obtain a physical prediction result of an optimization stage;h(Y _{k training}) Is composed ofkPredicting a target data knowledge model at a moment to obtain a data knowledge prediction result of an optimization stage;uand carrying out random error vector in the prediction process for the target data knowledge model.

In one embodiment, the optimization method may adopt an optimization toolkit, and adopt a quadratic optimization method to perform optimization solution, where the solution variables are modified fusion variablesaThe iteration termination condition is that the difference value of the target values of the two iterations is not more than 1% of the previous iteration, or the preset iteration times are reached.

In the embodiment, the optimal correction fusion coefficient can be rapidly and accurately determined in an iterative optimization mode, so that the generated power prediction accuracy of the new energy station is improved.

In another embodiment, as shown in fig. 5, a multi-dimensional fused new energy power parallel prediction method is provided, including:

and step S1, constructing a target data knowledge model.

The main steps of the construction of the target data knowledge model are as follows:

1) object data model construction

In the data preprocessing, the data cleaning, the data mining, the data set construction and the feature selection are mainly included. The target data model includes site data, weather data, and power generation history operation data for a plurality of new energy sites of the target new energy site (i.e., sites for which new energy power predictions are to be made). The data source is real-time cloud transmission and updating, and related data are directly extracted from a cloud database. Data cleaning is completed at the cloud end, and data mining, data set construction and feature selection are completed at the local end.

2) Target knowledge model construction

The target knowledge model mainly comprises a massive expert knowledge base and a migration model obtained by migration learning.

The migration model is mainly obtained by migration in models and knowledge of other new energy stations by adopting a migration learning method. In this embodiment, the other new energy field stations refer to new energy field stations similar to the target new energy field station in the aspects of terrain, landform, longitude and latitude, meteorological conditions, and the like, and information, data, and models of historical operating conditions, operating data, new energy power prediction conditions, new energy power prediction effects, and the like of the other new energy field stations are integrated to form a mass data-knowledge base. The target new energy station can migrate knowledge in the expert knowledge base and the data-knowledge base in the modes of data matching, knowledge learning, model migration, model integration and the like, so that a migrated target knowledge model is obtained.

The method comprises the following steps: and calculating the similar distance between the target new energy station and other new energy stations according to the operation data and station data of the plurality of new energy stations in the database. Selecting 20 new energy field stations with the minimum similar distance, and collecting and fusing the data and knowledge of the 20 new energy field stations, such as historical operation data, configuration data, prediction models, historical prediction effects and the like. The input of the integrated framework is knowledge content and model of 20 new energy stations and knowledge in an expert knowledge base, and the output is an integrated unified fusion model, namely a target knowledge model.

3) Data-knowledge fusion

And taking the target data model and the target knowledge model as input to construct a target data-knowledge model. The target data-knowledge model of data-knowledge fusion is constructed as follows:

in the above formula:

learning parameters for a target data-knowledge modelAn amount;

With actual true value data setyThe difference between them;kin order to be a set of knowledge,k _iis the first in the knowledge setiAn element;

is directed to the knowledgek _iThe hyper-parameter of (c);

e represents the E norm as a knowledge constraint operator;

is a constraint threshold;

And the mass data and the mass knowledge are fused through data-knowledge fusion, so that data-knowledge cooperative driving is realized.

And inputting the data to be processed of the target new energy station into the obtained target data-knowledge model, and outputting the predicted power value of the target new energy station. If the prediction scale is ultra-short-term prediction, the predicted power value is 4 hours in the future, one predicted power value is obtained every 15 minutes, and the 16 predicted power values are used for representing the power generation power of the target new energy station 4 hours in the future. If the prediction scale is short-term prediction, the predicted power value is 3 days in the future, one predicted power value is obtained every 15 minutes, and 288 predicted power values are used for representing the generated power of the target new energy station for 3 days in the future. If the prediction scale is the medium-long term prediction, the predicted power value is 10 days in the future, one predicted power value is obtained every 15 minutes, and 960 predicted power values are used for representing the generated power of the target new energy station 10 days in the future.

And step S2, constructing a power prediction model of data-physics-knowledge fusion.

The method for constructing the power prediction model of data-physics-knowledge fusion comprises the following steps: establishing a target power generation physical model, combining the target power generation physical model with a target data-knowledge model and performing parallel prediction.

(1) Establishing a target power generation physical model

The new energy power generation physical model comprises a fan model, a photovoltaic single machine model, a wind power plant model considering wake effect and a photovoltaic power station model.

The fan model mainly comprises a circuit model, a magnetic circuit model, a mechanical model and a power-wind speed model for fan power generation. The photovoltaic single-machine model comprises a silicon crystal plate model, a photovoltaic plate model and a power-irradiance model.

The wind power plant model mainly comprises a wake effect model and a wind power cluster model.

The photovoltaic power station model mainly comprises a photovoltaic array model and a photovoltaic cluster model.

(2) Combining a target power generation physical model and a target data-knowledge model

In the aspect of fusion of a target power generation physical model and a target data-knowledge model, two modes are adopted for fusion, and finally the two modes are output in parallel. The two fusion modes are learner fusion and weighted fusion respectively.

The model formula of the fusion mode of the learner is as follows:

in the above formula:kis a time tag;fandhrespectively reflecting mapping functions of input data and a prediction result in a target power generation physical model and a target data-knowledge model;x _k+1is composed ofkA first prediction result of the target new energy station at the +1 moment;x´ _k+1is composed ofkPredicting the target power generation physical model at the +1 moment to obtain a physical prediction result;ucarrying out random error vector in the prediction process for the target data-knowledge model; x_kAnd Y_kIdentical, vectors consisting of input data, where X_kInput to the physical model of target power generation, Y_kInput to the target data-knowledge model.

The model formula of the weighted fusion mode is as follows:

in the above formula:ato modify the fusion coefficient, takea∈[0,1]. The modified fusion coefficient can be obtained in advance based on actual requirement setting, and can also be obtained based on training data training. X_kAnd Y_kIdentical, are vectors of input data, where X_kInput to the physical model of target power generation, Y_kInput to the target data-knowledge model.kIn the form of a time stamp,fandhrespectively reflecting mapping functions of input data and a prediction result in a target power generation physical model and a target data-knowledge model;x _k+1is composed ofkA second prediction result of the target new energy station at the moment + 1;f(X_k) Is composed ofkPredicting a target power generation physical model at a moment to obtain a physical prediction result;h (Y_k) Is composed ofkPredicting a target data-knowledge model at a moment to obtain a data-knowledge prediction result;uand carrying out random error vectors in the prediction process for the target data-knowledge model.

And step S3, predicting and outputting the target new energy station in parallel.

After the results of the two fusion mode predictions are obtained, the two output feature vectors are used as new inputs and placed in a newly-built learner for learning, the new learner is referred to as a parallel output learner in this embodiment, and finally, the final prediction result fused with the two mode predictions is output. The parallel prediction steps are as follows:

a) fusing the target power generation physical model and the target data-knowledge model through a learner to obtain a target data-physics-knowledge fusion prediction model fused through the learner, wherein the target data-physics-knowledge fusion prediction model fused through the learner can also be called as a first fusion model in the embodiment;

b) the target power generation physical model and the target data-knowledge model are subjected to weighted fusion to obtain a target data-physics-knowledge fusion prediction model subjected to weighted fusion, and the target data-physics-knowledge fusion prediction model subjected to weighted fusion can also be called a second fusion model in the embodiment;

c) station data, numerical weather forecast data and power generation historical operation data of the target new energy station are used as input data sets and are respectively input into the two fusion prediction models in the step a) and the step b) to respectively obtain a prediction result 1 and a prediction result 2;

d) and inputting the prediction result 1 and the prediction result 2 into a parallel output learning device to obtain a parallel prediction result, wherein the prediction result is a power prediction result of the target new energy station.

The power prediction results are classified according to the prediction time scale and can comprise ultra-short term power prediction results, short term power prediction results and medium and long term power prediction results.

If the prediction time scale is ultra-short-term prediction, the power prediction result is 16 prediction power values determined by obtaining one prediction power value every 15 minutes from 4 hours in the future, and the prediction power values are used for representing the generated power of the target new energy station in 4 hours in the future. If the prediction time scale is short-term prediction, the power prediction result is 288 predicted power values determined by obtaining one predicted power value every 15 minutes for 3 days in the future, and the predicted power values are used for representing the generated power of the target new energy station for 3 days in the future. If the prediction time scale is medium-long term prediction, the power prediction result is 960 prediction power values determined by obtaining a prediction power value every 15 minutes in the future 10 days, and the power prediction result is used for representing the generated power of the target new energy station in the future 10 days.

In this embodiment, a technical route is adopted in which a data model (i.e., a target data model) and a knowledge model (i.e., a target knowledge model) are fused, and then the data-knowledge model and a physical model (i.e., a target power generation physical model) are fused. Compared with a method of first performing fusion of a data model and a physical model to obtain a physical-data model and then performing fusion of the physical-data model and a knowledge model, the method has the following advantages:

a) the data-knowledge model is fused firstly, mass data and knowledge can be combined, on one hand, the knowledge in the data is also mined, and on the other hand, the data driving data model and the knowledge model are fused.

b) Compared with a physical model, the data-knowledge model has the characteristic of easier migration, and is beneficial to data and knowledge migration from other similar new energy stations.

c) In the embodiment, the physical model is used for modeling physical characteristics and physical laws of equipment, units, stations, meteorological environments and the like related to the new energy station, the main modeling forms are physical formulas, engineering formulas, mathematical equations and the like, the physical formulas, the engineering formulas, the mathematical equations and the like are suitable for being combined at last and are not suitable for being combined and iterated in the intermediate process, and the calculated amount and the complexity of the fusion process are greatly increased if the physical models are iterated in the intermediate process.

In this embodiment, the result of predicting the new energy power by the data-physics-knowledge fusion prediction model fused by the learner and the result of predicting the new energy power by the data-physics-knowledge fusion prediction model fused by weighting are combined by the parallel output learner, which has the following advantages:

a) indirectly outputting in parallel, re-digging data in the data-physics-knowledge fusion prediction model, re-learning knowledge in the data-physics-knowledge fusion prediction model, and re-correcting the physical model in the data-physics-knowledge fusion prediction model;

b) the prediction results obtained by the two fusion modes are mutually corrected, and the prediction error can be reduced from the technical and statistical aspects.

It should be understood that, although the steps in the flowcharts related to the embodiments as described above are sequentially displayed as indicated by arrows, the steps are not necessarily performed sequentially as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a part of the steps in the flowcharts related to the embodiments described above may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the execution order of the steps or stages is not necessarily sequential, but may be rotated or alternated with other steps or at least a part of the steps or stages in other steps.

Based on the same inventive concept, the embodiment of the application also provides a multi-dimensional fusion new energy power parallel prediction device for realizing the multi-dimensional fusion new energy power parallel prediction method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the method, so that specific limitations in one or more embodiments of the multi-dimensional fused new energy power parallel prediction device provided below can be referred to the limitations of the multi-dimensional fused new energy power parallel prediction method in the foregoing, and details are not repeated herein.

In one embodiment, as shown in fig. 6, there is provided a multi-dimensional fused new energy power parallel prediction apparatus 600, including: an acquisition module 602, a prediction module 604, and an output module 606, wherein:

an obtaining module 602, configured to obtain a target power generation physical model and a target data knowledge model corresponding to the target new energy station.

The obtaining module 602 is further configured to determine a first fusion model and a second fusion model obtained by fusing the target power generation physical model and the target data knowledge model according to a first fusion mode and a second fusion mode, respectively.

The obtaining module 602 is further configured to obtain to-be-processed data corresponding to the target new energy station, where the to-be-processed data includes station data, weather data, and historical power generation operation data of the target new energy station.

The prediction module 604 is configured to input the data to be processed into the first fusion model to obtain a first prediction result, and input the data to be processed into the second fusion model to obtain a second prediction result.

And the output module 606 is used for inputting the first prediction result and the second prediction result into the parallel output learner for processing to obtain a power prediction result of the target new energy station.

In one embodiment, the multi-dimensional fused new energy power parallel prediction apparatus 600 further includes a construction module, configured to determine candidate new energy stations matching the target new energy station, and obtain candidate knowledge models of the candidate new energy stations; performing migration learning based on the candidate knowledge model of the candidate new energy station and the expert knowledge base to construct a target knowledge model corresponding to the target new energy station; constructing a target data model based on station data, weather data and power generation historical operation data of the target new energy station and each candidate new energy station; and fusing the target knowledge model and the target data model to obtain the target data knowledge model.

In one embodiment, the construction module is further configured to calculate similar distances between the target new energy station and other new energy stations according to station data of the plurality of new energy stations; and determining the new energy stations with similar distances meeting the close range condition as candidate new energy stations.

In one embodiment, the first fusion mode includes a learner fusion mode, and the prediction module 604 is further configured to input the data to be processed of the target new energy station into the target power generation physical model to obtain a physical prediction result; and inputting the data to be processed and the physical prediction result of the target new energy station into a target data knowledge model to obtain a first prediction result.

In one embodiment, the second fusion mode includes a weighted fusion mode, and the prediction module 604 is further configured to obtain a modified fusion coefficient and an error vector corresponding to the target data knowledge model; inputting data to be processed into a target power generation physical model to obtain a physical prediction result; inputting data to be processed into a target data knowledge model to obtain a data knowledge prediction result; and performing weighted fusion processing on the physical prediction result and the data knowledge prediction result based on the corrected fusion coefficient and the error vector to obtain a second prediction result.

In one embodiment, the prediction module 604 is further configured to obtain training data corresponding to the target new energy site, and a historical power signature; inputting training data into a target power generation physical model to obtain a physical prediction result of an optimization stage; inputting training data into a target data knowledge model to obtain a data knowledge prediction result in an optimization stage; determining a weighted prediction result according to the corrected fusion variable, the physical prediction result of the optimization stage, the data knowledge prediction result of the optimization stage and the error vector corresponding to the target data knowledge model; and solving the correction fusion variable in an iterative calculation mode towards the direction of minimizing the difference between the historical power label and the weighted prediction result to obtain a correction fusion coefficient.

All modules in the multi-dimensional integrated new energy power parallel prediction device can be wholly or partially realized through software, hardware and a combination of the software and the hardware. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.

In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 7. The computer apparatus includes a processor, a memory, an input/output interface, a communication interface, a display unit, and an input device. The processor, the memory and the input/output interface are connected by a system bus, and the communication interface, the display unit and the input device are connected by the input/output interface to the system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The input/output interface of the computer device is used for exchanging information between the processor and an external device. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless communication can be realized through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. The computer program is executed by a processor to realize a multi-dimensional fused new energy power parallel prediction method. The display unit of the computer equipment is used for forming a visual and visible picture, and can be a display screen, a projection device or a virtual reality imaging device, the display screen can be a liquid crystal display screen or an electronic ink display screen, the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, and can also be an external keyboard, a touch pad or a mouse and the like.

Those skilled in the art will appreciate that the architecture shown in fig. 7 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.

In one embodiment, a computer device is provided, comprising a memory and a processor, the memory having stored therein a computer program, the processor implementing the steps of the above-described method embodiments when executing the computer program.

In an embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which computer program, when being executed by a processor, carries out the steps of the above-mentioned method embodiments.

In an embodiment, a computer program product is provided, comprising a computer program which, when being executed by a processor, carries out the steps of the above-mentioned method embodiments.

It should be noted that, the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) referred to in the present application are information and data authorized by the user or sufficiently authorized by each party, and the collection, use and management of the related data need to comply with the relevant laws and regulations and standards of the relevant country and region.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, database, or other medium used in the embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high-density embedded nonvolatile Memory, resistive Random Access Memory (ReRAM), Magnetic Random Access Memory (MRAM), Ferroelectric Random Access Memory (FRAM), Phase Change Memory (PCM), graphene Memory, and the like. Volatile Memory can include Random Access Memory (RAM), external cache Memory, and the like. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others. The databases referred to in various embodiments provided herein may include at least one of relational and non-relational databases. The non-relational database may include, but is not limited to, a block chain based distributed database, and the like. The processors referred to in the embodiments provided herein may be general purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, quantum computing based data processing logic devices, etc., without limitation.

The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.

The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present application. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A multi-dimension fused new energy power parallel prediction method is characterized by comprising the following steps:

determining a first fusion model and a second fusion model obtained by fusing the target power generation physical model and the target data knowledge model respectively according to a first fusion mode and a second fusion mode;

inputting the data to be processed into the first fusion model to obtain a first prediction result, and inputting the data to be processed into the second fusion model to obtain a second prediction result;

2. The method of claim 1, wherein the step of constructing the target data knowledge model comprises:

determining candidate new energy stations matched with the target new energy station, and acquiring candidate knowledge models of the candidate new energy stations;

performing migration learning based on the candidate knowledge model of the candidate new energy station and an expert knowledge base to construct a target knowledge model corresponding to the target new energy station;

constructing a target data model based on station data, weather data and power generation historical operation data of the target new energy station and each candidate new energy station;

and fusing the target knowledge model and the target data model to obtain a target data knowledge model.

3. The method of claim 2, wherein determining candidate new energy sites that match the target new energy site comprises:

calculating similar distances between a target new energy station and other new energy stations according to station data of the new energy stations;

and determining the new energy stations with similar distances meeting the close range condition as candidate new energy stations.

4. The method according to claim 1, characterized in that in the case where the target new energy farm belongs to a wind power farm, the target generation physics model comprises a wind turbine model and a wind farm model taking into account wake effects; the wind power plant model comprises a wind power generation model, a wind power generation model and a wind power generation model, wherein the wind power generation model comprises at least one of a circuit model, a magnetic circuit model, a mechanical model and a power and wind speed model of wind power generation;

under the condition that the target new energy station belongs to a photovoltaic power generation station, the target power generation physical model comprises a photovoltaic single-machine model and a photovoltaic power station model; the photovoltaic single-machine model comprises at least one of a silicon crystal plate model, a photovoltaic plate model and a power irradiance model, and the photovoltaic power station model comprises at least one of a photovoltaic array model and a photovoltaic cluster model.

5. The method according to claim 1, wherein the first fusion mode comprises a learner fusion mode, and the inputting the data to be processed into the first fusion model to obtain a first prediction result comprises:

inputting the data to be processed of the target new energy station into the target power generation physical model to obtain a physical prediction result;

and inputting the data to be processed of the target new energy station and the physical prediction result into the target data knowledge model to obtain a first prediction result.

6. The method according to claim 5, wherein the second fusion mode comprises a weighted fusion mode, and the inputting the data to be processed into the second fusion model to obtain a second prediction result comprises:

acquiring a correction fusion coefficient and an error vector corresponding to the target data knowledge model;

inputting the data to be processed into the target power generation physical model to obtain a physical prediction result;

inputting the data to be processed into the target data knowledge model to obtain a data knowledge prediction result;

and performing weighted fusion processing on the physical prediction result and the data knowledge prediction result based on the corrected fusion coefficient and the error vector to obtain a second prediction result.

7. The method according to claim 6, wherein the obtaining of the modified fusion coefficients comprises:

acquiring training data corresponding to the target new energy station and a historical power label;

inputting the training data into the target power generation physical model to obtain a physical prediction result of an optimization stage;

inputting the training data into the target data knowledge model to obtain a data knowledge prediction result in an optimization stage;

determining a weighted prediction result according to the corrected fusion variable, the physical prediction result of the optimization stage, the data knowledge prediction result of the optimization stage and the error vector corresponding to the target data knowledge model;

and solving the correction fusion variable in an iterative calculation mode towards the direction of minimizing the difference between the historical power label and the weighted prediction result to obtain a correction fusion coefficient.

8. A multidimensional-fused new energy power parallel prediction device is characterized by comprising:

the acquisition module is further used for determining a first fusion model and a second fusion model obtained by fusing the target power generation physical model and the target data knowledge model respectively according to a first fusion mode and a second fusion mode;

and the output module is used for inputting the first prediction result and the second prediction result into a parallel output learning device for processing to obtain a power prediction result of the target new energy station.

9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the method of any of claims 1 to 7.

10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.