CN117878930B - Short-term power prediction method, device, terminal and storage medium for distributed photovoltaic - Google Patents

Short-term power prediction method, device, terminal and storage medium for distributed photovoltaic Download PDF

Info

Publication number
CN117878930B
CN117878930B CN202410275636.0A CN202410275636A CN117878930B CN 117878930 B CN117878930 B CN 117878930B CN 202410275636 A CN202410275636 A CN 202410275636A CN 117878930 B CN117878930 B CN 117878930B
Authority
CN
China
Prior art keywords
photovoltaic
data
time period
time
power
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.)
Active
Application number
CN202410275636.0A
Other languages
Chinese (zh)
Other versions
CN117878930A (en
Inventor
王会平
申光鹏
刘晓琳
聂泽
全佳杰
刘林杰
周彪
王宁
杜涛
杨萌
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shijiazhuang Kelin Yunneng Information Technology Co ltd
Shijiazhuang Kelin Electric Co Ltd
Original Assignee
Shijiazhuang Kelin Yunneng Information Technology Co ltd
Shijiazhuang Kelin Electric Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shijiazhuang Kelin Yunneng Information Technology Co ltd, Shijiazhuang Kelin Electric Co Ltd filed Critical Shijiazhuang Kelin Yunneng Information Technology Co ltd
Priority to CN202410275636.0A priority Critical patent/CN117878930B/en
Publication of CN117878930A publication Critical patent/CN117878930A/en
Application granted granted Critical
Publication of CN117878930B publication Critical patent/CN117878930B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/004—Generation forecast, e.g. methods or systems for forecasting future energy generation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
    • H02J3/381—Dispersed generators
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
    • H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J2300/00—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation
    • H02J2300/20—The dispersed energy generation being of renewable origin
    • H02J2300/22—The renewable source being solar energy
    • H02J2300/24—The renewable source being solar energy of photovoltaic origin
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Economics (AREA)
  • Power Engineering (AREA)
  • Physiology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)
  • Human Resources & Organizations (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Evolutionary Biology (AREA)
  • Photovoltaic Devices (AREA)

Abstract

The invention provides a short-term power prediction method, a device, a terminal and a storage medium of distributed photovoltaic. The method comprises the following steps: acquiring photovoltaic data at different moments in at least one time period before a predicted time period; inputting the photovoltaic data into a pre-trained power prediction model to obtain photovoltaic power generation power at different moments in a prediction time period output by the power prediction model; the power prediction model is obtained by training based on photovoltaic data at different moments in at least one time period before a plurality of time periods and photovoltaic power generation power at different moments in a corresponding time period, and is a negative feedback neural network model which comprises a plurality of processing units, and hidden states output by the processing units in each time step are correspondingly fed back to the processing units in the previous time step. The invention can improve the prediction precision of the generated power of the distributed photovoltaic.

Description

Short-term power prediction method, device, terminal and storage medium for distributed photovoltaic
Technical Field
The invention relates to the technical field of photovoltaic power generation, in particular to a short-term power prediction method, device, terminal and storage medium of distributed photovoltaic.
Background
Photovoltaic power generation utilizes the photovoltaic effect of semiconductor elements to convert light energy into electrical energy. Compared with other renewable resources, the photovoltaic power generation has the advantages of safety, low investment, convenient construction and the like. In recent years, photovoltaic power generation is widely popularized in the global scope due to the remarkable advantages of cleanliness, no pollution, convenience in distributed popularization and the like.
However, due to the influence of meteorological factors, the generated power of the distributed photovoltaic has strong instability, the fluctuation of the generated power of the distributed photovoltaic is severe under abrupt weather conditions, grid connection of the distributed photovoltaic can not only influence the electric energy quality of a power grid, but also cause difficulty in active economic dispatching of the power grid, and the safe and stable operation of the power grid is influenced.
Therefore, how to accurately predict the power generated by the distributed photovoltaic in a short period helps the power dispatching department to formulate a dispatching scheme, so as to optimally dispatch each distributed photovoltaic, reduce the influence of unstable output of photovoltaic power generation on the power grid, and become a problem to be solved urgently.
Disclosure of Invention
The embodiment of the invention provides a short-term power prediction method, device, terminal and storage medium of distributed photovoltaic, which are used for accurately predicting photovoltaic power generation power in a short term.
In a first aspect, an embodiment of the present invention provides a method for short-term power prediction of distributed photovoltaic, including:
Acquiring photovoltaic data at different moments in at least one time period before a predicted time period;
Inputting the photovoltaic data into a pre-trained power prediction model to obtain photovoltaic power generation power at different moments in the prediction time period output by the power prediction model;
the power prediction model is obtained by training based on photovoltaic data at different moments in at least one time period before a plurality of time periods and photovoltaic power generation power at different moments in a corresponding time period, and is a negative feedback neural network model which comprises a plurality of processing units, and hidden states output by the processing units in each time step are correspondingly fed back to the processing units in the previous time step.
In one possible implementation, the negative feedback neural network model is a modified LSTM neural network model, and the processing unit is an LSTM unit;
in the improved LSTM neural network model, the hidden state output by each LSTM unit in each time step is correspondingly fed back to the input end of each LSTM unit in the previous time step;
the improved LSTM neural network model further comprises a plurality of convolution layers, wherein the input end of each convolution layer is correspondingly connected to the output end of the last-stage LSTM unit under each time step, and the output end of each convolution layer serves as the output end of the improved LSTM neural network model.
In one possible implementation manner, before the photovoltaic data is input into a pre-trained power prediction model, obtaining photovoltaic power generation power at different moments in the prediction time period output by the power prediction model, the method further includes:
carrying out data cleaning on the photovoltaic data; the data cleaning comprises missing data complement processing and abnormal data correction processing;
inputting the photovoltaic data into a pre-trained power prediction model to obtain photovoltaic power generation power at different moments in the prediction time period output by the power prediction model, wherein the method comprises the following steps:
and inputting the photovoltaic data subjected to data cleaning into the power prediction model to obtain photovoltaic power generation power at different moments in the prediction time period output by the power prediction model.
In one possible implementation manner, the processing for correcting the abnormal data is performed on the photovoltaic data, including:
Respectively inputting the photovoltaic data into a pre-trained abnormal time period identification model, and correspondingly determining an abnormal time period; the abnormal time period identification model is obtained based on photovoltaic data at different moments in different time periods and abnormal labels in corresponding time periods through training;
And correspondingly determining abnormal data in the photovoltaic data in the abnormal time period based on the photovoltaic data at each time in the abnormal time period, and correcting the abnormal data.
In a possible implementation manner, based on the photovoltaic data at each time point in the abnormal time period, determining the abnormal data in the photovoltaic data in the abnormal time period correspondingly includes:
sliding a sliding window with a preset length along a time sequence data sequence formed by the photovoltaic data at each time in the abnormal time period according to a preset step;
Calculating a window average value corresponding to the sliding window at the current moment and a window average value corresponding to the sliding window at the last moment in real time, if the difference value between the window average value corresponding to the current moment and the window average value corresponding to the last moment is larger than a first preset difference value, shortening the length of the sliding window, and eliminating data corresponding to the tail end of the sliding window at the current moment to obtain a new sliding window;
The step of calculating the window average value corresponding to the sliding window at the current moment and the window average value corresponding to the sliding window at the last moment in real time is carried out in a skip mode, when the difference value between the window average value corresponding to the current moment and the window average value corresponding to the last moment is smaller than or equal to a second preset difference value, the data which are removed last time are determined to be abnormal data, the sliding window continues to slide along the time sequence, and the step of calculating the window average value corresponding to the sliding window at the current moment and the window average value corresponding to the sliding window at the last moment in real time is carried out in a skip mode until the sliding window traverses the time sequence, and all abnormal data in the photovoltaic data in the abnormal time period are obtained; wherein the second preset difference is smaller than the first preset difference.
In one possible implementation, the photovoltaic data includes: photovoltaic power generation power, corrected grid-connected point voltage and corrected irradiance;
The obtaining photovoltaic data at different moments in at least one time period before the predicted time period comprises the following steps:
Acquiring photovoltaic power generation power, grid-connected point voltage and irradiance at different moments in at least one time period before a predicted time period;
respectively obtaining the voltage of the grid-connected point and the regression coefficient corresponding to the irradiance; the regression coefficient is used for representing the influence degree of the power generated by the photovoltaic power generation;
And correspondingly determining the corrected grid-connected point voltage and the corrected irradiance based on the grid-connected point voltage, the irradiance and the corresponding regression coefficient.
In one possible implementation manner, the obtaining the regression coefficients corresponding to the grid-connected point voltage and the irradiance respectively includes:
According to Constructing a regression equation;
Wherein, Representing regression coefficient corresponding to grid-connected point voltage,/>Representing grid-tie point voltage,/>Representing regression coefficients corresponding to irradiance,/>Representing irradiance,/>Representing photovoltaic power generation;
And constructing a loss function, and solving the regression equation according to the principle of minimum loss function to obtain the grid-connected point voltage and the regression coefficient corresponding to the irradiance.
In a second aspect, an embodiment of the present invention provides a short-term power prediction apparatus for distributed photovoltaic, including:
the acquisition module is used for acquiring photovoltaic data at different moments in at least one time period before the predicted time period;
The prediction module is used for inputting the photovoltaic data into a pre-trained power prediction model to obtain photovoltaic power generation power at different moments in the prediction time period output by the power prediction model;
the power prediction model is obtained by training based on photovoltaic data at different moments in at least one time period before a plurality of time periods and photovoltaic power generation power at different moments in a corresponding time period, and is a negative feedback neural network model which comprises a plurality of processing units, and hidden states output by the processing units in each time step are correspondingly fed back to the processing units in the previous time step.
In a third aspect, embodiments of the present invention provide a terminal comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the method according to the first aspect or any one of the possible implementations of the first aspect, when the computer program is executed.
In a fourth aspect, embodiments of the present invention provide a computer readable storage medium storing a computer program which, when executed by a processor, implements the steps of the method as described above in the first aspect or any one of the possible implementations of the first aspect.
The embodiment of the invention provides a short-term power prediction method, a device, a terminal and a storage medium of distributed photovoltaic, which are used for obtaining photovoltaic power generation power at different moments in a predicted time period by inputting photovoltaic data at different moments in at least one time period before the predicted time period into a pre-trained power prediction model. The power prediction model is a negative feedback neural network model, the negative feedback neural network model comprises a plurality of processing units, and the hidden state output by each processing unit in each time step is correspondingly fed back to each processing unit in the previous time step. The embodiment of the invention provides a mode for predicting the photovoltaic power generation by utilizing the thought of model prediction, which not only rapidly predicts the photovoltaic power generation and helps a power dispatching department to formulate a dispatching scheme, but also enables the processing unit at the current time step to process data based on the hidden state at the next time step by correspondingly feeding back the hidden state output by the processing units at each time step in the model to the processing units at the previous time step, thereby being capable of capturing the time sequence dependency relationship in the photovoltaic data at different moments more effectively and improving the prediction precision of the photovoltaic power generation.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments or the description of the prior art will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of an implementation of a method for short-term power prediction of distributed photovoltaic provided by an embodiment of the present invention;
FIG. 2 is a schematic diagram of an LSTM neural network model according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of another LSTM neural network model according to an embodiment of the present invention;
FIG. 4 is a flowchart of an implementation of determining exception data provided by an embodiment of the present invention;
FIG. 5 is a schematic diagram of a process for shortening the length of a sliding window according to an embodiment of the present invention;
Fig. 6 is a schematic structural diagram of a short-term power prediction device of a distributed photovoltaic according to an embodiment of the present invention;
fig. 7 is a schematic diagram of a terminal according to an embodiment of the present invention.
Detailed Description
In the following description, for purposes of explanation and not limitation, specific details are set forth such as the particular system architecture, techniques, etc., in order to provide a thorough understanding of the embodiments of the present invention. It will be apparent, however, to one skilled in the art that the present invention may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present invention with unnecessary detail.
The photovoltaic power generation power has strong instability, and the photovoltaic power generation power fluctuation is more severe under abrupt weather conditions. The fluctuation of the photovoltaic power generation power obviously affects the safe and stable operation of the power grid. Therefore, how to accurately predict the photovoltaic power generation power in a short period, so as to optimally schedule each distributed photovoltaic in time, so as to reduce the influence of unstable output of photovoltaic power generation on a power grid, and the method is a problem to be solved urgently.
In order to improve the accuracy of photovoltaic power generation prediction in a short period of time, the photovoltaic power generation is predicted based on a trained power prediction model in the embodiment of the present application. The power prediction model is a negative feedback neural network model, and the hidden state output by each processing unit in each time step is correspondingly fed back to each processing unit in the previous time step, so that information flow and updating processes can be established in different time steps, the negative feedback neural network model can effectively capture time sequence dependency relations in photovoltaic data at different moments, and the prediction accuracy of photovoltaic power generation power is improved.
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the following description will be made by way of specific embodiments with reference to the accompanying drawings.
Fig. 1 is a flowchart of an implementation of a method for predicting short-term power of a distributed photovoltaic according to an embodiment of the present invention, which is described in detail below:
step 101, obtaining photovoltaic data at different moments in at least one time period before a predicted time period.
The photovoltaic data herein may include: photovoltaic power generation, grid-tie point voltage, and irradiance. That is, the embodiment of the invention can predict and obtain the photovoltaic power generation power at different moments in the prediction time period based on the photovoltaic power generation power, the grid-connected point voltage and the irradiance at different moments in different time periods before the prediction time period.
Here, the duration of the time period may be determined according to practical situations, which is not particularly limited in the embodiment of the present invention. Illustratively, the time period may be one day long. That is, if photovoltaic power generation power is predicted for a certain day, photovoltaic power generation power, grid-connected point voltage, and irradiance at different times within at least one day before the predicted day are obtained.
It should be noted that there is a time-series dependency relationship between the photovoltaic power generation powers at different times. Therefore, the embodiment of the invention can acquire the photovoltaic data at different moments in at least one time period which is nearest to the predicted time period and is used for carrying out the photovoltaic power generation power at different moments in the predicted time period, thereby improving the prediction accuracy of the photovoltaic power generation power.
And 102, inputting the photovoltaic data into a pre-trained power prediction model to obtain photovoltaic power generation power at different moments in a prediction time period output by the power prediction model. The power prediction model is obtained based on photovoltaic data at different moments in at least one time period before a plurality of time periods and photovoltaic power generation power at different moments in the corresponding time period. The power prediction model is a negative feedback neural network model. The negative feedback neural network model comprises a plurality of processing units, and the hidden state output by each processing unit in each time step is correspondingly fed back to each processing unit in the previous time step.
According to the embodiment of the invention, the photovoltaic power generation power, the grid-connected point voltage and the irradiance are converted into a matrix form, and the matrix form is used as an input sequence to be input into a power prediction model for carrying out photovoltaic power generation power prediction. Wherein the input sequence in matrix form can be expressed as:。
Wherein, Representing the photovoltaic power generation power at a first time in a first time period,/>Representing the first time period of the first/>Photovoltaic power generation power at each moment,/>Representing the voltage of the grid-connected point at the first moment in the first time period,/>Representing the first time period of the first/>Grid-connected point voltage at each moment,/>Representing irradiance at a first time instant during a first time period,/>Representing the first time period of the first/>Irradiance at a time,/>Represents the/>Photovoltaic power generation power at first moment in each time period,/>Represents the/>Within each time period/>Photovoltaic power generation power at each moment,/>Represents the/>Grid-connected point voltage at first time in each time period,/>Represents the/>Within each time period/>Grid-connected point voltage at each moment,/>Represents the/>Irradiance at a first time during a time period,Represents the/>Within each time period/>Irradiance at each time instant.
Before the input sequence in the matrix form is input into the pre-trained power prediction model, the embodiment of the invention can clean the input sequence in advance to clean some bad data, improve the accuracy of subsequent processing results, and normalize the cleaned input sequence by the data so that each data in the input sequence is between-1 and 1, thereby facilitating the subsequent model processing.
And each processing unit in the same time step is used for processing the photovoltaic data in the same time in the negative feedback neural network model. And the negative feedback neural network model correspondingly processes the photovoltaic data at different moments in at least one time period before the predicted time period by utilizing each processing unit under different time steps, and correspondingly outputs the photovoltaic power generation power at different moments in the predicted time period.
In some embodiments, the negative feedback neural network model is a modified LSTM neural network model and the processing unit is a LSTM unit.
In the improved LSTM neural network model, the hidden state output by each LSTM unit in each time step is correspondingly fed back to the input end of each LSTM unit in the previous time step.
The improved LSTM neural network model further comprises a plurality of convolution layers, wherein the input end of each convolution layer is correspondingly connected to the output end of the last LSTM unit under each time step, and the output end of each convolution layer serves as the output end of the improved LSTM neural network model.
In the LSTM neural network model, the LSTM units under each time step are used for correspondingly processing the photovoltaic data under different moments and correspondingly outputting the photovoltaic power generation power under the corresponding moments. The hidden state output by each LSTM unit in each time step is correspondingly transmitted to each LSTM unit in the next time step in turn, so that the LSTM can keep memorizing the past information, and the time sequence data can be processed conveniently. The hidden state contains sequence information contained in the photovoltaic data up to the current time step. For example, referring to fig. 2, in the LSTM neural network model, each LSTM unit at time t-1 corresponds to processing the photovoltaic data Xt-1 at time t-1 in each time period, and outputs the photovoltaic power Yt-1 at time t-1 in the predicted time period. And (3) correspondingly processing the photovoltaic data Xt at the t moment in each time period by each LSTM unit under the t time step, and correspondingly outputting the photovoltaic power Yt at the t moment in the predicted time period. And each LSTM unit in the t+1 time step correspondingly processes the photovoltaic data Xt+1 in the t+1 time in each time period and correspondingly outputs the photovoltaic power Yt+1 in the t+1 time in the predicted time period. the hidden state ht-1 output by each LSTM unit under the t-1 time step is correspondingly transmitted to each LSTM unit under the t time step in sequence. the hidden state ht output by each LSTM unit in the t time step is correspondingly transmitted to each LSTM unit in the t+1 time step in sequence.
Referring to fig. 3, the embodiment of the present invention further improves the LSTM neural network model, and by feeding back the hidden state output by each LSTM unit in each time step to the input end of each LSTM unit in the previous time step, when the photovoltaic data at the current moment is processed by each LSTM unit in each time step, the hidden state in the previous time step can be referred to, and the hidden state in the next time step can be referred to, so that the time correlation between the photovoltaic data at different moments is enhanced, and further the technical effect of improving the photovoltaic power prediction accuracy is achieved.
In addition, referring to fig. 3, in the embodiment of the invention, the output end of the last-stage LSTM unit under each time step is correspondingly provided with a convolution layer, so that the convolution operation is performed on the photovoltaic power generated by each last-stage LSTM unit, thereby filtering high-frequency impurity information in the photovoltaic power generated, and further improving the prediction precision of the photovoltaic power generated.
Here, the convolution kernel of each convolution layer may be set according to the actual situation. Illustratively, the convolution kernels of the convolution layers in embodiments of the present disclosure may be designed to:。
It should be noted that fig. 2 and 3 are only exemplary drawings, and exemplarily show t-1 time step, t time step, and 3 LSTM cells under t+1 time step. But this is not a limitation on the LSTM neural network model. The user can determine the number of time steps and the number of LSTM units under each time step according to the actual situation.
In some embodiments, model training may also be performed before photovoltaic data is input into a pre-trained power prediction model to obtain photovoltaic power generated at different times within a predicted time period output by the power prediction model. For each time period, inputting the photovoltaic data at different moments in at least one time period before the time period into a power prediction model to obtain photovoltaic power generation power at different moments in the time period;
And adjusting the power prediction model according to the photovoltaic power generation power at different moments in the corresponding time period, the real photovoltaic power generation power at different moments in the corresponding time period and the loss function, and obtaining the trained power prediction model.
In this embodiment, photovoltaic data at different moments in time before each time period is used as input data of a power prediction model, real photovoltaic power generated at different moments in time during the corresponding time period is used as tag data of the power prediction model, and the power prediction model is subjected to supervised training, so that the photovoltaic power generated at different moments in time predicted by the power prediction model approaches to the real photovoltaic power generated at different moments in time during the corresponding time period, and at the moment, the value of a loss function is smaller, and a trained power prediction model is obtained.
Compared with the prior art, the photovoltaic power generation method and device provided by the embodiment of the invention have the advantages that the photovoltaic data at different moments in at least one time period before the predicted time period are input into the power prediction model, so that the photovoltaic power generation power at different moments in the predicted time period is predicted. The power prediction model is a negative feedback neural network model, the negative feedback neural network model comprises a plurality of processing units, and the hidden state output by each processing unit in each time step is correspondingly fed back to each processing unit in the previous time step. The hidden states output by the processing units in each time step are correspondingly fed back to the processing units in the previous time step, so that the processing units in the current time step can process data based on the hidden states in the next time step, the whole negative feedback neural network model can effectively capture time sequence dependency relations in the photovoltaic data at different moments, and the prediction accuracy of the photovoltaic power generation power is improved.
In some embodiments, to avoid affecting the accuracy of the prediction of the photovoltaic power due to the deviation of the grid-tie voltage and irradiance, the grid-tie voltage and irradiance may be modified, and the power prediction may be performed based on the modified grid-tie voltage, the modified irradiance data, and the photovoltaic power. That is, the photovoltaic data may include: photovoltaic power generation power, corrected grid-connected point voltage and corrected irradiance.
Correspondingly, the obtaining the photovoltaic data at different moments in at least one time period before the predicted time period may include:
and acquiring photovoltaic power generation power, grid-connected point voltage and irradiance at different moments in at least one time period before the predicted time period.
And respectively acquiring regression coefficients corresponding to the grid-connected point voltage and the irradiance. The regression coefficient is used for representing the influence degree of photovoltaic power generation power.
And correspondingly determining the corrected grid-connected point voltage and the corrected irradiance based on the grid-connected point voltage, the irradiance and the corresponding regression coefficient.
Here, the corrected grid-connected point voltage can be determined by calculating the product of the grid-connected point voltage and the regression coefficient corresponding to the grid-connected point voltage. Accordingly, the corrected irradiance can be determined by calculating the product of the irradiance and its corresponding regression coefficient.
In some embodiments, the obtaining regression coefficients corresponding to the grid-connected point voltage and the irradiance may include:
According to And constructing a regression equation.
Wherein,Representing regression coefficient corresponding to grid-connected point voltage,/>Representing grid-tie point voltage,/>Representing regression coefficients corresponding to irradiance,/>Representing irradiance,/>Representing the photovoltaic power generation.
And constructing a loss function, and solving a regression equation according to the principle of minimum loss function to obtain regression coefficients corresponding to grid-connected point voltage and irradiance.
In the embodiment of the invention, the photovoltaic power generation power, the grid-connected point voltage and the irradiance at different moments in at least one time period before the predicted time period are brought into the aboveAnd carrying out regression calculation by taking a least square method as a loss function to correspondingly obtain regression coefficients corresponding to the grid-connected point voltage and irradiance.
According to the embodiment of the invention, the photovoltaic power generation power, the corrected grid-connected point voltage and the corrected irradiance are converted into a matrix form and are input into a power prediction model as an input sequence for photovoltaic power generation power prediction. The input sequence in matrix form can be expressed as:
。
Wherein, Representing the corrected grid-connected point voltage at the first moment in the first time period,/>Representing the first time period of the first/>Grid-connected point voltage corrected at each moment,/>Representing corrected irradiance at a first time during a first time period,/>Representing the first time period of the first/>Corrected irradiance at each instant,/>Represents the/>Grid-connected point voltage corrected at first moment in each time period,/>Represents the/>Within each time period/>Corrected grid-connected point voltage at each moment,/>Represents the/>Corrected irradiance at the first instant in time period,/>Represents the/>Within each time period/>Corrected irradiance at each instant.
In the embodiment of the invention, the grid-connected point voltage and irradiance are used as influencing factors of the photovoltaic power generation power. The grid-connected point voltage and the regression coefficient corresponding to the irradiance are respectively used for representing the influence degree of the grid-connected point voltage and the irradiance on the photovoltaic power generation power. And on the basis of the regression coefficient, correcting the grid-connected point voltage and irradiance respectively, and on the basis of the corrected grid-connected point voltage and corrected irradiance, predicting the photovoltaic power generation power, so that the prediction accuracy of the photovoltaic power generation power can be further improved.
In some embodiments, before inputting the photovoltaic data into the pre-trained power prediction model to obtain the photovoltaic power generation power at different moments in the prediction time period output by the power prediction model, the method further includes:
carrying out data cleaning on the photovoltaic data; the data cleansing includes missing data complement processing and abnormal data correction processing.
Inputting the photovoltaic data into a power prediction model to obtain photovoltaic power generation power at different moments in a prediction time period output by the power prediction model, wherein the photovoltaic power generation power comprises the following components:
and inputting the photovoltaic data subjected to data cleaning into a power prediction model to obtain photovoltaic power generation power at different moments in a prediction time period output by the power prediction model.
The photovoltaic data herein may include: photovoltaic power generation, corrected grid-tie point voltage, or corrected irradiance. According to the embodiment of the invention, the abnormal data in the photovoltaic data can be corrected and the missing data can be complemented by cleaning the data of the photovoltaic data. And the prediction of the photovoltaic power generation power is carried out based on the photovoltaic data after data cleaning, so that the prediction precision of the photovoltaic power generation power can be further improved.
In some embodiments, performing the abnormal data correction process on the photovoltaic data may include:
And respectively inputting the photovoltaic data in each time period into a pre-trained abnormal time period identification model, and correspondingly determining the abnormal time period. The abnormal time period identification model is obtained based on photovoltaic data at different moments in different time periods and abnormal labels in corresponding time periods through training.
And correspondingly determining the abnormal data in the photovoltaic data in the abnormal time period based on the photovoltaic data at each time in the abnormal time period, and correcting the abnormal data.
Here, when the photovoltaic data in each period is input into the pre-trained abnormal period identification model, the photovoltaic data in each period may be converted into a matrix form and then input into the pre-trained abnormal period identification model.
When converting the photovoltaic data in each time period into a matrix form, determining the nearest square value corresponding to the photovoltaic data according to the photovoltaic data quantity in the current time period. The matrix size is then determined based on the nearest neighbor square value. And finally, filling the photovoltaic data into each row in the matrix in sequence to obtain the photovoltaic data in the matrix form.
The nearest square value corresponding to the photovoltaic data quantity is as follows: greater than or equal to the smallest of all squares of the number of photovoltaic data. The code representation mode is as follows:
Wherein, Representing nearest square value,/>Representing the amount of photovoltaic data.
Correspondingly, the matrix size is. Sequentially filling photovoltaic data in the current time period to/>Is defined in the matrix of (a) and (b). If the photovoltaic data quantity is insufficient, the rest positions in the matrix are all zero-padded.
Illustratively, if the number of photovoltaic data in the current time period is 96, the corresponding nearest square value is 100. Accordingly, the matrix size is. I.e. the matrix is a square matrix of 10 rows and 10 columns. And filling the photovoltaic data in the current time period into each row in the square matrix of 10 rows and 10 columns in sequence, and filling all the positions which are not filled with the photovoltaic data with zero.
In some embodiments, the photovoltaic data in each time period is input into a pre-trained abnormal time period identification model, and model training may also be performed before the abnormal time period is correspondingly determined. Illustratively, photovoltaic data at different moments in each time period are respectively input into an abnormal time period identification model, and abnormal time periods are correspondingly determined;
And adjusting model parameters of the abnormal time period identification model according to the determined abnormal time period, the abnormal label and the loss function of the corresponding time period to obtain a trained abnormal time period identification model. Wherein the anomaly flag of the respective time period is used to indicate whether the time period is an anomaly time period.
In this embodiment, photovoltaic data in each time period is used as input data of an abnormal time period identification model, an abnormal label in each time period is used as label data of the abnormal time period identification model, and the abnormal time period identification model is subjected to supervised training, so that a trained abnormal time period identification model is obtained.
Here, the abnormal period identification model may be a convolutional neural network model. The convolutional neural network model includes a convolutional layer and a classification layer. The convolution layer is used for carrying out convolution processing on input data in sequence and outputting a convolution result correspondingly. The classification layer is used for carrying out classification recognition based on all convolution results and outputting classification labels so as to determine whether the current time period is an abnormal time period or not. The convolution kernel of the convolution layer can be determined according to the actual situation. Exemplary embodiments of the present invention employ、/>/>And (5) a convolution kernel. The convolution kernel is specifically designed as follows:
in addition, the embodiments of the present invention take the above into consideration Zero value data may exist in the matrix of the (2), and in order to avoid that the zero value data affects the recognition accuracy of the abnormal time period recognition model, when the classification layer in the abnormal time period recognition model performs classification recognition, the convolution result received last time is removed, and classification recognition is performed based on the remaining convolution result.
In some embodiments, referring to fig. 4, the determining, based on the photovoltaic data at each time point in the abnormal time period, the abnormal data in the photovoltaic data in the abnormal time period correspondingly may include:
Step 401, sliding a sliding window with a preset length along a time sequence data sequence formed by photovoltaic data at each time in an abnormal time period according to a preset step.
In order to avoid the influence on the accuracy of identifying abnormal data due to the photovoltaic data during the photovoltaic shutdown period, the method is used. In a time sequence data sequence formed by photovoltaic data at each time in an abnormal time period, the sliding window slides along the time sequence data sequence according to preset steps from the photovoltaic data corresponding to the time when the photovoltaic power generation power is greater than 0.
The preset length and the preset step can be determined according to practical situations. The embodiment of the present invention is not particularly limited thereto. Illustratively, the preset step may be 1.
Step 402, calculating a window average value corresponding to the sliding window at the current moment and a window average value corresponding to the sliding window at the last moment in real time, and if the difference between the window average value corresponding to the current moment and the window average value corresponding to the last moment is greater than a first preset difference value, shortening the length of the sliding window to remove the data corresponding to the tail end of the sliding window at the current moment, thereby obtaining a new sliding window.
The calculation formula of the window average value corresponding to the sliding window can be expressed as:。
Wherein, Represent window mean value,/>Representing the length of the sliding window, i.e. the corresponding photovoltaic data quantity of the sliding window,/>Indicate the corresponding/>, of the sliding windowAnd (5) photovoltaic data.
Illustratively, referring to fig. 5, the sliding window has a length of 5, i.e., the sliding window corresponds to 5 data in the time series data sequence at each time. The preset step is 1, i.e. the sliding window slides through 1 data in the time series data sequence at each moment. And in the sliding process of the sliding window, respectively calculating the window average value corresponding to the sliding window at the current moment and the window average value corresponding to the sliding window at the last moment. If the difference between the window average value corresponding to the current moment and the window average value corresponding to the last moment is larger than the first preset difference, the data corresponding to the current moment at the tail end of the sliding window is removed, so that the length of the sliding window is shortened, and a new sliding window is obtained. And re-calculating the window average value corresponding to the new sliding window at the current moment and the window average value corresponding to the new sliding window at the last moment.
Step 403, skip execution of the step of calculating the window average value corresponding to the current time of the sliding window and the window average value corresponding to the last time of the sliding window in real time until the difference between the window average value corresponding to the current time and the window average value corresponding to the last time is smaller than or equal to a second preset difference value, determining the last removed data as abnormal data, continuing sliding the sliding window along the time sequence data sequence, skip execution of the step of calculating the window average value corresponding to the current time of the sliding window and the window average value corresponding to the last time of the sliding window in real time until the sliding window traverses the sequence data, and obtaining all abnormal data in the photovoltaic data in the abnormal time period. The second preset difference value is smaller than the first preset difference value.
That is, in the process of sliding the sliding window on the time sequence data sequence, the window average value corresponding to the current time and the last time of the sliding window is calculated in real time. If the difference between the window average value at the current moment and the window average value at the last moment is larger than the first preset difference value, shortening the length of the sliding window, removing the data corresponding to the tail end of the sliding window at the current moment, obtaining a new sliding window, and recalculating the window average value corresponding to the new sliding window at the current moment and the window average value corresponding to the sliding window at the last moment. And determining the data which is removed last time as abnormal data until the difference value between the window average value corresponding to the current moment and the window average value corresponding to the last moment is smaller than or equal to a second preset difference value. And continuing to slide the sliding window along the time sequence data sequence until the sliding window traverses all data in the time sequence data sequence, so as to determine all abnormal data in the time sequence data sequence.
The first preset difference value and the second preset difference value can be determined according to actual conditions. By way of example, the embodiment of the invention can utilize the sliding window with the preset length to sequentially slide along the time sequence data sequence formed by the photovoltaic data in at least one normal time period according to the preset steps, and calculate the average value of the window corresponding to the sliding window at different moments in real time. The embodiment of the invention determines the maximum value in all the window average values as a first preset difference value, and determines the minimum value in all the window average values as a second preset difference value.
In some embodiments, the correcting the abnormal data may include:
Here, the time corresponding to each of the different data in the abnormal time period is determined as an abnormal time, and for each abnormal time, the time corresponding to the abnormal time is determined from at least one time period adjacent to the abnormal time period, and the average value of the photovoltaic data at each corresponding time is calculated. The average value is replaced with the abnormal data, thereby realizing the correction of the abnormal data.
In the embodiment of the invention, when the abnormal data is corrected, the time corresponding to the abnormal time in N time periods which are positioned before the abnormal time period and are nearest to the abnormal time period and the time corresponding to the abnormal time in N time periods which are positioned after the abnormal time period and are nearest to the abnormal time are respectively selected as all the time corresponding to the abnormal time. And calculating the average value of the photovoltaic data at all corresponding moments, and replacing the abnormal data with the average value. The calculation formula can be expressed as follows:
。
Wherein, Representing replacement values of anomalous data,/>Represents the first/>, located before and closest to the abnormal time periodPhotovoltaic data at a time corresponding to the abnormal time in a time period,/>Represents the second/>, located after and closest to the abnormal time periodPhotovoltaic data at a time corresponding to the abnormal time in a time period,/>Representing the number of time periods,/>。
For ease of understanding, the duration of the time period is one day, as exemplified herein for simplicity. Number of time periods. If 8 months No. 2 8:00 photovoltaic data are abnormal data, and 8 months No. 2, 8:00 is the abnormal time. 8 months 1 and 8:00 photovoltaic data, 8 month No. 3 8:00, and calculate 8 month No. 1 8: photovoltaic data of 00 and 8 month No. 3 8: average of the photovoltaic data of 00. The average value was used to replace 8 of 8 months No. 2: the photovoltaic data of 00a is obtained,
Similarly, when the missing data is subjected to the missing data completion processing, missing time corresponding to the missing data in the current time period is obtained, for each missing time, a time corresponding to the missing time is determined from at least one time period adjacent to the current time period, and the average value of the photovoltaic data at each corresponding time is calculated. The average value is used as the complement value of the missing data, thereby realizing the complement of the missing data.
According to the embodiment of the invention, the missing data is accurately complemented and the abnormal data is corrected by carrying out the missing data complement processing and the abnormal data correction processing on the photovoltaic data, so that the influence of the missing data and the abnormal data in the photovoltaic data on the prediction accuracy of the photovoltaic power generation power is avoided, and the prediction accuracy of the photovoltaic power generation power is further improved.
It should be understood that the sequence number of each step in the foregoing embodiment does not mean that the execution sequence of each process should be determined by the function and the internal logic, and should not limit the implementation process of the embodiment of the present invention.
The following are device embodiments of the invention, for details not described in detail therein, reference may be made to the corresponding method embodiments described above.
Fig. 6 is a schematic structural diagram of a distributed photovoltaic short-term power prediction apparatus according to an embodiment of the present invention, and for convenience of explanation, only a portion related to the embodiment of the present invention is shown, which is described in detail below:
As shown in fig. 6, the short-term power prediction device 6 of the distributed photovoltaic includes: an acquisition module 61 and a prediction module 62.
An obtaining module 61, configured to obtain photovoltaic data at different moments in time at least one time period before the predicted time period;
The prediction module 62 is configured to input the photovoltaic data into a pre-trained power prediction model, so as to obtain photovoltaic power generation power output by the power prediction model at different moments in a prediction time period;
The power prediction model is obtained by training based on photovoltaic data at different moments in at least one time period before a plurality of time periods and photovoltaic power generation power at different moments in a corresponding time period, and is a negative feedback neural network model which comprises a plurality of processing units, and hidden states output by the processing units in each time step are correspondingly fed back to the processing units in the previous time step.
In one possible implementation, the negative feedback neural network model is a modified LSTM neural network model, and the processing unit is an LSTM unit;
In the improved LSTM neural network model, the hidden state output by each LSTM unit in each time step is correspondingly fed back to the input end of each LSTM unit in the previous time step;
The improved LSTM neural network model further comprises a plurality of convolution layers, wherein the input end of each convolution layer is correspondingly connected to the output end of the last LSTM unit under each time step, and the output end of each convolution layer serves as the output end of the improved LSTM neural network model.
In one possible implementation, the prediction module 62 is further configured to:
carrying out data cleaning on the photovoltaic data; the data cleaning comprises missing data complement processing and abnormal data correction processing;
the prediction module 62 is specifically configured to:
and inputting the photovoltaic data subjected to data cleaning into a power prediction model to obtain photovoltaic power generation power at different moments in a prediction time period output by the power prediction model.
In one possible implementation, the prediction module 62 is specifically configured to:
Respectively inputting the photovoltaic data into a pre-trained abnormal time period identification model, and correspondingly determining an abnormal time period; the abnormal time period identification model is obtained based on photovoltaic data at different moments in different time periods and abnormal labels in corresponding time periods through training;
And correspondingly determining the abnormal data in the photovoltaic data in the abnormal time period based on the photovoltaic data at each time in the abnormal time period, and correcting the abnormal data.
In one possible implementation, the prediction module 62 is specifically configured to:
sliding a sliding window with a preset length according to a preset step along a time sequence data sequence formed by photovoltaic data at each moment in an abnormal time period;
calculating window average values corresponding to the sliding window at the current moment and window average values corresponding to the sliding window at the last moment in real time, if the difference value between the window average value corresponding to the current moment and the window average value corresponding to the last moment is larger than a first preset difference value, shortening the length of the sliding window, and eliminating data corresponding to the tail end of the sliding window at the current moment to obtain a new sliding window;
The step of calculating the window average value corresponding to the current moment of the sliding window and the window average value corresponding to the last moment of the sliding window in real time is carried out in a skip mode, when the difference value between the window average value corresponding to the current moment and the window average value corresponding to the last moment is smaller than or equal to a second preset difference value, the data which are removed last time are determined to be abnormal data, the sliding window continues to slide along the time sequence data, and the step of calculating the window average value corresponding to the current moment of the sliding window and the window average value corresponding to the last moment of the sliding window in real time is carried out in a skip mode until the sliding window traverses the sequence data, and all abnormal data in the photovoltaic data in the abnormal time period are obtained; the second preset difference value is smaller than the first preset difference value.
In one possible implementation, the photovoltaic data includes: photovoltaic power generation power, corrected grid-connected point voltage and corrected irradiance;
The obtaining module 61 is specifically configured to:
Acquiring photovoltaic power generation power, grid-connected point voltage and irradiance at different moments in at least one time period before a predicted time period;
Respectively obtaining regression coefficients corresponding to the grid-connected point voltage and the irradiance; the regression coefficient is used for representing the influence degree of photovoltaic power generation power;
and correspondingly determining the corrected grid-connected point voltage and the corrected irradiance based on the grid-connected point voltage, the irradiance and the corresponding regression coefficient.
In one possible implementation, the obtaining module 61 is specifically configured to:
the regression coefficients corresponding to the grid-connected point voltage and the irradiance are respectively obtained, and the regression coefficients comprise:
According to Constructing a regression equation;
Wherein, Representing regression coefficient corresponding to grid-connected point voltage,/>Representing grid-tie point voltage,/>Representing regression coefficients corresponding to irradiance,/>Representing irradiance,/>Representing photovoltaic power generation;
And constructing a loss function, and solving a regression equation according to the principle of minimum loss function to obtain regression coefficients corresponding to grid-connected point voltage and irradiance.
The prediction module 62 in the embodiment of the present invention inputs the photovoltaic data at different moments in time at least one time period before the predicted time period into the power prediction model, so as to obtain the photovoltaic power generation power at different moments in time in the predicted time period. The power prediction model is a negative feedback neural network model, the negative feedback neural network model comprises a plurality of processing units, and the hidden state output by each processing unit in each time step is correspondingly fed back to each processing unit in the previous time step. The hidden states output by the processing units in each time step are correspondingly fed back to the processing units in the previous time step, so that the processing units in the current time step can process data based on the hidden states in the next time step, the time sequence dependency relationship in the photovoltaic data in different moments can be captured more effectively, and the prediction accuracy of the photovoltaic power generation power is improved.
Fig. 7 is a schematic diagram of a terminal according to an embodiment of the present invention. As shown in fig. 7, the terminal 7 of this embodiment includes: a processor 70, a memory 71, and a computer program 72 stored in the memory 71 and executable on the processor 70. The processor 70, when executing the computer program 72, implements the steps of the above-described embodiments of the short-term power prediction method for each distributed photovoltaic, such as steps 101 through 102 shown in fig. 1. Or the processor 70, when executing the computer program 72, performs the functions of the modules/units of the apparatus embodiments described above, such as the functions of the modules 61-62 shown in fig. 6.
By way of example, the computer program 72 may be partitioned into one or more modules/units that are stored in the memory 71 and executed by the processor 70 to complete the present invention. The one or more modules/units may be a series of computer program instruction segments capable of performing the specified functions, which instruction segments are used for describing the execution of the computer program 72 in the terminal 7. For example, the computer program 72 may be divided into modules 61 to 62 shown in fig. 6.
The terminal 7 may be a computing device such as a desktop computer, a notebook computer, a palm computer, a cloud server, etc. The terminal 7 may include, but is not limited to, a processor 70, a memory 71. It will be appreciated by those skilled in the art that fig. 7 is merely an example of the terminal 7 and is not limiting of the terminal 7, and may include more or fewer components than shown, or may combine some components, or different components, e.g., the terminal may further include input and output devices, network access devices, buses, etc.
The Processor 70 may be a central processing unit (Central Processing Unit, CPU), or may be another general purpose Processor, a digital signal Processor (DIGITAL SIGNAL Processor, DSP), an Application SPECIFIC INTEGRATED Circuit (ASIC), a Field-Programmable gate array (Field-Programmable GATE ARRAY, FPGA) or other Programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory 71 may be an internal storage unit of the terminal 7, such as a hard disk or a memory of the terminal 7. The memory 71 may be an external storage device of the terminal 7, such as a plug-in hard disk, a smart memory card (SMART MEDIA CARD, SMC), a Secure Digital (SD) card, a flash memory card (FLASH CARD) or the like, which are provided on the terminal 7. Further, the memory 71 may also include both an internal storage unit and an external storage device of the terminal 7. The memory 71 is used for storing the computer program as well as other programs and data required by the terminal. The memory 71 may also be used for temporarily storing data that has been output or is to be output.
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-described division of the functional units and modules is illustrated, and in practical application, the above-described functional distribution may be performed by different functional units and modules according to needs, i.e. the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-described functions. The functional units and modules in the embodiment may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit, where the integrated units may be implemented in a form of hardware or a form of a software functional unit. In addition, the specific names of the functional units and modules are only for distinguishing from each other, and are not used for limiting the protection scope of the present application. The specific working process of the units and modules in the above system may refer to the corresponding process in the foregoing method embodiment, which is not described herein again.
In the foregoing embodiments, the descriptions of the embodiments are emphasized, and in part, not described or illustrated in any particular embodiment, reference is made to the related descriptions of other embodiments.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
In the embodiments provided in the present invention, it should be understood that the disclosed apparatus/terminal and method may be implemented in other manners. For example, the apparatus/terminal embodiments described above are merely illustrative, e.g., the division of the modules or units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection via interfaces, devices or units, which may be in electrical, mechanical or other forms.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.
The integrated modules/units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the present invention may also be implemented by implementing all or part of the above-described embodiment method, or by implementing relevant hardware by a computer program, where the computer program may be stored in a computer readable storage medium, and the computer program may be executed by a processor to implement the steps of the above-described embodiments of the distributed photovoltaic short-term power prediction method. Wherein the computer program comprises computer program code which may be in source code form, object code form, executable file or some intermediate form etc. The computer readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), an electrical carrier signal, a telecommunications signal, a software distribution medium, and so forth. The above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention, and are intended to be included in the scope of the present invention.

Claims (10)

1. A method of short-term power prediction for distributed photovoltaics, comprising:
Acquiring photovoltaic data at different moments in at least one time period before a predicted time period;
Inputting the photovoltaic data into a pre-trained power prediction model to obtain photovoltaic power generation power at different moments in the prediction time period output by the power prediction model;
the power prediction model is obtained by training based on photovoltaic data at different moments in at least one time period before a plurality of time periods and photovoltaic power generation power at different moments in a corresponding time period, and is a negative feedback neural network model which comprises a plurality of processing units, and hidden states output by the processing units in each time step are correspondingly fed back to the processing units in the previous time step.
2. The method of claim 1, wherein the negative feedback neural network model is a modified LSTM neural network model and the processing unit is an LSTM unit;
in the improved LSTM neural network model, the hidden state output by each LSTM unit in each time step is correspondingly fed back to the input end of each LSTM unit in the previous time step;
the improved LSTM neural network model further comprises a plurality of convolution layers, wherein the input end of each convolution layer is correspondingly connected to the output end of the last-stage LSTM unit under each time step, and the output end of each convolution layer serves as the output end of the improved LSTM neural network model.
3. The method for short-term power prediction of distributed photovoltaic according to claim 1 or 2, characterized in that before the photovoltaic data is input into a pre-trained power prediction model to obtain photovoltaic power generation power at different moments in the prediction time period output by the power prediction model, the method further comprises:
carrying out data cleaning on the photovoltaic data; the data cleaning comprises missing data complement processing and abnormal data correction processing;
inputting the photovoltaic data into a pre-trained power prediction model to obtain photovoltaic power generation power at different moments in the prediction time period output by the power prediction model, wherein the method comprises the following steps:
and inputting the photovoltaic data subjected to data cleaning into the power prediction model to obtain photovoltaic power generation power at different moments in the prediction time period output by the power prediction model.
4. A method of short term power prediction for distributed photovoltaics according to claim 3, wherein performing anomaly data correction processing on the photovoltaic data comprises:
Respectively inputting the photovoltaic data into a pre-trained abnormal time period identification model, and correspondingly determining an abnormal time period; the abnormal time period identification model is obtained based on photovoltaic data at different moments in different time periods and abnormal labels in corresponding time periods through training;
And correspondingly determining abnormal data in the photovoltaic data in the abnormal time period based on the photovoltaic data at each time in the abnormal time period, and correcting the abnormal data.
5. The method for short-term power prediction of distributed photovoltaic according to claim 4, wherein the correspondingly determining the abnormal data in the photovoltaic data in the abnormal time period based on the photovoltaic data at each time point in the abnormal time period comprises:
sliding a sliding window with a preset length along a time sequence data sequence formed by the photovoltaic data at each time in the abnormal time period according to a preset step;
Calculating a window average value corresponding to the sliding window at the current moment and a window average value corresponding to the sliding window at the last moment in real time, if the difference value between the window average value corresponding to the current moment and the window average value corresponding to the last moment is larger than a first preset difference value, shortening the length of the sliding window, and eliminating data corresponding to the tail end of the sliding window at the current moment to obtain a new sliding window;
The step of calculating the window average value corresponding to the sliding window at the current moment and the window average value corresponding to the sliding window at the last moment in real time is carried out in a skip mode, when the difference value between the window average value corresponding to the current moment and the window average value corresponding to the last moment is smaller than or equal to a second preset difference value, the data which are removed last time are determined to be abnormal data, the sliding window continues to slide along the time sequence, and the step of calculating the window average value corresponding to the sliding window at the current moment and the window average value corresponding to the sliding window at the last moment in real time is carried out in a skip mode until the sliding window traverses the time sequence, and all abnormal data in the photovoltaic data in the abnormal time period are obtained; wherein the second preset difference is smaller than the first preset difference.
6. The method of short term power prediction of distributed photovoltaic according to claim 1 or 2, characterized in that the photovoltaic data comprises: photovoltaic power generation power, corrected grid-connected point voltage and corrected irradiance;
The obtaining photovoltaic data at different moments in at least one time period before the predicted time period comprises the following steps:
Acquiring photovoltaic power generation power, grid-connected point voltage and irradiance at different moments in at least one time period before a predicted time period;
respectively obtaining the voltage of the grid-connected point and the regression coefficient corresponding to the irradiance; the regression coefficient is used for representing the influence degree of the power generated by the photovoltaic power generation;
And correspondingly determining the corrected grid-connected point voltage and the corrected irradiance based on the grid-connected point voltage, the irradiance and the corresponding regression coefficient.
7. The method for short-term power prediction of a distributed photovoltaic according to claim 6, wherein the obtaining regression coefficients corresponding to the grid-tie voltage and the irradiance, respectively, comprises:
According to Constructing a regression equation;
Wherein, Representing regression coefficient corresponding to grid-connected point voltage,/>Representing grid-tie point voltage,/>Representing regression coefficients corresponding to irradiance,/>Representing irradiance,/>Representing photovoltaic power generation;
And constructing a loss function, and solving the regression equation according to the principle of minimum loss function to obtain the grid-connected point voltage and the regression coefficient corresponding to the irradiance.
8. A distributed photovoltaic short-term power prediction apparatus, comprising:
the acquisition module is used for acquiring photovoltaic data at different moments in at least one time period before the predicted time period;
The prediction module is used for inputting the photovoltaic data into a pre-trained power prediction model to obtain photovoltaic power generation power at different moments in the prediction time period output by the power prediction model;
the power prediction model is obtained by training based on photovoltaic data at different moments in at least one time period before a plurality of time periods and photovoltaic power generation power at different moments in a corresponding time period, and is a negative feedback neural network model which comprises a plurality of processing units, and hidden states output by the processing units in each time step are correspondingly fed back to the processing units in the previous time step.
9. A terminal comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the steps of the distributed photovoltaic short-term power prediction method according to any of the preceding claims 1 to 7 when the computer program is executed.
10. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the steps of the method of short-term power prediction of distributed photovoltaics according to any one of the preceding claims 1 to 7.
CN202410275636.0A 2024-03-12 2024-03-12 Short-term power prediction method, device, terminal and storage medium for distributed photovoltaic Active CN117878930B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202410275636.0A CN117878930B (en) 2024-03-12 2024-03-12 Short-term power prediction method, device, terminal and storage medium for distributed photovoltaic

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202410275636.0A CN117878930B (en) 2024-03-12 2024-03-12 Short-term power prediction method, device, terminal and storage medium for distributed photovoltaic

Publications (2)

Publication Number Publication Date
CN117878930A CN117878930A (en) 2024-04-12
CN117878930B true CN117878930B (en) 2024-05-28

Family

ID=90579580

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202410275636.0A Active CN117878930B (en) 2024-03-12 2024-03-12 Short-term power prediction method, device, terminal and storage medium for distributed photovoltaic

Country Status (1)

Country Link
CN (1) CN117878930B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111506868A (en) * 2020-04-07 2020-08-07 河海大学 Ultrashort-term wind speed prediction method based on HHT weight optimization
CN112461537A (en) * 2020-10-16 2021-03-09 浙江工业大学 Wind power gear box state monitoring method based on long-time neural network and automatic coding machine
CN114091766A (en) * 2021-11-24 2022-02-25 东北电力大学 Space load prediction method based on CEEMDAN-LSTM
CN114787837A (en) * 2019-11-26 2022-07-22 巴斯夫欧洲公司 Predicting industrial aging process using machine learning method
CN116435998A (en) * 2023-04-17 2023-07-14 石家庄科林电气股份有限公司 Prediction method of photovoltaic power generation power

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114787837A (en) * 2019-11-26 2022-07-22 巴斯夫欧洲公司 Predicting industrial aging process using machine learning method
CN111506868A (en) * 2020-04-07 2020-08-07 河海大学 Ultrashort-term wind speed prediction method based on HHT weight optimization
CN112461537A (en) * 2020-10-16 2021-03-09 浙江工业大学 Wind power gear box state monitoring method based on long-time neural network and automatic coding machine
CN114091766A (en) * 2021-11-24 2022-02-25 东北电力大学 Space load prediction method based on CEEMDAN-LSTM
CN116435998A (en) * 2023-04-17 2023-07-14 石家庄科林电气股份有限公司 Prediction method of photovoltaic power generation power

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
Improved day-ahead predictions of load and renewable generation by optimally exploiting multi-scale dependencies Jean-François Toubeau et al.."Improved day-ahead predictions of load and renewable generation by optimally exploiting multi-scale dependencies".《2017 IEEE Innovative Smart Grid Technologies - Asia (ISGT-Asia)》.2018,第1-5页. *

Also Published As

Publication number Publication date
CN117878930A (en) 2024-04-12

Similar Documents

Publication Publication Date Title
Li et al. An integrated missing-data tolerant model for probabilistic PV power generation forecasting
CN116011686B (en) Charging shed photovoltaic power generation reserve prediction method based on multi-data fusion
EP3195438B1 (en) System, method and apparatus for determining parameter settings for a power generation system and a tangible computer readable medium
CN111625399A (en) Method and system for recovering metering data
CN115498623A (en) Energy storage configuration optimization method, device, equipment and storage medium for multiple micro-grids
CN116799796A (en) Photovoltaic power generation power prediction method, device, equipment and medium
CN115436812B (en) Electrochemical model parameter identification method and system based on splicing factors
CN116435998A (en) Prediction method of photovoltaic power generation power
CN111090926A (en) Photovoltaic power curve modeling method and device, computer equipment and storage medium
CN115618922A (en) Photovoltaic power prediction method and device, photovoltaic power generation system and storage medium
WO2023136774A2 (en) Method and apparatus for determining energy storage charge and discharge scheme, device, and storage medium
CN115796338A (en) Photovoltaic power generation power prediction model construction and photovoltaic power generation power prediction method
CN117748482B (en) Method and device for predicting generated power, electronic equipment and storage medium
CN117878930B (en) Short-term power prediction method, device, terminal and storage medium for distributed photovoltaic
CN118174297A (en) Photovoltaic power prediction method, system, terminal and medium based on multi-model fusion
CN116706903B (en) Photovoltaic power generation amount prediction method, device, equipment and medium
CN111950752A (en) Photovoltaic power station generating capacity prediction method, device and system and storage medium thereof
CN116992274B (en) Short-term wind speed prediction method and system based on improved principal component regression model
CN114239945A (en) Short-term power load prediction method, device, equipment and storage medium
CN113780640A (en) TCN-Attention-based solar radiation prediction method
CN116307632B (en) Hydropower station economic load distribution method and device, electronic equipment and storage medium
CN116739130A (en) Multi-time scale load prediction method of TCN-BiLSTM network
CN111126645A (en) Wind power prediction algorithm based on data mining technology and improved support vector machine
CN115689067A (en) Solar irradiance prediction method, device and storage medium
CN112613674A (en) Medium-and-long-term wind power generation capacity prediction method and device, electronic equipment and storage medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant