CN113536655B - Heliostat artificial intelligence deviation rectifying method, heliostat artificial intelligence deviation rectifying device, electronic equipment and storage medium - Google Patents
Heliostat artificial intelligence deviation rectifying method, heliostat artificial intelligence deviation rectifying device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN113536655B CN113536655B CN202110372862.7A CN202110372862A CN113536655B CN 113536655 B CN113536655 B CN 113536655B CN 202110372862 A CN202110372862 A CN 202110372862A CN 113536655 B CN113536655 B CN 113536655B
- Authority
- CN
- China
- Prior art keywords
- data
- heliostat
- sample
- lstm network
- lstm
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 27
- 238000012549 training Methods 0.000 claims abstract description 50
- 230000015654 memory Effects 0.000 claims abstract description 33
- 238000004364 calculation method Methods 0.000 claims abstract description 20
- 238000004891 communication Methods 0.000 claims description 17
- 238000007781 pre-processing Methods 0.000 claims description 17
- 238000012937 correction Methods 0.000 claims description 15
- 238000004590 computer program Methods 0.000 claims description 11
- 238000005457 optimization Methods 0.000 claims description 11
- 239000011159 matrix material Substances 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 9
- 238000003491 array Methods 0.000 claims description 4
- 238000012821 model calculation Methods 0.000 claims description 4
- 238000010801 machine learning Methods 0.000 abstract description 2
- 230000008569 process Effects 0.000 description 10
- 230000001413 cellular effect Effects 0.000 description 5
- 230000001360 synchronised effect Effects 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000007787 long-term memory Effects 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 238000009434 installation Methods 0.000 description 2
- 238000010248 power generation Methods 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 230000003749 cleanliness Effects 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 230000020169 heat generation Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
-
- 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
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02S—GENERATION OF ELECTRIC POWER BY CONVERSION OF INFRARED RADIATION, VISIBLE LIGHT OR ULTRAVIOLET LIGHT, e.g. USING PHOTOVOLTAIC [PV] MODULES
- H02S20/00—Supporting structures for PV modules
- H02S20/30—Supporting structures being movable or adjustable, e.g. for angle adjustment
- H02S20/32—Supporting structures being movable or adjustable, e.g. for angle adjustment specially adapted for solar tracking
-
- 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
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E10/00—Energy generation through renewable energy sources
- Y02E10/50—Photovoltaic [PV] energy
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The disclosure relates to a heliostat artificial intelligence deviation rectifying method, a heliostat artificial intelligence deviation rectifying device, electronic equipment and a storage medium, wherein the method comprises the following steps: reading sample data; training an LSTM network according to the sample data to obtain an LSTM network model; inputting the sequential data of the heliostat into the LSTM network model, and outputting the deviation correcting data of the heliostat after calculation by the LSTM network model; and correcting the daily tracking operation parameters of the heliostat by using the deviation correcting data of the heliostat. The method comprises the step of learning sample data of the heliostat efficiently by using a long-short term memory artificial intelligence machine learning method, so that the heliostat can track with higher precision.
Description
Technical Field
The disclosure relates to the technical field of solar energy application, and more particularly to a heliostat artificial intelligence deviation rectifying method, device, electronic equipment and storage medium.
Background
Solar energy is increasingly utilized in human life due to its advantages of large total energy, wide distribution, cleanliness, no pollution and high economy. In the process of utilizing solar energy, whether in the form of solar photovoltaic panels, solar concentrating mirrors or other forms, sunlight tracking technology is increasingly used in order to receive more solar radiation for utilization. Especially in a large-scale solar power station, the sunlight tracking system can be used for improving the generating efficiency of the solar power generation device so as to improve the yield.
For example, when the sunlight tracking technology is applied to solar photovoltaic power generation, a control system of the sunlight tracking technology calculates according to a tracking formula obtained by a mathematical principle to control a solar photovoltaic cell panel to track so as to keep the photovoltaic cell panel facing the sun at any time and enable light rays of sunlight to vertically irradiate the photovoltaic cell panel at any time; when the solar energy heat collection system is applied to the tower type solar energy heat generation system, the control system calculates according to a tracking formula obtained by a mathematical principle to control the movement of the reflector in real time so as to reflect sunlight to the heat collection tower at any time.
After various measurements and initial installation and debugging are carried out on the photovoltaic cell panel and the reflector, fixed input state parameters for carrying out mathematical principle calculation, such as longitude and latitude parameters of the photovoltaic cell panel and the reflector, an inclination angle and an altitude of an installation base and the like, are obtained, and at the initial stage, the precision of the parameters is high, and the precision of equipment tracking is also high. However, as the equipment operates, due to the influence of wind, vibration, geographical conditions and mechanical stress, these parameters change gradually, and errors accumulate gradually, eventually resulting in a significant drop in tracking accuracy.
In this regard, the prior art provides various methods for obtaining tracking errors, i.e., offsets, tracked by the individual photovoltaic panels or mirrors at different times. In addition, the prior art also discloses a method for fitting a tracking error curve according to the obtained discrete tracking error, calculating the current tracking error in real time by using the tracking error curve, and correcting the current tracking parameter according to the current tracking error curve to perform accurate tracking.
The method for calculating the current tracking error by fitting the tracking error curve according to the historical deviation value is low in speed, high in consumed calculation resource and low in accuracy of the fitted result.
Disclosure of Invention
In order to solve the technical problems or at least partially solve the technical problems, the disclosure provides a heliostat artificial intelligence deviation rectifying method, device, electronic equipment and storage medium.
In a first aspect, the present disclosure relates to a heliostat artificial intelligence deviation rectification method, which includes:
reading sample data;
training an LSTM network according to the sample data to obtain an LSTM network model;
inputting the sequence data of the heliostat into the LSTM network model, and outputting the deviation rectifying data of the heliostat after calculation by the LSTM network model; and
and correcting the daily tracking operation parameters of the heliostat by using the deviation correcting data of the heliostat.
In some embodiments, training the LSTM network according to the sample data to obtain an LSTM network model comprises:
preprocessing the sample data, wherein the sample data comprises sample sequence data and sample deviation correcting data;
in the LSTM network, connecting the preprocessed sample sequence data serving as input with a first LSTM hidden layer with a first number of units, connecting the first LSTM hidden layer with a first discarding layer, and connecting the first discarding layer with the sample deviation rectifying data serving as output through a full connection layer; and
iteratively training the LSTM network using an optimizer to obtain the LSTM network model.
In some embodiments, training the LSTM network according to the sample data to obtain an LSTM network model comprises:
preprocessing the sample data, wherein the sample data comprises sample sequence data and sample deviation correcting data;
in the LSTM network, the preprocessed sample sequence data is used as input and connected with a first LSTM hidden layer with a first quantity unit, the first LSTM hidden layer is then connected with a first discarding layer, the first discarding layer is then connected with a second LSTM hidden layer with a second quantity unit, the second LSTM hidden layer is then connected with a second discarding layer, and the second discarding layer is connected with the sample deviation rectifying data used as output through a full connection layer; and
and performing iterative optimization training on the LSTM network by using an optimizer to obtain the LSTM network model.
In certain implementations, the pre-processing the sample data comprises:
splicing all groups of data in the same day of the same heliostat into a matrix;
combining the matrixes by using the cellular arrays to obtain sample cellular arrays; and
and processing the sample cell array to ensure that the normalized distribution of the sample cell array is between 0 and 1, and storing the mean value and the standard deviation.
In some embodiments, the iteratively training the LSTM network using the optimizer comprises: and performing iterative optimization training on the LSTM network by adopting a gradient descent optimizer.
In certain embodiments, the first number is less than the second number, and the drop rate of the first drop layer is less than the drop rate of the second drop layer.
In certain implementations, the reading the sample data comprises: at least 6 sets of sample sequence data for each heliostat scattered over each day are read, and irrelevant partial data are ignored.
In a second aspect, the present disclosure relates to a heliostat artificial intelligence deviation rectification device, comprising:
a data reading module configured to read sample data;
the model training module is configured to train the LSTM network according to the sample data to obtain an LSTM network model;
the model calculation module is configured to input the sequence data of the heliostat into the LSTM network model, and output the deviation correction data of the heliostat after calculation of the LSTM network model; and
a result application module configured to use the deskew data for the heliostat to correct a daily tracking operational parameter of the heliostat.
In a third aspect, the present disclosure relates to an electronic device comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus;
the memory is used for storing computer programs;
the processor is configured to implement any heliostat artificial intelligence deskewing method according to the first aspect of the disclosure when executing the computer program stored on the memory.
In a fourth aspect, the present disclosure relates to a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements any heliostat artificial intelligence deskewing method according to the first aspect of the present disclosure.
According to the method, the method comprising long-term and short-term memory artificial intelligence machine learning is used for efficiently learning the sample data of the heliostat so as to obtain a model for calculating deviation correction data of the heliostat, and therefore more accurate tracking operation parameters are obtained, and the heliostat is tracked with higher precision.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the disclosure.
In order to more clearly illustrate the embodiments or technical solutions in the prior art of the present disclosure, the drawings used in the embodiments or technical solutions in the prior art description will be briefly described below, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without inventive labor.
FIG. 1 is a schematic flow chart diagram illustrating a heliostat artificial intelligence deviation rectification method according to an embodiment of the disclosure;
FIG. 2 is a schematic flow chart illustrating training of an LSTM network according to sample data in an artificial intelligence deviation rectification method for a heliostat according to an embodiment of the disclosure;
FIG. 3 is a schematic flow chart illustrating training of an LSTM network according to sample data in a heliostat artificial intelligence deviation rectification method according to another embodiment of the disclosure;
FIG. 4 is a schematic flow chart illustrating sample data preprocessing in a heliostat artificial intelligence deviation rectification method according to an embodiment of the disclosure;
FIG. 5 shows a schematic diagram of a heliostat artificial intelligence deviation rectification device, in accordance with an embodiment of the present disclosure; and
fig. 6 shows a schematic structural diagram of an electronic device according to an embodiment of the present disclosure.
Detailed Description
To make the objects, technical solutions and advantages of the embodiments of the present disclosure more clear, the technical solutions of the embodiments of the present disclosure will be described clearly and completely with reference to the drawings in the embodiments of the present disclosure, and it is obvious that the described embodiments are some embodiments of the present disclosure, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments disclosed herein without making any creative effort, shall fall within the protection scope of the present disclosure.
The heliostat referred to in the present disclosure may be a reflector which tracks the moving track of the sun in real time and reflects the sunlight to a specific position by using mirror reflection, which is common in a concentrated solar thermal power station, or a photovoltaic cell panel which tracks the sunlight in real time by a solar photovoltaic power station tracked in dual axes to face the incident direction of the sunlight, as long as the heliostat needs to form certain specific angles with the sunlight in real time to use the energy of the sunlight. The heliostat used in the embodiment of the present disclosure is exemplified by a heliostat used in a tower-type concentrated solar thermal power plant, and the heliostat adjusts the posture of a mirror surface in an azimuth-elevation manner to track sunlight and reflects the sunlight to a concentrated heat collector fixed at a certain position.
Fig. 1 shows a schematic flow chart of an artificial intelligence deviation rectifying method for a heliostat according to an embodiment of the present disclosure. The heliostat artificial intelligence deviation rectifying method can comprise the following steps:
s11, reading sample data;
s13, training an LSTM network according to the sample data to obtain an LSTM network model;
s15, inputting the sequential data of the heliostat into an LSTM network model, and outputting the deviation correcting data of the heliostat after calculation by the LSTM network model; and
and S17, correcting the daily tracking operation parameters of the heliostat by using the deviation correcting data of the heliostat.
In the embodiment of the disclosure, because the heliostat tracks in a dual-axis manner of pitch plus azimuth, the tracking operation parameters of the heliostat include the azimuth angle and the pitch angle of the heliostat. The operation parameters may also include other parameters that define the attitude of the heliostat according to the operation tracking mode of the heliostat, which is not limited by the present disclosure.
The sample data of the heliostat in S11 is sample data used for training the model, which may include data reflecting deviation amounts of the heliostat at different time instants, status position data of the heliostat itself, other relevant data of the heliostat field, sample deviation correction data obtained by calculation, and other data related to training. Wherein all data except the sample deskew data may be collectively referred to as sample sequence data. In the embodiment of the present disclosure, the data about the deviation amount is obtained by recording an initial operating parameter (at this time, although an instruction for controlling the heliostat to irradiate the light spot on the center of the light target is issued, the light spot will deviate relative to the center of the light target due to the existence of a tracking error) for controlling the heliostat to irradiate the light spot on the center of a certain light target, and a corrected operating parameter when the light spot is actually irradiated on the center of the light target after calibration; and the sample rectification data is determined by performing regression operation according to the data related to the deviation amount and other related data of the equipment and the mirror field. The rectification data can be used for adjusting a position-pitch tracking formula of the heliostat, so that more accurate tracking operation parameters can be obtained through the adjusted position-pitch tracking formula to perform more accurate tracking operation. The data reflecting the deviation of the heliostat at different time can be obtained by other methods, such as directly observing the state of the heliostat by using a camera or installing various sensors on the heliostat, as long as the obtained data can reflect various differences between the operating parameters before and after the calibration of the heliostat.
The LSTM (Long-Short Term Memory) Network in S13 is a Long-Short Term Memory Neural Network, which is one of Recurrent Neural Networks (RNN). The general RNN is greatly influenced by short-term memory, and important information in the early stage can be gradually omitted when the general RNN is learned to the later stage. The LSTM incorporates a mechanism for adjusting the flow of information, and can memorize necessary data and forget useless data. The sample trained LSTM network may be saved as an LSTM network model for later use.
In S15, the sequence data including the deviation characteristic acquired when the heliostat is currently operating is input as input data to the LSTM network model stored after training, and the rectification data output from the LSTM network model can be acquired. The deviation correcting data is the deviation correcting data corresponding to the sequence data of the heliostat during the current operation, namely the deviation correcting data can be used for correcting the deviation of the heliostat during the current operation with higher precision.
In S17, the deviation correcting data is used for adjusting the orientation-pitching tracking formula of the heliostat, so that the real-time tracking operation parameters of the heliostat are corrected, the heliostat tracking effect with higher precision is obtained, and the efficiency of a power plant is ensured.
Therefore, the LSTM network is trained by using the sample data to obtain the LSTM network model which can be used for generating heliostat deviation correction data, the calculation amount in the process is small, the calculation speed is high, and the deviation correction data can be obtained more conveniently; the LSTM network can transmit related information along a long-chain sequence to perform long-term memory, and the trained LSTM network model has higher accuracy of an output result due to the long-term memory of useful information in sample data when in use.
Fig. 2 shows a schematic flow chart of training an LSTM network according to sample data in the heliostat artificial intelligence deviation rectification method according to an embodiment of the present disclosure. Training the LSTM network according to the sample data in S13, and obtaining the LSTM network model may include:
s131, preprocessing sample data, wherein the sample data comprises sample sequence data and sample deviation correcting data;
s133a, in LSTM network, connecting preprocessed sample sequence data as input with a first LSTM hidden layer with a first number of units, the first LSTM hidden layer is connected with a first discarding layer, and the first discarding layer is connected with output sample deviation correcting data through a full connection layer; and
and S135, performing iterative optimization training on the LSTM network by using an optimizer to obtain an LSTM network model.
In S131, the sample data is preprocessed to be processed into a data format suitable for and capable of being directly input to the LSTM network for learning.
In S133a, the preprocessed sample data is connected as input to a first LSTM hidden layer with 100 cells, which is then connected to a first drop layer (drop layer) to regularize the parameters in the network using the drop layer, where the drop rate of the first drop layer is 0.4. The first number is not limited to 100, the discarding rate of the first discarding layer is not limited to 0.4, and the first number may be selected by comprehensive consideration of the implementer according to the dimensions of the sample data, the number of samples, the consumption of the computing resources, and other factors. The effect of using the discard layer is to prevent over-fitting of the data so that the finally obtained LSTM network model can output accurate results.
And in S133a, the first discarding layer is connected to the sample correction data as an output through the full connection layer, and the full connection layer is used to convert the nonlinear calculation into a linear calculation, so as to output a result.
In S135, since the input and output of the LSTM network are determined, the LSTM network may be iteratively optimally trained by the optimizer to optimize various parameters in the neural network and learn the characteristics of the sample data.
FIG. 3 shows a schematic flow chart of training an LSTM network according to sample data in a heliostat artificial intelligence deviation rectification method according to another embodiment of the disclosure. Training the LSTM network according to the sample data in S13, obtaining an LSTM network model may include:
s131, preprocessing sample data, wherein the sample data comprises sample sequence data and sample deviation correcting data;
s133b, in LSTM network, connecting preprocessed sample sequence data as input with a first LSTM hidden layer with a first quantity unit, the first LSTM hidden layer is connected with a first discarding layer, the first discarding layer is connected with a second LSTM hidden layer with a second quantity unit, the second LSTM hidden layer is connected with a second discarding layer, and the second discarding layer is connected with output sample deviation rectifying data through a full connection layer; and
and S135, performing iterative optimization training on the LSTM network by using an optimizer to obtain an LSTM network model.
In S133b, in addition to including the first LSTM hidden layer and the first discard layer as in S133a, the first discard layer is further connected to a second LSTM hidden layer having 200 cells, which is then connected to a second discard layer, wherein the discard rate of the second discard layer is 0.5. The second number is not limited to 200, the discarding rate of the second discarding layer is not limited to 0.5, and the second number may be selected by comprehensive consideration of the implementer according to the dimensions of the sample data, the number of samples, the consumption of the computing resources, and the like.
And in S133b, the second discarding layer is connected to the sample rectification data as an output through the full connection layer to convert the nonlinear calculation into a linear calculation for outputting the result.
The method flow as shown in fig. 3 introduces a second LSTM hidden layer and a second discarded layer with a second number of units in S133b, and the second number is different from the first number, and the discard rate of the first discarded layer is different from the discard rate of the second discarded layer. The two pairs of parameters are selected by considering various factors according to practical situations, but not limited to the first unit number of 100, the second unit number of 200, and the discard rate of the first discard layer of 0.4 and the discard rate of the second discard layer of 0.5, which are described in this disclosure, they may be the same.
In some embodiments, when the complexity of the project is high and the requirement on the accuracy of the result is high, a method of introducing a second LSTM hidden layer and a second discarded layer as shown in fig. 3 may be adopted, so that an LSTM network model with high accuracy can be obtained through further computational learning. And in some other embodiments, the similar steps may be repeated multiple times according to actual situations to obtain the LSTM network model with higher accuracy, which is not limited by the present disclosure.
Therefore, the hidden layer is trained by using the sample data, then the hidden layer and the discarded layer are repeatedly used for training, and under the condition that the total number of units of the hidden layer is equal, the training result is better than that of the training by directly using a single hidden layer. Therefore, the LSTM network model with higher accuracy can be obtained by relatively reducing the number of LSTM hidden layer units, and resources are saved.
In some embodiments, the first number is less than the second number and the drop rate of the first drop layer is less than the drop rate of the second drop layer, i.e. the number of cells of the first-in LSTM hidden layer is less than the number of cells of the second-in LSTM hidden layer and the drop rate of the first-in drop layer is less than the drop rate of the second-in drop layer. Therefore, firstly, a few LSTM hidden layer units are used for carrying out preliminary training on sample data, a small discarding rate is set, more useful information is reserved, and therefore a preliminarily formed LSTM network is obtained; and then more accurate training is carried out on the sample data by using more LSTM hidden layer units, and a larger discarding rate is set, so that a more accurate LSTM network is obtained. The structures of the units in the first hidden layer and the second hidden layer may be the same or different according to the characteristics of the items, and the disclosure does not limit this. By setting the number of LSTM hidden layers and the discarding rate of the discarded layers, an LSTM network model with higher accuracy can be obtained with less computing resources.
Fig. 4 shows a schematic flow chart of preprocessing sample data in the heliostat artificial intelligence deviation rectifying method according to an embodiment of the present disclosure. The preprocessing of the sample data in S131 may include:
s1311, splicing all groups of data of the same heliostat in the same day into a matrix;
s1313, obtaining a sample cell array by using the cell array combination matrix; and
s1315, processing the sample cell array to enable the sample cell array to be distributed in a normalized mode between 0 and 1, and storing the mean value and the standard deviation.
In S1311, sample data of the same heliostat on the same day is spliced into a matrix. In the embodiment of the present disclosure, there are approximately 10000 heliostats, and each heliostat may have data for multiple days, i.e., multiple sets of sample data, and there may be multiple sets of sample sequence data for each heliostat for each day. Each sample sequence data comprises 20 rows of parameters, and the parameters record data such as deviation amount of heliostat tracking, state parameters of the heliostat, related parameters of a heliostat field and the like; the daily sample correction data contains 4 columns of parameters which record the sample correction data which is obtained after regression calculation and used for adjusting the tracking formula to obtain the calibrated tracking operation parameters. Firstly, selecting 6 sample sequence data with 20-row parameters of the same heliostat on the same day for all heliostats and splicing the data into a 20 multiplied by 6 matrix; and the sample deviation correcting data only have 1 piece each day, and each 4 columns are spliced into a matrix of 1 x 4. The selected sample sequence data of the same heliostat on the same day is preferably at least 6 groups, so that enough useful information is provided for the LSTM network to learn, and the accuracy of the training result is ensured.
In S1313, all matrices are combined into a sample cell array, where the sample sequence data is combined into an array XTrain; the sample deskew data is combined into the array YTrain. The cellular array is a special data combination type, and internal elements of the cellular array can belong to different data types, so that the cellular array and the internal elements can be combined to form a whole to perform subsequent steps, and operation and calculation are facilitated.
In S1315, XTrain and YTrain are processed to be normalized to be distributed between 0 and 1, so as to be suitable for input into the LSTM network, and to accelerate the training process of the LSTM network. The mean and standard deviation are preserved during normal distribution to facilitate their reduction in subsequent processes.
In some embodiments, the optimizer used to train the LSTM network in S135 is a gradient descent optimizer, which has an initial learning rate of 0.01, and selects a piecewise descent learning rate, which has a piecewise descent factor of 0.5, and descends every 30 rounds, i.e., the learning rate x piecewise descent factor is updated to a new learning rate. During the training process, the minimum Batch Size (Batch Size) of each training is 10, and the maximum number of rounds is 80. The parameters such as the learning rate, the segment descent factor, the descent turn, and the like can be selected by the implementer in a comprehensive consideration according to the characteristics of the sample data, the consumption condition of the computing resources, and the like, and the values used in the embodiment are not limited. Because the gradient descent optimizer is adopted, the gradient calculation speed is high, the tolerance to noise is large, convergence is easy, and the training speed is high when the data set is large, so that the training speed and the training efficiency are improved.
In some embodiments, in addition to tuning optimization during training, reading data from the database and preprocessing may take a significant amount of time, taking all the grouped mirror queues by mirror number and date, and taking only 6 pieces of data per group modulo.
In some embodiments, two fully-connected layers may be included to allow better splicing with the output.
In some embodiments, some sample data may be reserved before training as test data to test the trained LSTM model after obtaining the model, and the mean square error of the output of the test data and the output of the model calculation and the accuracy within the fault tolerance range are compared, and based on this, the training process is optimized to obtain a more accurate model.
Fig. 5 shows an artificial intelligence deviation rectification device for a heliostat according to an embodiment of the disclosure, which includes:
a data reading module 501 configured to read sample data;
the model training module 503 is configured to train the LSTM network according to the sample data to obtain an LSTM network model;
the model calculation module 505 is configured to input the sequential data of the heliostat into the LSTM network model, and output the deviation correction data of the heliostat after calculation by the LSTM network model; and
the result application module 507 is configured to modify daily tracking operational parameters of the heliostat using deskew data for the heliostat.
In certain embodiments, the model training module 503 may include:
the system comprises a preprocessing module, a data processing module and a data processing module, wherein the preprocessing module is configured to preprocess sample data, and the sample data comprises sample sequence data and sample deviation correcting data;
the first connection module is configured to connect the preprocessed sample sequence data serving as input with a first LSTM hidden layer with a first number of units in an LSTM network, the first LSTM hidden layer is connected with a first discarding layer, and the first discarding layer is connected with the sample deviation correction data serving as output through a full connection layer; and
and the optimization module is configured to perform iterative optimization training on the LSTM network by using an optimizer so as to obtain an LSTM network model.
In certain embodiments, the model training module 503 may further include:
the system comprises a preprocessing module, a data processing module and a data processing module, wherein the preprocessing module is configured to preprocess sample data, and the sample data comprises sample sequence data and sample deviation correcting data;
a second connection module, configured to connect the preprocessed sample sequence data as input to a first LSTM hidden layer with a first number of units in an LSTM network, the first LSTM hidden layer is then connected to a first discarding layer, the first discarding layer is then connected to a second LSTM hidden layer with a second number of units, the second LSTM hidden layer is then connected to a second discarding layer, and the second discarding layer is connected to the sample rectification data as output through a full connection layer; and
and the optimization module is configured to perform iterative optimization training on the LSTM network by using an optimizer so as to obtain an LSTM network model.
In certain embodiments, the pre-processing module may further comprise:
a data splicing module configured to splice each group of data of the same heliostat in the same day into a matrix
The matrix combination module is configured to obtain a sample cell array by using a cell array combination matrix; and
and the array processing module is configured to process the sample cell array, so that the normalized distribution of the sample cell array is between 0 and 1, and the mean value and the standard deviation are stored.
Fig. 6 shows a structure of an electronic device according to an embodiment of the present disclosure, which includes a processor 601, a communication interface 602, a memory 603, and a communication bus 604, where the processor 601, the communication interface 602, and the memory 603 complete communication with each other through the communication bus 604;
a memory 603 for storing a computer program;
when the processor 601 executes the program stored in the memory 603, the method for rectifying the artificial intelligence of any heliostat provided by the disclosure is implemented, and the method comprises the following steps:
reading sample data;
training an LSTM network according to the sample data to obtain an LSTM network model;
inputting the sequence data of the heliostat into an LSTM network model, and outputting the deviation rectifying data of the heliostat after calculation of the LSTM network model; and
and correcting the daily tracking operation parameters of the heliostat by using the deviation correction data of the heliostat.
The communication bus 604 mentioned in the above electronic device may be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The communication bus may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is shown, but this is not intended to represent only one bus or type of bus.
The communication interface 602 is used for communication between the electronic apparatus and other apparatuses.
The memory 603 may be either volatile memory or nonvolatile memory, or may include both volatile and nonvolatile memory. The non-volatile Memory may be a Read-Only Memory (ROM), a Programmable ROM (PROM), an Erasable PROM (EPROM), an Electrically Erasable PROM (EEPROM), or a flash Memory. Volatile Memory can be Random Access Memory (RAM), which acts as external cache Memory. By way of example, and not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDRSDRAM), Enhanced Synchronous SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), and Direct Rambus RAM (DRRAM). The memory 602 described herein is intended to comprise, without being limited to, these and any other suitable types of memory.
The Processor may be a general-purpose Processor, and includes a Central Processing Unit (CPU), a Network Processor (NP), and the like; the Integrated Circuit may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic device, or discrete hardware components.
The embodiment of the disclosure further provides a computer readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the method for rectifying deviation of artificial intelligence of any heliostat provided by the disclosure is implemented. Among others, storage media may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as read-only memory, flash memory, a hard disk, or a solid state disk; the memory may also comprise a combination of the above kinds of memories.
It is noted that, in this document, relational terms such as "first" and "second," and the like, are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
As will be appreciated by one of skill in the art, embodiments of the present disclosure may be provided as a method, system, or computer program product. Accordingly, the present disclosure may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present disclosure may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and so forth) having computer-usable program code embodied therein.
The previous description is only for the purpose of describing particular embodiments of the present disclosure, so as to enable those skilled in the art to understand or implement the present disclosure. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the disclosure. Thus, the present disclosure is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. The heliostat artificial intelligence deviation rectifying method comprises the following steps:
reading sample data, wherein the sample data comprises sample sequence data and sample deviation correction data, the sample sequence data comprises data related to deviation amount, and the sample deviation correction data is determined by performing regression operation according to the data related to deviation amount;
training an LSTM network according to the sample data to obtain an LSTM network model;
inputting the sequential data of the heliostat into the LSTM network model, and outputting the deviation correcting data of the heliostat after calculation by the LSTM network model; and
and correcting the daily tracking operation parameters of the heliostat by using the deviation correcting data of the heliostat.
2. The method of claim 1, wherein the training of the LSTM network according to the specimen data, resulting in an LSTM network model comprises:
preprocessing the sample data;
in the LSTM network, the preprocessed sample sequence data is used as input to be connected with a first LSTM hidden layer with a first number of units, the first LSTM hidden layer is then connected with a first discarding layer, and the first discarding layer is connected with the sample deviation correction data used as output through a full connection layer; and
and performing iterative optimization training on the LSTM network by using an optimizer to obtain the LSTM network model.
3. The method of claim 1, wherein the training of the LSTM network according to the specimen data, resulting in an LSTM network model comprises:
preprocessing the sample data;
in the LSTM network, the preprocessed sample sequence data is used as input to be connected with a first LSTM hidden layer with a first quantity unit, the first LSTM hidden layer is then connected with a first discarding layer, the first discarding layer is then connected with a second LSTM hidden layer with a second quantity unit, the second LSTM hidden layer is then connected with a second discarding layer, and the second discarding layer is connected with the sample deviation rectifying data used as output through a full connection layer; and
iteratively training the LSTM network using an optimizer to obtain the LSTM network model.
4. A method as claimed in claim 2 or 3, wherein said pre-processing said sample data comprises:
splicing all groups of data in the same day of the same heliostat into a matrix;
combining the matrixes by using the cell arrays to obtain sample cell arrays; and
and processing the sample cell array to ensure that the normalized distribution of the sample cell array is between 0 and 1, and storing the mean value and the standard deviation.
5. The method of claim 2 or 3, wherein the iteratively optimally training the LSTM network using an optimizer comprises:
and performing iterative optimization training on the LSTM network by adopting a gradient descent optimizer.
6. The method of claim 3, wherein,
the first number is less than the second number and a drop rate of the first drop layer is less than a drop rate of the second drop layer.
7. The method of claim 1, wherein said reading sample data comprises:
at least 6 sets of sample sequence data for each heliostat scattered over each day are read, and irrelevant partial data are ignored.
8. Heliostat artificial intelligence deviation correcting device includes:
the data reading module is configured to read sample data, wherein the sample data comprises sample sequence data and sample deviation rectifying data, the sample sequence data comprises data related to deviation amount, and the sample deviation rectifying data is determined by performing regression operation according to the data related to deviation amount;
the model training module is configured to train the LSTM network according to the sample data to obtain an LSTM network model;
the model calculation module is configured to input the sequence data of the heliostat into the LSTM network model, and output the deviation correction data of the heliostat after calculation of the LSTM network model; and
a result application module configured to modify the daily tracking operational parameters of the heliostat using the deskew data for the heliostat.
9. The electronic equipment comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are communicated with each other through the communication bus;
the memory is used for storing computer programs;
the processor is configured to implement the method of any one of claims 1-7 when executing the computer program stored on the memory.
10. Computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the method according to any one of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110372862.7A CN113536655B (en) | 2021-04-07 | 2021-04-07 | Heliostat artificial intelligence deviation rectifying method, heliostat artificial intelligence deviation rectifying device, electronic equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110372862.7A CN113536655B (en) | 2021-04-07 | 2021-04-07 | Heliostat artificial intelligence deviation rectifying method, heliostat artificial intelligence deviation rectifying device, electronic equipment and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113536655A CN113536655A (en) | 2021-10-22 |
CN113536655B true CN113536655B (en) | 2022-07-22 |
Family
ID=78094422
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110372862.7A Active CN113536655B (en) | 2021-04-07 | 2021-04-07 | Heliostat artificial intelligence deviation rectifying method, heliostat artificial intelligence deviation rectifying device, electronic equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113536655B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114111067B (en) * | 2021-11-26 | 2022-09-16 | 北京聚树核科技有限公司 | Mirror field deviation rectifying method and device for tower type molten salt photo-thermal power generation system |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101997446A (en) * | 2009-08-14 | 2011-03-30 | 珠海市奥凯励光电技术有限公司 | Artificial intelligent environment self-recognition multifunctional controller used for solar battery |
CN106444056A (en) * | 2016-12-09 | 2017-02-22 | 中国科学院光电技术研究所 | Sparse optical synthetic aperture imaging device based on three apertures and light beam combination correction method thereof |
CN108536123A (en) * | 2018-03-26 | 2018-09-14 | 北京交通大学 | The method for diagnosing faults of the train control on board equipment of the long neural network of memory network combination in short-term |
CN111241744A (en) * | 2020-01-07 | 2020-06-05 | 浙江大学 | Low-pressure casting machine time sequence data abnormity detection method based on bidirectional LSTM |
CN112001433A (en) * | 2020-08-12 | 2020-11-27 | 西安交通大学 | Flight path association method, system, equipment and readable storage medium |
Family Cites Families (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102980313A (en) * | 2012-12-26 | 2013-03-20 | 首航节能光热技术股份有限公司 | Heliostat error correction system and method for solar tower optical-thermal power station |
CN108562811B (en) * | 2018-03-12 | 2020-05-22 | 西安理工大学 | Bidirectional long-short term memory-based complex power quality disturbance analysis method |
CN109062269A (en) * | 2018-08-23 | 2018-12-21 | 杨军峰 | A kind of heliostat focuses method for correcting error, apparatus and system automatically |
CN110070102B (en) * | 2019-03-13 | 2023-05-23 | 西安理工大学 | Method for establishing sequence-to-sequence model for identifying power quality disturbance type |
CN110233661B (en) * | 2019-04-28 | 2021-06-29 | 国网安徽省电力有限公司信息通信分公司 | Long-short term memory neural network training method, channel parameter adjusting system and method |
CN111367323B (en) * | 2020-04-16 | 2021-01-01 | 宁***新能源科技有限公司 | Heliostat monitoring control system based on computer vision |
CN112508245A (en) * | 2020-11-26 | 2021-03-16 | 郑州迈拓信息技术有限公司 | Method and system for predicting power generation capacity based on artificial intelligence photovoltaic cell panel glass breakage |
-
2021
- 2021-04-07 CN CN202110372862.7A patent/CN113536655B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101997446A (en) * | 2009-08-14 | 2011-03-30 | 珠海市奥凯励光电技术有限公司 | Artificial intelligent environment self-recognition multifunctional controller used for solar battery |
CN106444056A (en) * | 2016-12-09 | 2017-02-22 | 中国科学院光电技术研究所 | Sparse optical synthetic aperture imaging device based on three apertures and light beam combination correction method thereof |
CN108536123A (en) * | 2018-03-26 | 2018-09-14 | 北京交通大学 | The method for diagnosing faults of the train control on board equipment of the long neural network of memory network combination in short-term |
CN111241744A (en) * | 2020-01-07 | 2020-06-05 | 浙江大学 | Low-pressure casting machine time sequence data abnormity detection method based on bidirectional LSTM |
CN112001433A (en) * | 2020-08-12 | 2020-11-27 | 西安交通大学 | Flight path association method, system, equipment and readable storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN113536655A (en) | 2021-10-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Durrani et al. | Photovoltaic yield prediction using an irradiance forecast model based on multiple neural networks | |
KR100328187B1 (en) | Method and apparatus for estimating generated energy of solar cell | |
CN113536655B (en) | Heliostat artificial intelligence deviation rectifying method, heliostat artificial intelligence deviation rectifying device, electronic equipment and storage medium | |
Alblawi et al. | ANN for assessment of energy consumption of 4 kW PV modules over a year considering the impacts of temperature and irradiance | |
Kim et al. | Tracking control of high‐concentration photovoltaic systems for minimizing power losses | |
Memon et al. | Determining optimum tilt angle for 1 MW photovoltaic system at Sukkur, Pakistan | |
CN115936177A (en) | Photovoltaic output power prediction method and system based on neural network | |
CN113437938B (en) | Photovoltaic array output power calculation method considering field characteristics due to regional differences | |
CN117993283A (en) | Particle swarm optimization-based photovoltaic power generation array optimization design method and system | |
Ashetehe et al. | Development of optimal tilt angle models of a photovoltaic module for maximum power production: Ethiopia | |
Sánchez-García et al. | Short term photovoltaic power production using a hybrid of nearest neighbor and artificial neural networks | |
Tchao et al. | An Implementation of an optimized dual-axis solar tracking algorithm for concentrating solar power plants deployment | |
WO2023024822A1 (en) | Low-cost high-precision measurement method for solar radiation | |
Deville et al. | Open‐source photovoltaic model pipeline validation against well‐characterized system data | |
Tan et al. | Rectifying structural deflection effect of large solar concentrator via correction of sun-tracking angle in the concentrator photovoltaic system | |
CN101929684A (en) | Coal consumption calculating method of composite firing low heat value gas unit | |
US11486606B2 (en) | Method of forecasting heat output of solar collectors | |
Guangyu et al. | Heliostat attitude angle detection method based on BP neural network | |
CN114239910A (en) | Household photovoltaic power prediction method and device | |
CN111476393B (en) | Method for quantitatively evaluating feasibility of photovoltaic project | |
Cros et al. | Benchmarking different approaches to convert surface solar irradiance into PV power production: a case study with an operational forecast system for a roof-top PV farm | |
Durisch et al. | Efficiency and degradation of a copper indium diselenide photovoltaic module and yearly output at a sunny site in Jordan | |
Wild | Comparative assessment of the productive performance of three solar PV technologies at Gatton | |
US11823409B1 (en) | Systems and methods for single-axis tracking via sky imaging and machine leanring comprising a neural network to determine an angular position of a photovoltaic power system | |
CN114900124B (en) | Photovoltaic power generation detection system and detection method thereof |
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 |