CN113011645A - Power grid strong wind disaster early warning method and device based on deep learning - Google Patents
Power grid strong wind disaster early warning method and device based on deep learning Download PDFInfo
- Publication number
- CN113011645A CN113011645A CN202110275805.7A CN202110275805A CN113011645A CN 113011645 A CN113011645 A CN 113011645A CN 202110275805 A CN202110275805 A CN 202110275805A CN 113011645 A CN113011645 A CN 113011645A
- Authority
- CN
- China
- Prior art keywords
- wind speed
- wind
- early warning
- data
- sample data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000013135 deep learning Methods 0.000 title claims abstract description 38
- 230000001360 synchronised effect Effects 0.000 claims abstract description 13
- 238000012549 training Methods 0.000 claims description 35
- 238000012937 correction Methods 0.000 claims description 21
- 230000006870 function Effects 0.000 claims description 15
- 238000002310 reflectometry Methods 0.000 claims description 12
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 11
- 238000002592 echocardiography Methods 0.000 claims description 9
- 239000000178 monomer Substances 0.000 claims description 9
- 238000012360 testing method Methods 0.000 claims description 9
- 230000004913 activation Effects 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 6
- 238000003062 neural network model Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 2
- 230000002265 prevention Effects 0.000 abstract description 3
- 230000009467 reduction Effects 0.000 abstract description 3
- 239000010410 layer Substances 0.000 description 66
- 238000013528 artificial neural network Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 230000005540 biological transmission Effects 0.000 description 3
- 230000019771 cognition Effects 0.000 description 3
- 238000013277 forecasting method Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 239000002184 metal Substances 0.000 description 3
- 210000004027 cell Anatomy 0.000 description 2
- 230000001149 cognitive effect Effects 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 239000002356 single layer Substances 0.000 description 1
- 230000008667 sleep stage Effects 0.000 description 1
- 239000002904 solvent Substances 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 230000002618 waking effect Effects 0.000 description 1
Images
Classifications
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
- G06N3/047—Probabilistic or stochastic 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
-
- 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
-
- 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
- G06Q50/10—Services
- G06Q50/26—Government or public services
-
- 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
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Development Economics (AREA)
- Data Mining & Analysis (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Primary Health Care (AREA)
- Probability & Statistics with Applications (AREA)
- Educational Administration (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A power grid strong wind disaster early warning method and device based on deep learning are disclosed, wherein radar data are collected as input data, input into a pre-trained wind speed prediction model and output wind speed information; the wind speed information is used as an input element of a pre-trained wind disaster early warning model; the wind disaster early warning model is combined with the data of the synchronous automatic meteorological observation station to correct the wind speed information, the wind speed revision information after errors caused by micro-terrain information are eliminated is output, the gale disaster is early warned in a grading way, graded gale position early warning is provided in real time, the fine forecasting problem of the wind damage of the power grid is solved, wherein the strong convection gale forms a 1km multiplied by 1km spatial resolution, and a forecasting result is obtained within 0-120 minutes; and a service operation software system is formed, the precision of the early warning and forecasting of the gale disaster is further improved, the pertinence of disaster prevention and reduction and the emergency rescue efficiency of the power grid and equipment are improved, and the safe operation level of the power grid and the reliability of social power utilization are improved.
Description
Technical Field
The invention relates to the technical field of weather forecasting, in particular to a power grid gale disaster early warning method and device based on deep learning.
Background
The normal operation of the power grid is closely related to meteorological conditions, meteorological changes have obvious influence on the peak time and peak value of the power grid load, and meteorological disasters can also threaten the safe operation of the power grid.
In meteorological disasters, the influence of gale on the transmission line in the power grid system is the most serious, and the gale disasters generally include: the tower is caused to fall, the power transmission is interrupted, foreign matters are blown up to cause line short circuit and cause linkage faults to cause a larger power failure range, line galloping and windage yaw flashover are caused due to abnormal wind power, and the success rate of reclosing is reduced.
In the prior art, a Chinese patent (CN109902885B) proposes a typhoon prediction method based on a deep learning mixed CNN-LSTM model, and predicts whether a typhoon is formed and a formed path and strength; chinese patent application (CN109492756A) proposes a multi-element wire galloping early warning method and a related device based on deep learning to obtain effective early warning information of the galloping of a power transmission line; chinese patent application (CN109447315A) proposes a 'power meteorological element numerical forecasting method', a short-term forecasting result is obtained based on fused power meteorological data and a mesoscale numerical weather forecasting mode, then based on the short-term forecasting result and radar monitoring data, a deep convolutional neural network method is adopted to identify characteristic information of a radar echo jigsaw puzzle, and finally based on the characteristic information, a deep learning method is adopted to carry out short-term forecasting on the power meteorological element to obtain a short-term forecasting result; chinese patent (CN106126896B) proposes 'a hybrid model wind speed prediction method and system based on empirical mode decomposition and deep learning', an original wind speed time sequence is obtained, the original wind speed time sequence is decomposed according to the empirical mode decomposition to obtain a plurality of intrinsic mode functions, the wind speed prediction submodels corresponding to the original wind speed time sequence are used for prediction to obtain a prediction output value of each wind speed prediction submodel, and then the prediction output values are combined and overlapped to obtain a final overall prediction output value; chinese patent (CN103413174B) proposes a short-term wind speed multi-step prediction method based on a deep learning method, and a deep neural network regression model with a multi-input multi-output structure is used for multi-step prediction of wind speed.
In the prior art, the characteristics of gales generated by different weather systems are very different, the early warning and forecasting methods are also different, how to automatically identify different types of gales by using three-dimensional echo intensity data observed by a radar is difficult, and the steps of identifying and tracking a flow system, extracting the main characteristics of a squall line, identifying the squall line and the like are required; the time, location, and level of the gale generation are then analyzed based on a determination of the squall line stage of development, a determination of the squall line shape, and the like. The effect of different types of high winds on the power production and the equipment is not only related to the wind speed but also to the wind direction, as well as to the duration of the wind, the instantaneous changes in the wind speed and direction, etc. The existing strong wind forecasting and early warning technology aims at more public services, less professional strong wind disaster early warning services for power grid production, no classification aiming at wind damage of different production stages of a power grid, and no systematic strong wind early warning and forecasting method and hierarchical forecasting products.
Disclosure of Invention
In order to solve the defects in the prior art, the invention aims to provide a power grid gale disaster early warning method and device based on deep learning.
The invention adopts the following technical scheme.
A power grid strong wind disaster early warning method based on deep learning comprises the following steps:
step 1, collecting radar data to form input data;
step 2, inputting input data into a pre-trained wind speed prediction model; the wind speed prediction model is obtained by training sample data by adopting a deep learning algorithm, and can output wind speed information according to the input sample data;
step 3, acquiring wind speed information output by a wind speed prediction model, and inputting the wind speed information into the wind disaster early warning model as an input element of a pre-trained wind disaster early warning model; the wind disaster early warning model is combined with the data of the synchronous automatic meteorological observation station to perform fitting correction on the wind speed information, and can output the wind speed revision information with errors generated by the micro-terrain information eliminated;
and 4, acquiring wind speed revision information and carrying out graded early warning on the gale disasters.
Preferably, in step 1, the radar data includes maximum echo intensity obtained from the current volume sweep echo, maximum echo intensity corresponding height, convective cell echo top height, time variation of echo intensity, radial velocity, geometric center position, cloud water content, cloud shape, wind shear, near-earth humidity, wind speed corresponding to the mobile station, mesocyclone, reflectivity, latitude and longitude.
Preferably, the first and second electrodes are formed of a metal,
in step 2, the training method of the wind speed prediction model comprises the following steps:
step 2.1, collecting radar historical data as input sample data of a wind speed prediction model, wherein the radar historical data comprises maximum echo intensity obtained from historical volume sweep echoes, corresponding height of the maximum echo intensity, echo crest height of a convection monomer, time variation of the echo intensity, radial speed, geometric center position, cloud water content, cloud shape, wind shear, ground proximity humidity, wind speed corresponding to an automatic station, mesocyclone, reflectivity and longitude and latitude;
step 2.2, obtaining wind speed information used for representing wind speed and wind direction according to the input sample data, and taking the wind speed information as output sample data of a wind speed prediction model;
step 2.3, combining input sample data and output sample data into sample data, wherein the sample data comprises maximum echo intensity obtained from historical volume-sweep echoes, corresponding height of the maximum echo intensity, echo top height of a convection monomer, time variation of the echo intensity, radial velocity, geometric center position, cloud water content, cloud shape, wind shear, near-earth humidity, wind speed, cyclone, reflectivity and longitude and latitude of a corresponding automatic station, and wind speed and wind direction corresponding to the historical volume-sweep echoes;
step 2.4, combining a plurality of sample data into a sample data set, taking all input sample data in the sample data set as input data of the wind speed prediction model, taking all output sample data in the sample data set as output data of the wind speed prediction model, carrying out unsupervised learning layer by layer from the bottom layer to the top layer, and training weight parameters of each layer;
step 2.5, based on the weight parameters of each layer obtained by training, applying a wake-sleep algorithm, carrying out supervised learning from the top layer to the bottom layer by layer, and adjusting the weight parameters of each layer;
step 2.6, inputting all input sample data in the sample data set into the trained and adjusted wind speed prediction model, and acquiring test data output by the trained and adjusted wind speed prediction model; when the error between the test data and the output sample data meets the preset requirement, taking the trained and adjusted model as a final available wind speed prediction model; and when the error between the test data and the output sample data cannot meet the preset requirement, returning to the step 2.4, and training and adjusting the model.
The wind speed prediction model is a deep learning neural network model and comprises an input layer, a 3-layer hidden layer and an output layer; the input layer comprises 32 network nodes, the output layer comprises 3 network nodes, and the number of the network nodes in the hidden layer from the input layer to the output layer is 16, 8 or 4 in sequence; the learning rate is set to 0.001.
The hidden layer adopts a network structure formed by stacking 3 independently trained restricted Boltzmann machines.
The hidden layer is used as an activation function and the output layer is used as an activation function.
Preferably, the sample data needs to be normalized before being input into the model.
Preferably, the first and second electrodes are formed of a metal,
in step 3, the training method of the wind disaster early warning model comprises the following steps:
step 3.1, collecting synchronous automatic meteorological observation station data and micro-terrain information of a wind speed prediction area in the wind speed prediction area;
step 3.2, taking the difference value of the wind speed information and the data of the synchronous automatic meteorological observation station as a target function;
step 3.3, constructing a micro-terrain correction coefficient according to the micro-terrain information;
step 3.4, adjusting the micro-terrain correction coefficient by taking the optimal value of the objective function as a training target;
and 3.5, constructing a finally available wind disaster early warning model by using the adjusted micro-terrain correction coefficient pair.
The microtopography correction coefficient satisfies the following relation:
S＝1+K1K2K3
in the formula,
K1representing the terrain type parameters, and respectively taking the values of 2.2H/L and 1.4H/L for the mountain peak and the mountain slope, wherein H is the height of the mountain and L is the width of the windward side of the mountain,
K2represents a horizontal coordinate parameter and satisfies K21- | x |/L, wherein x is the horizontal coordinate from the mountain top,
K3represents a height coordinate parameter, satisfies K31-z/2.5H, wherein z is the height from the surface of the mountain,
when z >2.5H, let S ═ 1;
and when the H/L is more than 0.3, taking the H/L as 0.3.
A power grid gale disaster early warning device based on deep learning comprises: the wind disaster early warning system comprises an acquisition module, a wind speed prediction module, a wind disaster early warning module and an output module;
the acquisition module is used for acquiring radar data as input data; the radar data comprise maximum echo intensity obtained from the current body scanning echo, the corresponding height of the maximum echo intensity, the echo peak height of a convection monomer, the time variation of the echo intensity, the radial velocity, the geometric center position, the cloud water content, the cloud shape, the wind shear, the ground proximity humidity, the wind speed corresponding to the automatic station, the mesowhirl, the reflectivity and the longitude and latitude;
the wind speed prediction module is used for inputting input data into a pre-trained wind speed prediction model; the wind speed prediction model is obtained by training sample data by adopting a deep learning algorithm, and can output wind speed information according to the input sample data;
the wind disaster early warning module is used for acquiring wind speed information output by the wind speed prediction model, serving as an input element of a pre-trained wind disaster early warning model and inputting the input element into the wind disaster early warning model; the wind disaster early warning model is combined with the data of the synchronous automatic meteorological observation station to perform fitting correction on the wind speed information, and can output the wind speed revision information with errors generated by the micro-terrain information eliminated;
and the output module is used for acquiring the wind speed revision information and sending the gale disaster grading early warning to the user.
Compared with the prior art, the method has the advantages that the method realizes graded early warning and forecast of the gale disaster, provides graded gale position early warning in real time, and solves the fine forecast problem of the wind damage of the power grid, wherein the strong convection gale forms a spatial resolution of 1km multiplied by 1km and forecast results in 0-120 minutes; and a service operation software system is formed, the precision of the early warning and forecasting of the gale disaster is further improved, the pertinence of disaster prevention and reduction and the emergency rescue efficiency of the power grid and equipment are improved, and the safe operation level of the power grid and the reliability of social power utilization are improved.
Drawings
FIG. 1 is a flow chart of a power grid gale disaster early warning method based on deep learning according to the invention;
FIG. 2 is a frame schematic diagram of a wind speed prediction model and a wind disaster early warning model in the deep learning-based power grid strong wind disaster early warning method of the invention;
fig. 3 is a schematic structural diagram of the power grid gale disaster early warning device based on deep learning.
Detailed Description
The present application is further described below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and the protection scope of the present application is not limited thereby.
Referring to fig. 1, a power grid gale disaster early warning method based on deep learning includes the following steps:
step 1, radar data are collected to form input data.
Specifically, in step 1, the radar data includes, but is not limited to, maximum echo intensity obtained from the current volume sweep echo, maximum echo intensity corresponding height, convective cell echo ceiling height, time variation of echo intensity, radial velocity, geometric center position, cloud moisture content, cloud shape, wind shear, ground proximity humidity, wind speed corresponding to the mobile station, mesocyclone, reflectivity, latitude and longitude.
In the wind speed prediction stage, the deep learning deep neural network input parameters total 31 characteristic parameters, and those skilled in the art select different types and different quantities of characteristic parameters as the input parameters of the deep neural network according to actual requirements, wherein the selected input parameters in the preferred embodiment are a non-limiting preferred selection.
Step 2, as shown in fig. 2, inputting input data into a pre-trained wind speed prediction model; the wind speed prediction model is obtained by training sample data by adopting a deep learning algorithm, and can output wind speed information according to the input sample data.
In particular, the amount of the solvent to be used,
in step 2, the training method of the wind speed prediction model comprises the following steps:
step 2.1, collecting radar historical data as input sample data of a wind speed prediction model, wherein the radar historical data comprises maximum echo intensity obtained from historical volume sweep echoes, corresponding height of the maximum echo intensity, echo crest height of a convection monomer, time variation of the echo intensity, radial speed, geometric center position, cloud water content, cloud shape, wind shear, ground proximity humidity, wind speed corresponding to an automatic station, mesocyclone, reflectivity and longitude and latitude;
step 2.2, obtaining wind speed information used for representing wind speed and wind direction according to the input sample data, and taking the wind speed information as output sample data of a wind speed prediction model;
step 2.3, combining input sample data and output sample data into sample data, wherein the sample data comprises maximum echo intensity obtained from historical volume-sweep echoes, corresponding height of the maximum echo intensity, echo top height of a convection monomer, time variation of the echo intensity, radial velocity, geometric center position, cloud water content, cloud shape, wind shear, near-earth humidity, wind speed, cyclone, reflectivity and longitude and latitude of a corresponding automatic station, and wind speed and wind direction corresponding to the historical volume-sweep echoes;
step 2.4, combining a plurality of sample data into a sample data set, taking all input sample data in the sample data set as input data of the wind speed prediction model, taking all output sample data in the sample data set as output data of the wind speed prediction model, carrying out unsupervised learning layer by layer from the bottom layer to the top layer, and training weight parameters of each layer;
in the preferred embodiment, each layer of parameters is trained in a layering mode by using non-calibration data, and the method is an unsupervised training process and comprises the following steps: firstly, training a first layer by using calibration-free data, and learning parameters of the first layer during training, wherein the obtained model can learn the structure of the data due to the limitation and sparsity constraint of the model, so that the characteristic with more expressive ability than the input is obtained; after learning to obtain the n-1 th layer, the output of the n-1 th layer is used as the input of the n-1 th layer, and the n-th layer is trained, thereby obtaining the parameters of each layer.
Step 2.5, based on the weight parameters of each layer obtained by training, applying a wake-sleep algorithm, carrying out supervised learning from the top layer to the bottom layer by layer, and adjusting the weight parameters of each layer;
in the preferred embodiment, the data with the label is adopted for training, the error is transmitted from top to bottom, the network is finely adjusted, and the process is as follows: further optimizing and adjusting the parameters of the whole multilayer model based on the parameters of each layer obtained by training in the step 2.4, wherein the process is a supervised training process; the first step is similar to a random initialization initial value process of a neural network, and the first step of deep learning is not random initialization but obtained by learning the structure of input data, so that the initial value is closer to global optimum, and a better effect can be achieved.
Step 2.6, inputting all input sample data in the sample data set into the trained and adjusted wind speed prediction model, and acquiring test data output by the trained and adjusted wind speed prediction model; when the error between the test data and the output sample data meets the preset requirement, taking the trained and adjusted model as a final available wind speed prediction model; and when the error between the test data and the output sample data cannot meet the preset requirement, returning to the step 2.4, and training and adjusting the model.
In actual training, the time complexity is too high if all layers are trained simultaneously; if one layer is trained at a time, the bias is passed layer by layer, resulting in severe overfitting due to too many neurons and parameters in the deep neural network. Therefore, the preferred embodiment proposes a training method of the wind speed prediction model.
The training method changes the weights between other layers except the topmost layer into two-way, so that the topmost layer is still a single-layer neural network, and other layers become graph models. The upward weight is used for cognition, the downward weight is used for generation, and all the weights are adjusted by using a Wake-Sleep algorithm to make cognition and generation consistent, namely, the generated topmost expression can be used for correctly restoring the node at the bottom layer as much as possible.
The Wake-Sleep algorithm is divided into two parts, namely waking (Wake) and sleeping (Sleep). Wherein:
(1) a wake stage: and the cognitive process generates an abstract representation of each layer through external features and upward weight, namely cognitive weight, and expresses the abstract representation as a node state, and downward weight between layers is modified by using gradient descent, namely the cognitive weight is generated.
(2) sleep stage: the generation process generates the state of the bottom layer by the top-level representation, i.e., the concepts learned at wake stage, and the downward weights, while modifying the upward weights between the layers.
Specifically, the wind speed prediction model is a deep learning neural network model and comprises an input layer, a 3-layer hidden layer and an output layer; the input layer comprises 32 network nodes, the output layer comprises 3 network nodes, and the number of the network nodes in the hidden layer from the input layer to the output layer is 16, 8 or 4 in sequence; the learning rate is set to 0.001.
The hidden layer adopts a network structure formed by stacking 3 singly trained Restricted Boltzmann Machines (RBMs). The self-descending unsupervised pre-training between deep neural networks is achieved using RBM, whose learning goal is to maximize likelihood values.
The hidden layer is used as an activation function and the output layer is used as an activation function.
The sample data needs to be normalized before being input into the model. Since the unit and the value range of different sample data are not uniform, the sample data needs to be normalized.
In the preferred embodiment, the normalization method is a Gaussian normalization algorithm with 0 mean and 1 variance; the value range of the sample data after normalization processing is limited to a smaller range, the radar data after normalization is used as input sample data during model training, and the observation data of the automatic meteorological station after normalization is used as output sample data.
Step 3, acquiring wind speed information output by a wind speed prediction model, and inputting the wind speed information into the wind disaster early warning model as an input element of a pre-trained wind disaster early warning model; the wind disaster early warning model is combined with the data of the synchronous automatic meteorological observation station to perform fitting correction on the wind speed information, and can output the wind speed revision information with errors generated by the micro-terrain information eliminated;
preferably, the first and second electrodes are formed of a metal,
in step 3, the training method of the wind disaster early warning model comprises the following steps:
step 3.1, collecting synchronous automatic meteorological observation station data and micro-terrain information of a wind speed prediction area in the wind speed prediction area;
step 3.2, taking the difference value of the wind speed information and the data of the synchronous automatic meteorological observation station as a target function;
step 3.3, constructing a micro-terrain correction coefficient according to the micro-terrain information;
step 3.4, adjusting the micro-terrain correction coefficient by taking the optimal value of the objective function as a training target;
and 3.5, constructing a finally available wind disaster early warning model by using the adjusted micro-terrain correction coefficient pair.
The microtopography correction coefficient satisfies the following relation:
S＝1+K1K2K3
in the formula,
K1representing the terrain type parameters, and respectively taking the values of 2.2H/L and 1.4H/L for the mountain peak and the mountain slope, wherein H is the height of the mountain and L is the width of the windward side of the mountain,
K2represents a horizontal coordinate parameter and satisfies K21- | x |/L, wherein x is the horizontal coordinate from the mountain top,
K3represents a height coordinate parameter, satisfies K31-z/2.5H, wherein z is the height from the surface of the mountain,
when z >2.5H, let S ═ 1;
and when the H/L is more than 0.3, taking the H/L as 0.3.
In the preferred embodiment, a micro-terrain correction coefficient is constructed according to the micro-terrain information to obtain corrected wind speed information, and in an actual application scene, errors are generated on the actual wind speed by the power grid operation condition, the line state information and the like, and the errors need to be eliminated by adopting a parameter correction method. It should be noted that the inventive concept of modifying the wind speed information by using other information to construct the correction factor is also within the scope of the present invention.
And 4, acquiring wind speed revision information and carrying out graded early warning on the gale disasters.
In the preferred embodiment, the wind damage is divided into four levels, which are detailed in table 1.
TABLE 1 gale disaster early warning level
| Threshold value for early warning wind speed | Early warning level |
| V≤13m/s | Class I |
| 13m/s<V≤20m/s | Class II |
| 20m/s<V≤29m/s | Class III |
| V>29m/s | Grade IV |
As shown in fig. 3, a power grid gale disaster early warning device based on deep learning includes: the wind disaster early warning system comprises an acquisition module, a wind speed prediction module, a wind disaster early warning module and an output module.
The acquisition module is used for acquiring radar data as input data; the radar data comprise maximum echo intensity obtained from the current body scanning echo, the corresponding height of the maximum echo intensity, the echo peak height of a convection monomer, the time variation of the echo intensity, the radial velocity, the geometric center position, the cloud water content, the cloud shape, the wind shear, the ground proximity humidity, the wind speed corresponding to the automatic station, the mesowhirl, the reflectivity and the longitude and latitude;
the wind speed prediction module is used for inputting input data into a pre-trained wind speed prediction model; the wind speed prediction model is obtained by training sample data by adopting a deep learning algorithm, and can output wind speed information according to the input sample data;
the wind disaster early warning module is used for acquiring wind speed information output by the wind speed prediction model, serving as an input element of a pre-trained wind disaster early warning model and inputting the input element into the wind disaster early warning model; the wind disaster early warning model is combined with the data of the synchronous automatic meteorological observation station to perform fitting correction on the wind speed information, and can output the wind speed revision information with errors generated by the micro-terrain information eliminated;
and the output module is used for acquiring the wind speed revision information and sending the gale disaster grading early warning to the user.
Compared with the prior art, the method has the advantages that the method realizes the graded early warning and forecast of the gale disaster, provides graded gale position early warning in real time, and solves the fine forecast problem of the wind damage of the power grid, wherein the strong convection gale forms a spatial resolution of 1km multiplied by 1km and forecast results in 0-120 minutes; and a service operation software system is formed, the precision of the early warning and forecasting of the gale disaster is further improved, the pertinence of disaster prevention and reduction and the emergency rescue efficiency of the power grid and equipment are improved, and the safe operation level of the power grid and the reliability of social power utilization are improved.
The present applicant has described and illustrated embodiments of the present invention in detail with reference to the accompanying drawings, but it should be understood by those skilled in the art that the above embodiments are merely preferred embodiments of the present invention, and the detailed description is only for the purpose of helping the reader to better understand the spirit of the present invention, and not for limiting the scope of the present invention, and on the contrary, any improvement or modification made based on the spirit of the present invention should fall within the scope of the present invention.
Claims (10)
1. A power grid strong wind disaster early warning method based on deep learning is characterized in that,
the method comprises the following steps:
step 1, collecting radar data to form input data;
step 2, inputting input data into a pre-trained wind speed prediction model; the wind speed prediction model is obtained by training sample data by adopting a deep learning algorithm, and can output wind speed information according to the input sample data;
step 3, acquiring wind speed information output by a wind speed prediction model, and inputting the wind speed information into the wind disaster early warning model as an input element of a pre-trained wind disaster early warning model; the wind disaster early warning model is combined with synchronous automatic meteorological observation station data to perform fitting correction on wind speed information, and can output wind speed revision information with errors generated by micro-terrain information eliminated;
and 4, acquiring wind speed revision information and carrying out graded early warning on the gale disasters.
2. The power grid strong wind disaster early warning method based on deep learning of claim 1,
in the step 1, the radar data comprises maximum echo intensity obtained from the current body-scanning echo, corresponding height of the maximum echo intensity, echo top height of a convection monomer, time variation of the echo intensity, radial speed, geometric center position, cloud water content, cloud shape, wind shear, near-earth humidity, wind speed corresponding to an automatic station, cyclone in the middle, reflectivity and longitude and latitude.
3. The power grid strong wind disaster early warning method based on deep learning of claim 2,
in step 2, the training method of the wind speed prediction model comprises the following steps:
step 2.1, collecting radar historical data as input sample data of a wind speed prediction model, wherein the radar historical data comprises maximum echo intensity obtained from historical volume sweep echoes, corresponding height of the maximum echo intensity, echo top height of a convection monomer, time variation of the echo intensity, radial speed, geometric center position, cloud water content, cloud shape, wind shear, ground proximity humidity, wind speed corresponding to an automatic station, mesocyclone, reflectivity and longitude and latitude;
step 2.2, obtaining wind speed information used for representing wind speed and wind direction according to the input sample data, and taking the wind speed information as output sample data of a wind speed prediction model;
step 2.3, combining input sample data and output sample data into sample data, wherein the sample data comprises maximum echo intensity obtained from historical volume-sweep echoes, corresponding height of the maximum echo intensity, echo top height of a convection monomer, time variation of the echo intensity, radial velocity, geometric center position, cloud water content, cloud shape, wind shear, near-earth humidity, wind speed, cyclone, reflectivity and longitude and latitude of a corresponding automatic station, and wind speed and wind direction corresponding to the historical volume-sweep echoes;
step 2.4, combining a plurality of sample data into a sample data set, taking all input sample data in the sample data set as input data of the wind speed prediction model, taking all output sample data in the sample data set as output data of the wind speed prediction model, carrying out unsupervised learning layer by layer from the bottom layer to the top layer, and training weight parameters of each layer;
step 2.5, based on the weight parameters of each layer obtained by training, applying a wake-sleep algorithm, carrying out supervised learning from the top layer to the bottom layer by layer, and adjusting the weight parameters of each layer;
step 2.6, inputting all input sample data in the sample data set into the trained and adjusted wind speed prediction model, and acquiring test data output by the trained and adjusted wind speed prediction model; when the error between the test data and the output sample data meets a preset requirement, taking the trained and adjusted model as a final available wind speed prediction model; and when the error between the test data and the output sample data cannot meet the preset requirement, returning to the step 2.4, and training and adjusting the model.
4. The power grid strong wind disaster early warning method based on deep learning of claim 3,
the wind speed prediction model is a deep learning neural network model and comprises an input layer, a 3-layer hidden layer and an output layer; wherein,
the input layer comprises 32 network nodes, the output layer comprises 3 network nodes, and the number of the network nodes in the hidden layer from the input layer to the output layer is 16, 8 or 4 in sequence;
the learning rate is set to 0.001.
5. The power grid strong wind disaster early warning method based on deep learning of claim 4,
the hidden layer adopts a network structure formed by stacking 3 independently trained limited Boltzmann machines.
6. The power grid strong wind disaster early warning method based on deep learning of claim 4,
the hidden layer is characterized in that a sigmoid function is adopted as an activation function, and a softmax function is adopted as an activation function of the output layer.
7. The power grid strong wind disaster early warning method based on deep learning according to claim 1 or 3,
before the sample data is input into the model, normalization processing is required.
8. The power grid strong wind disaster early warning method based on deep learning of claim 1,
in step 3, the training method of the wind disaster early warning model comprises the following steps:
step 3.1, collecting synchronous automatic meteorological observation station data and micro-terrain information of a wind speed prediction area in the wind speed prediction area;
step 3.2, taking the difference value of the wind speed information and the data of the synchronous automatic meteorological observation station as a target function;
step 3.3, constructing a micro-terrain correction coefficient according to the micro-terrain information;
step 3.4, adjusting the micro-terrain correction coefficient by taking the optimal value of the objective function as a training target;
and 3.5, constructing a finally available wind disaster early warning model by using the adjusted micro-terrain correction coefficient pair.
9. The power grid strong wind disaster early warning method based on deep learning of claim 8,
the microtopography correction coefficient satisfies the following relation:
S＝1+K1K2K3
in the formula,
K1representing the terrain type parameters, and respectively taking the values of 2.2H/L and 1.4H/L for the mountain peak and the mountain slope, wherein H is the height of the mountain and L is the width of the windward side of the mountain,
K2represents a horizontal coordinate parameter and satisfies K21- | x |/L, wherein x is the horizontal coordinate from the mountain top,
K3represents a height coordinate parameter, satisfies K31-z/2.5H, wherein z is the height from the surface of the mountain,
when z >2.5H, let S be 1;
when H/L is more than 0.3, the ratio of H/L to 0.3 is taken.
10. A power grid gale disaster early warning device based on deep learning is characterized in that,
the early warning device includes: the wind disaster early warning system comprises an acquisition module, a wind speed prediction module, a wind disaster early warning module and an output module;
the acquisition module is used for acquiring radar data as input data; the radar data comprise maximum echo intensity obtained from the current body scanning echo, corresponding height of the maximum echo intensity, convection single echo top height, time variation of the echo intensity, radial velocity, geometric center position, cloud water content, cloud shape, wind shear, near-earth humidity, wind speed, mesocyclone, reflectivity and longitude and latitude of a corresponding automatic station;
the wind speed prediction module is used for inputting input data into a pre-trained wind speed prediction model; the wind speed prediction model is obtained by training sample data by adopting a deep learning algorithm, and can output wind speed information according to the input sample data;
the wind disaster early warning module is used for acquiring wind speed information output by the wind speed prediction model, serving as an input element of a pre-trained wind disaster early warning model and inputting the input element into the wind disaster early warning model; the wind disaster early warning model is combined with synchronous automatic meteorological observation station data to perform fitting correction on wind speed information, and can output wind speed revision information with errors generated by micro-terrain information eliminated;
and the output module is used for acquiring the wind speed revision information and sending the gale disaster grading early warning to the user.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110275805.7A CN113011645A (en) | 2021-03-15 | 2021-03-15 | Power grid strong wind disaster early warning method and device based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110275805.7A CN113011645A (en) | 2021-03-15 | 2021-03-15 | Power grid strong wind disaster early warning method and device based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113011645A true CN113011645A (en) | 2021-06-22 |
Family
ID=76407152
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110275805.7A Pending CN113011645A (en) | 2021-03-15 | 2021-03-15 | Power grid strong wind disaster early warning method and device based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113011645A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113570133A (en) * | 2021-07-26 | 2021-10-29 | 广西电网有限责任公司电力科学研究院 | Power transmission and distribution line risk prediction method and system for dealing with heavy rainfall |
| CN113792382A (en) * | 2021-09-10 | 2021-12-14 | 中国航空工业集团公司西安飞行自动控制研究所 | Multi-task mode ground proximity warning method based on machine learning |
| CN115238365A (en) * | 2022-09-07 | 2022-10-25 | 西南交通大学 | Tunnel post-disaster damage early warning method and system based on dynamic deep learning |
| CN115426358A (en) * | 2022-09-07 | 2022-12-02 | 河海大学 | Slope safety early warning method and system based on big data and storable medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106709820A (en) * | 2017-01-11 | 2017-05-24 | 中国南方电网有限责任公司电网技术研究中心 | Power system load prediction method and device based on deep belief network |
| US20180223804A1 (en) * | 2017-02-07 | 2018-08-09 | International Business Machines Corporation | Reducing curtailment of wind power generation |
| CN110929963A (en) * | 2019-12-16 | 2020-03-27 | 深圳智润新能源电力勘测设计院有限公司 | Wind speed prediction method, wind speed prediction device, and storage medium |
| CN111612315A (en) * | 2020-04-30 | 2020-09-01 | 国网江苏省电力有限公司电力科学研究院 | Novel power grid disastrous gale early warning method |
-
2021
- 2021-03-15 CN CN202110275805.7A patent/CN113011645A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106709820A (en) * | 2017-01-11 | 2017-05-24 | 中国南方电网有限责任公司电网技术研究中心 | Power system load prediction method and device based on deep belief network |
| US20180223804A1 (en) * | 2017-02-07 | 2018-08-09 | International Business Machines Corporation | Reducing curtailment of wind power generation |
| CN110929963A (en) * | 2019-12-16 | 2020-03-27 | 深圳智润新能源电力勘测设计院有限公司 | Wind speed prediction method, wind speed prediction device, and storage medium |
| CN111612315A (en) * | 2020-04-30 | 2020-09-01 | 国网江苏省电力有限公司电力科学研究院 | Novel power grid disastrous gale early warning method |
Non-Patent Citations (5)
| Title |
|---|
| LONG JIN: "An_Improved_Method_on_Meteorological_Prediction_Modeling_using_Genetic_Algorithm_and_Artificial_Neural_Network", 《2006 6TH WORLD CONGRESS ON INTELLIGENT CONTROL AND AUTOMATION》 * |
| 包博: "一种计及微地形修正的输电线台风风险预警方法", 《电力系统保护与控制》 * |
| 李帅: "气象因素对电力行业的影响研究", 《东北电力技术》 * |
| 胡伟成: "多国规范山地风速地形修正系数对比研究", 《工程力学》 * |
| 马婉贞: "基于WRF 数值模式的DBN 风速预测模型研究", 《计算机仿真》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113570133A (en) * | 2021-07-26 | 2021-10-29 | 广西电网有限责任公司电力科学研究院 | Power transmission and distribution line risk prediction method and system for dealing with heavy rainfall |
| CN113570133B (en) * | 2021-07-26 | 2024-05-24 | 广西电网有限责任公司电力科学研究院 | Power transmission and distribution line risk prediction method and system for coping with heavy rainfall |
| CN113792382A (en) * | 2021-09-10 | 2021-12-14 | 中国航空工业集团公司西安飞行自动控制研究所 | Multi-task mode ground proximity warning method based on machine learning |
| CN115238365A (en) * | 2022-09-07 | 2022-10-25 | 西南交通大学 | Tunnel post-disaster damage early warning method and system based on dynamic deep learning |
| CN115426358A (en) * | 2022-09-07 | 2022-12-02 | 河海大学 | Slope safety early warning method and system based on big data and storable medium |
| CN115238365B (en) * | 2022-09-07 | 2022-12-16 | 西南交通大学 | Tunnel post-disaster damage early warning method and system based on dynamic deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113011645A (en) | Power grid strong wind disaster early warning method and device based on deep learning | |
| WO2023045278A1 (en) | Data dual-drive method, apparatus, and device for predicting power grid failure during typhoon | |
| Giebel et al. | Wind power forecasting—a review of the state of the art | |
| CN113361186B (en) | Complete data-based wind turbine generator fault diagnosis method | |
| US20160223600A1 (en) | Power generation performance evaluation method and apparatus for power generator set | |
| US10345486B2 (en) | Apparatus and method for solar energy micro-forecasts for solar generation sources and utilities | |
| CN108898251A (en) | Consider the marine wind electric field power forecasting method of meteorological similitude and power swing | |
| CN111753893A (en) | Wind turbine generator power cluster prediction method based on clustering and deep learning | |
| CN106803135A (en) | The Forecasting Methodology and device of a kind of photovoltaic power generation system output power | |
| CN110362045A (en) | A kind of marine double-fed fan motor unit fault distinguishing method considering maritime meteorology factor | |
| CN117556197B (en) | Typhoon vortex initialization method based on artificial intelligence | |
| Dimitrov et al. | From SCADA to lifetime assessment and performance optimization: how to use models and machine learning to extract useful insights from limited data | |
| CN115935215B (en) | Power transmission line icing early warning method and system based on deep learning in extreme weather | |
| CN104794546B (en) | A kind of wind power climbing prediction technique based on depth confidence network class method | |
| CN114935759A (en) | Wave field missing value filling method and system based on high-frequency ground wave radar observation | |
| CN114564513A (en) | Sea fog prediction method, device, equipment and storage medium based on neural network | |
| CN111488974A (en) | Deep learning neural network-based ocean wind energy downscaling method | |
| CN116976227B (en) | Storm water increasing forecasting method and system based on LSTM machine learning | |
| CN109886488A (en) | Distributed wind power plant layered hybrid short-term prediction method considering wind speed time lag | |
| CN108038518A (en) | A kind of photovoltaic generation power based on meteorological data determines method and system | |
| KR20190071174A (en) | Method and system for short-term wind speed prediction based on pressure data | |
| CN116663393A (en) | Random forest-based power distribution network continuous high-temperature fault risk level prediction method | |
| CN115065166A (en) | Low-voltage distribution network state sensing and abnormal alarming method | |
| CN112070160B (en) | Multilayer collaborative real-time classification early warning method for wind power climbing event | |
| CN114297908A (en) | Wind turbine generator set energy efficiency state abnormity detection method and system |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210622 |