CN115097548B

CN115097548B - Sea fog classification early warning method, device, equipment and medium based on intelligent prediction

Info

Publication number: CN115097548B
Application number: CN202210944283.XA
Authority: CN
Inventors: 殷美祥; 宋清涛; 罗瑞婷; 王刚; 朱平; 郑延庆; 张志华; 王明辉
Original assignee: Guangdong Meteorological Public Service Center (guangdong Meteorological Film And Television Publicity Center)
Current assignee: Guangdong Meteorological Public Service Center (guangdong Meteorological Film And Television Publicity Center)
Priority date: 2022-08-08
Filing date: 2022-08-08
Publication date: 2023-04-07
Anticipated expiration: 2042-08-08
Also published as: CN115097548A

Abstract

The invention discloses a sea fog classification early warning method, a sea fog classification early warning device, sea fog classification early warning equipment and a sea fog classification early warning medium based on intelligent prediction, wherein the method comprises the following steps: receiving current meteorological characteristic data from terminal equipment in real time and inputting the current meteorological characteristic data into each prediction model of the prediction model set to obtain an initial prediction result of each prediction model; selecting a target prediction model with an initial prediction result corresponding to the actual sea fog information from the prediction model set in real time according to the actual sea fog information; predicting the current meteorological feature data according to a target prediction model to obtain a corresponding sea fog prediction result; and generating risk early warning information corresponding to the sea fog prediction result and sending the risk early warning information to each terminal device. The invention belongs to the technical field of meteorological early warning, and aims to predict sea fog by selecting a target prediction model corresponding to actual sea fog information in a prediction model set to obtain a sea fog prediction result, generate corresponding risk early warning information and send the risk early warning information to terminal equipment, so that the accurate prediction of various types of sea fog is greatly improved, and accurate early warning information is sent out.

Description

Sea fog classification early warning method, device, equipment and medium based on intelligent prediction

Technical Field

The invention relates to the technical field of meteorological early warning, in particular to a sea fog classification early warning method, device, equipment and medium based on intelligent prediction.

Background

Sea fog refers to a weather phenomenon that the visibility is less than 1 km caused by the condensation of a large amount of water vapor in an atmospheric boundary layer when the sea fog occurs on the sea surface or in an area near the shore. At present, due to the sudden and discontinuous changes of sea fog, and the great difference of the change mechanism and law of different types of sea fog, the forecast of sea fog visibility is still one of the more difficult fields in the weather forecast business. Although the sea fog can be predicted through the artificial intelligence model in the prior art, the prediction accuracy of different sea fog type models is obviously different, so that the prior art method cannot accurately predict various types of sea fog through the unified model and send accurate early warning information, and the port and the ship going out of the sea cannot timely deal with the sea fog due to the fact that the port and the ship going out of the sea cannot receive the accurate early warning information of the sea fog. Therefore, the prior art method has the problems that the multi-type sea fog cannot be accurately predicted and accurate early warning information cannot be sent out.

Disclosure of Invention

The embodiment of the invention provides an intelligent prediction-based sea fog classification early warning method, device, equipment and medium, and aims to solve the problem that the prior art method cannot accurately predict various types of sea fog and send accurate early warning information.

In a first aspect, an embodiment of the present invention provides an intelligent prediction-based sea fog classification early warning method, where the method includes:

receiving current meteorological characteristic data from the terminal equipment in real time and inputting the current meteorological characteristic data into each prediction model of a prediction model set to obtain an initial prediction result of each prediction model;

selecting a target prediction model with an initial prediction result corresponding to the actual sea fog information from the prediction model set in real time according to the actual sea fog information corresponding to the current meteorological characteristic data;

predicting the current meteorological feature data according to the target prediction model to obtain a corresponding sea fog prediction result;

generating risk early warning information corresponding to the sea fog prediction result according to a preset early warning classification rule;

and sending the risk early warning information to each terminal device.

In a second aspect, an embodiment of the present invention provides an intelligent prediction-based sea fog classification early warning apparatus, which includes:

the initial prediction result acquisition unit is used for receiving current meteorological feature data from the terminal equipment in real time and inputting the current meteorological feature data into each prediction model of the prediction model set to obtain an initial prediction result of each prediction model;

the target prediction model construction unit is used for selecting a target prediction model of which the initial prediction result corresponds to the actual sea fog information from the prediction model set in real time according to the actual sea fog information corresponding to the current meteorological characteristic data;

the sea fog prediction result acquisition unit is used for predicting the current meteorological feature data according to the target prediction model to obtain a corresponding sea fog prediction result;

a risk early warning information generating unit, configured to generate risk early warning information corresponding to the sea fog prediction result according to a preset early warning classification rule;

and the risk early warning information sending unit is used for sending the risk early warning information to each terminal device.

In a third aspect, an embodiment of the present invention further provides a computer device, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor, when executing the computer program, implements the method for classifying and warning fog based on intelligent prediction according to the first aspect.

In a fourth aspect, the embodiment of the present invention further provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and the computer program, when executed by a processor, causes the processor to execute the method for early warning based on intelligent prediction for sea fog classification according to the first aspect.

The embodiment of the invention provides a sea fog classification early warning method, a sea fog classification early warning device, sea fog classification early warning equipment and a sea fog classification early warning medium based on intelligent prediction. Receiving current meteorological characteristic data from terminal equipment in real time and inputting the current meteorological characteristic data into each prediction model of a prediction model set to obtain an initial prediction result of each prediction model; selecting a target prediction model with an initial prediction result corresponding to the actual sea fog information in real time from the prediction model set according to the actual sea fog information corresponding to the current meteorological characteristic data; predicting the current meteorological feature data according to a target prediction model to obtain a corresponding sea fog prediction result; generating risk early warning information corresponding to the sea fog prediction result according to a preset early warning classification rule; and sending the risk early warning information to each terminal device. By the method, the current meteorological characteristic data of the terminal device can be predicted through the prediction model set comprising the multiple prediction models to obtain the initial prediction result, the target prediction model corresponding to the initial prediction result and the actual sea fog information in the prediction model set is selected to carry out sea fog prediction to obtain the sea fog prediction result, corresponding risk early warning information is generated and sent to the terminal device, and accurate prediction on various types of sea fog and accurate early warning information sending are greatly improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

Fig. 1 is a schematic flow chart of an intelligent prediction-based sea fog classification early warning method according to an embodiment of the present invention;

fig. 2 is a schematic view of an application scenario of the intelligent prediction based sea fog classification early warning method according to the embodiment of the present invention;

fig. 3 is another schematic flow chart of the sea fog classification early warning method based on intelligent prediction according to the embodiment of the present invention;

fig. 4 is another schematic flow chart of the sea fog classification early warning method based on intelligent prediction according to the embodiment of the present invention;

fig. 5 is a schematic sub-flow diagram of a sea fog classification early warning method based on intelligent prediction according to an embodiment of the present invention;

fig. 6 is a schematic view of another sub-flow of the sea fog classification early warning method based on intelligent prediction according to the embodiment of the present invention;

fig. 7 is a schematic view of another sub-flow of the sea fog classification early warning method based on intelligent prediction according to the embodiment of the present invention;

fig. 8 is a schematic view of another sub-flow of the sea fog classification early warning method based on intelligent prediction according to the embodiment of the present invention;

fig. 9 is a schematic block diagram of an intelligent prediction based sea fog classification early warning device provided in an embodiment of the present invention;

FIG. 10 is a schematic block diagram of a computer device provided by an embodiment of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It will be understood that the terms "comprises" and/or "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.

It is also to be understood that the terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the specification of the present invention and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

It should be further understood that the term "and/or" as used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

Referring to fig. 1 and fig. 2, fig. 1 is a schematic flow chart of an intelligent prediction based sea fog classification early warning method according to an embodiment of the present invention, and fig. 2 is a schematic application scenario of the intelligent prediction based sea fog classification early warning method according to the embodiment of the present invention; the intelligent prediction based sea fog classification early warning method is applied to a management server 10, the management server 10 and a plurality of terminal devices 20 simultaneously establish network connection to transmit data information, and the intelligent prediction based sea fog classification early warning method is executed through application software installed in the management server 10; the management server 10 is a server side for executing an intelligent prediction-based sea fog classification early warning method to receive the current meteorological feature data prediction uploaded by the terminal device to obtain a sea fog prediction result and send corresponding risk early warning information, such as a server side built in an enterprise or a government department; the terminal device 20 is a device for transmitting the detected weather characteristic data to the management server, and the terminal device 20 can receive and display the risk early warning information sent by the management server. As shown in FIG. 1, the method includes steps S110 to S150.

And S110, receiving the current meteorological feature data from the terminal equipment in real time and inputting the current meteorological feature data into each prediction model of the prediction model set to obtain an initial prediction result of each prediction model.

Receiving current meteorological characteristic data from the terminal equipment in real time and inputting the current meteorological characteristic data into each prediction model of a prediction model set to obtain an initial prediction result of each prediction model; the set of prediction models includes prediction models corresponding to each sea fog type. The terminal device may be a weather monitoring and information receiving device disposed in a port, or a weather monitoring and information receiving device disposed on a ship.

Sea fog prediction is a very complex problem, and the difference of the characteristics of different sea fog change processes is very large; therefore, different parameterization schemes are needed for different sea fog process predictions, so that the optimal prediction result can be achieved. The project combines sea fog prediction and real-time evaluation, dynamically adjusts a prediction parameterization scheme and a prediction algorithm, constructs a sea fog prediction model set of a hybrid network, and greatly improves the prediction accuracy by selecting a specific prediction model in the prediction model set.

The current meteorological feature data received in real time can be input into each prediction model, and the current meteorological feature data is predicted through each prediction model, so that an initial prediction result corresponding to each prediction model is obtained. For example, at the t time node, the meteorological feature data of past M time nodes (the number of the time node is 0 to t-1) are used as the current meteorological feature data, and the initial prediction result obtained by predicting the current meteorological feature data by each prediction model in the prediction model set is calculated.

If the meteorological characteristic data of the t-1 time node is input into the prediction model, the initial prediction result of the sea fog type at the time t can be obtained.

In an embodiment, as shown in fig. 3, step S110 further includes steps S1101, S1102 and S1103.

S1101, if the input historical meteorological information is received, dividing the historical meteorological information according to preset interval time to obtain meteorological feature data corresponding to each time node.

If the input historical meteorological information is received, dividing the historical meteorological information according to preset interval time to obtain meteorological feature data corresponding to each time node; the meteorological characteristic data comprise characteristic data corresponding to various sea fog types.

The input historical meteorological information can be obtained, and the historical meteorological information covers meteorological information of various types of sea fog. Specifically, the sea fog can be generally classified into advection fog (advection fog), frontal fog (frontal fog), and radiation fog (radiation fog). The sea fog processes of different types are different in forming mechanism, atmospheric circulation field characteristics, micro-physical structure and boundary layer characteristics, and the relation between sea fog visibility and various meteorological element change characteristics is also greatly different, so that the types of the sea fog are divided into three types, namely advection fog, frontal fog and radiation fog. In addition, because sea fog visibility changes are linked to other meteorological elements in the spatial dimension as well as in the temporal dimension. The historical weather information is taken as the historical weather information of the port for illustration, except that the weather element data of the weather observation station local to the port is selected, the weather element data of the weather station surrounding the port is also selected as the source of the prediction model parameter, that is, the historical weather information may include the historical weather data of a target site and other sites surrounding the target site, for example, the historical weather information includes the historical weather data of the target port and other ports surrounding the target port. The historical meteorological information comprises a sea fog change mechanism and characteristics of a certain target port, including the monthly and daily change characteristics of sea fog, the relation between sea fog visibility and other meteorological elements, and the types of the sea fog, such as advection fog, radiation fog, mixed fog, terrain fog and the like, wherein the sea fog visibility can be used as characteristic parameters corresponding to the types of the sea fog, and the types of the sea fog can be used as target parameters for model training.

For example, the historical meteorological information includes m meteorological observation points of the target port and the surrounding ports, and the set of meteorological observation points can be represented as D = { D = { ₁ ,d ₂ ,...,d _i ,...,d _m }. The instrument for observing the meteorological information at the meteorological observation point comprises a novel automatic meteorological station (equipment model: DZZ 1-2), a photoelectric digital sunshine recorder (equipment model: DFC 3) and a precipitation phenomenon instrument (equipment model: DSG 5), and for example, the meteorological information is observed and recorded every 5 minutes at each meteorological observation point.

The historical weather information comprises weather information of past x hours, the historical weather information can be divided according to the interval time, if the interval time is 5 minutes, the weather information of the past x hours can be divided into 5 groups, each divided time period corresponds to one time node, and if the weather information of the past x hours can be divided into a plurality of time nodes: t is t ₁ ,t ₂ ,...,t _j ,...,t _x*60/5 . Acquiring meteorological characteristic data corresponding to each time node in historical meteorological information, wherein the meteorological characteristic data comprise meteorological element characteristic data, high-sensitivity factor data and time characteristic data.

The meteorological element characteristic data comprises visibility, air temperature, air pressure, relative humidity, wind direction, wind speed, precipitation, evaporation, ground temperature and grass temperature meteorological element data of x 60/5 time nodes (every 5 minutes) in the past x hours. The high-sensitivity factor data is determined according to the characteristics of the sea fog generation and the dissipation in the atmosphere humidity, and the temperature dew point difference (T-T) is constructed _d ) As high-sensitivity factor data for sea fog prediction; the temperature dew point difference is the difference between the temperature and the dew point, and is a physical quantity which is commonly used in daily weather analysis and forecast business to represent the dryness and humidity of air; in the weather, the temperature is represented by T and the dew point is represented by T _d Analysis on isobaric surfaces of the layers, etc. (T-T) _d ) The line is used to indicate the degree of saturation of the air with water vapor. The time characteristic data includes the number of hours and the number of months for a total of x × 60/5 time nodes (5 minutes by 5 hours) in the past x hours, i.e., the number of hours (an integer in the range of 1-24) to which the time node belongs in one day (24 hours); the number of months is the value of the month to which the node belongs in the year (an integer ranging from 1 to 12).

And S1102, performing correlation analysis on the visibility information of various types of fog in the meteorological characteristic data and other items of characteristic data in the meteorological characteristic data according to a preset correlation analysis rule to obtain corresponding correlation analysis information.

And performing correlation analysis on the visibility information of various types of fog in the meteorological characteristic data and other items of characteristic data in the meteorological characteristic data according to a preset correlation analysis rule to obtain corresponding correlation analysis information. Specifically, the meteorological characteristic data comprise characteristic data corresponding to various types of sea fog, and the characteristic data can be analyzed according to the types of the sea fog respectively; the meteorological characteristic data also comprises visibility information of various types of fog, the visibility information and other characteristic data in the same type of fog can be obtained and are respectively subjected to correlation analysis, correlation coefficients between the visibility information and the characteristic data in the same type of fog are obtained, and correlation analysis information corresponding to the meteorological characteristic data is obtained through combination.

In one embodiment, as shown in fig. 4, step S1110 is further included before step S1102.

And S1110, completing missing data in the meteorological feature data according to a preset preprocessing rule.

Before the meteorological characteristic data are processed, partial data missing in the meteorological characteristic data can be completed according to a preprocessing rule. For example, the lack measurement data is complemented by a Lagrange interpolation method, and after the data is complemented by standardization processing, errors can be reduced and the operation speed can be improved; data at two ends of the missing data can be calculated through Lagrange interpolation, so that the completion data corresponding to the missing data is obtained and filled to the position corresponding to the missing data.

In one embodiment, as shown in fig. 5, step S1102 includes sub-steps S1121 and S1122.

And S1121, acquiring visibility information corresponding to each type of sea fog at each time node from the meteorological characteristic data.

Specifically, visibility information corresponding to each type of sea fog at each time node can be respectively obtained, specific contents contained in the visibility information are represented in a numerical form, and the greater the numerical value of the visibility value is, the greater the visibility is.

And S1122, respectively carrying out association analysis on the visibility information of each type of sea fog at a plurality of time nodes and each item of characteristic data, corresponding to each type of sea fog at a plurality of time nodes, in the meteorological characteristic data to obtain corresponding association analysis information.

The visibility information of each sea fog type at a plurality of time nodes and each item of characteristic data of the corresponding time node in the meteorological characteristic data can be subjected to correlation analysis respectively, so that corresponding correlation analysis information is obtained. Specifically, the correlation analysis information includes correlation coefficients between visibility information of various types of fog and multiple items of characteristic data corresponding to the types of fog.

For example, the correlation analysis can be calculated using the following equation (1):

（1）；

wherein, X _i Is the characteristic data corresponding to the ith time node in a certain project,

is the average of all characteristic data of a certain item, Y _i Is the visibility value corresponding to the ith time node in the visibility information of a certain type of sea fog,

the average value of all visibility values in the visibility information of a certain type of sea fog, and R is a correlation coefficient obtained by calculation.

Through the formula, the correlation coefficient corresponding to the visibility information of a certain item of feature data and the visibility information of the same type of sea fog can be calculated, the correlation coefficient corresponding to the visibility information of each sea fog type and each item of feature data is obtained, and the corresponding correlation analysis information can be obtained.

S1103, constructing prediction models corresponding to various types of sea mists according to the correlation analysis information and the meteorological feature data to form the prediction model set.

A prediction model set can be constructed according to the correlation analysis information and the meteorological feature data, and the prediction model set comprises prediction models corresponding to various types of sea mists.

In one embodiment, as shown in fig. 6, step S1103 includes sub-steps S1131 and S1132.

S1131, establishing a parameter scheme set corresponding to each type of sea fog according to the correlation analysis information and the meteorological characteristic data.

And constructing a parameter scheme set corresponding to each type of sea fog according to the correlation analysis information and the meteorological feature data. Multiple groups of key parameters can be screened from meteorological characteristic data according to the correlation analysis information, and a parameter scheme set corresponding to each type of sea fog is obtained through combination.

The method comprises the following specific steps: respectively sorting the multiple items of feature data of each type of sea fog according to the correlation coefficient corresponding to each item of feature data to obtain a feature sorting result of each type of sea fog; selecting feature data from feature sorting results of various types of sea mists for multiple times according to preset selection times to combine; and acquiring parameters selected each time by each type of the sea fog to form the parameter scheme set, wherein the number of items of the characteristic data contained in each group of parameters of each type of the sea fog is increased in sequence.

The correlation coefficients of various characteristics contained in various types of sea mists can be used for sequencing the corresponding multiple items of characteristic data in the meteorological characteristic data, for example, the multiple items of characteristic data corresponding to various types of sea mists are respectively sequenced from large to small according to the correlation coefficients to obtain corresponding characteristic sequencing results, and each sea mist type corresponds to one group of characteristic sequencing results.

And sequentially selecting feature data from feature sequencing results of various types of sea fog for combination for multiple times according to the selection times, obtaining parameter combinations selected each time by each sea fog type to obtain parameter schemes of various sea fog types, wherein the number of items of the feature data contained in each group of parameters of various types of sea fog is increased in sequence.

For example, corresponding to three types of sea fog, namely advection fog, frontal fog and radiation fog, N meteorological elements which have the closest relation with the visibility of the sea fog are selected to construct a prediction parameter scheme set, and R weather elements are respectively selected _a 、R _f 、R _r (ii) a Because the more meteorological factors are not, the best prediction effect is. Advection fog parameter scheme set R _a ＝(a ₁ ，a ₂ ，…，a _i ，…，a _n ) Wherein a is ₁ The parameter is one item of feature data which is selected to be closest (the correlation coefficient is maximum), namely, the feature data corresponding to the first item of feature in the feature sorting result, a ₂ The parameter is that two items of feature data which are more closely cut (two items with larger correlation coefficient) are selected, namely feature data corresponding to the first two items of features in the feature sequencing result, and so on, a _n The parameter is that more closely cut N items of feature data are picked. Frontal fog parameter scheme set R _f ＝(f ₁ ，f ₂ ，…，f _i ，…，f _n ) Wherein f is ₁ The parameter is that the most dense is selectedOne item of feature data (with the largest correlation coefficient) is cut, namely the feature data corresponding to the first item of feature in the feature sorting result, f ₂ The parameters are two items of feature data which are more closely cut (two items with larger correlation coefficients) are selected, namely feature data corresponding to the first two items of features in the feature sorting result, and so on. Advection fog parameter scheme set R _r ＝(r ₁ ，r ₂ ，…，r _i ，…，r _n ) Wherein r is ₁ The parameter is the characteristic data which selects the most close item (the maximum correlation coefficient), namely the characteristic data corresponding to the first item of characteristic in the characteristic sorting result, r ₂ The parameters are two items of feature data which are more closely cut (two items with larger correlation coefficients) are selected, namely feature data corresponding to the first two items of features in the feature sorting result, and so on.

S1132, inputting parameter scheme sets corresponding to various types of sea fog and corresponding visibility information into a preset initial model for training to obtain a corresponding prediction model set; the set of prediction models includes a prediction model corresponding to each of the combination parameters.

Inputting parameter scheme sets corresponding to the same type of fog and corresponding visibility information into a preset initial model for training, collecting to obtain prediction models respectively corresponding to each parameter scheme set in each type of fog, and combining a plurality of prediction models to form a prediction model set. In a specific using process, a prediction model with an accurate result can be selected for use according to the accuracy of each prediction model in the prediction model set; the prediction model set obtained by training is composed of the prediction models obtained by training the characteristic data of various types of sea fog, so that the coverage of various types of sea fog can be realized through the prediction model set, and the accurate prediction of various types of sea fog can also be realized.

The method comprises the following specific steps: combining each group of parameters in the parameter scheme set of each type of sea fog with the visibility information of the corresponding type to obtain a combined parameter corresponding to each group of parameters; training the initial model according to each combination parameter to obtain a corresponding prediction model set; the set of prediction models includes a prediction model corresponding to each of the combination parameters.

And combining each group of parameters in the parameter scheme set of each type of sea fog with the visibility information of the same type of sea fog to obtain a combined parameter corresponding to each group of parameters.

For example, the number of parameter groups included in the three parameter scheme sets is 3 × N, each group of parameters in the first parameter scheme set is combined with visibility information of the same type of fog to obtain N combined parameters, and then the three parameter scheme sets can correspondingly obtain 3 × N combined parameters.

And respectively inputting each combination parameter into the initial model, namely training the initial model through a group of combination parameters to obtain a corresponding prediction model, and obtaining the prediction model obtained by training corresponding to each combination parameter to obtain a prediction model set.

Wherein, the steps specifically include: inputting each combination parameter into the initial model respectively; and calculating a gradient value corresponding to the output result of the initial model and the sea fog type of each combined parameter, and performing optimization training on the parameters of the initial model to obtain a prediction model corresponding to each combined parameter through training.

Specifically, each group of combination parameters can be respectively input into the initial model, the gradient value corresponding to the output result of the initial model and the visibility information of the sea fog type to which each group of parameters belongs is calculated, and the parameters of the initial model are optimally trained according to a gradient descent training mode, so that the adjusted prediction model is obtained. Because a group of combination parameters can carry out iterative training on the initial model to obtain a prediction model, the prediction model corresponding to each group of combination parameters can be obtained through corresponding training. Wherein the sea fog type can be advection fog, frontal fog, radiation fog or no sea fog.

Specifically, the initial model includes a plurality of independent neural networks. In order to improve the training accuracy, a plurality of independent neural networks can be configured in the initial model, each neural network is trained independently to form a prediction model, and through the setting mode, the number of the prediction models contained in the prediction model set can be greatly expanded, so that the applicability of prediction on various types of sea mists is further improved.

Specifically, acquiring output information of each neural network in the initial model; performing gradient calculation on the output information of each neural network and visibility information corresponding to the sea fog type to obtain a gradient value corresponding to each output information; and performing optimization training on the parameters of each neural network corresponding to each gradient value through a preset optimizer and the gradient value.

Specifically, the output information of each neural network in the initial model can be obtained, gradient calculation is performed on the output information of each neural network and visibility information corresponding to the type of sea fog, the gradient value of each neural network is obtained, optimization training is performed on parameters of the neural network corresponding to each gradient value through an optimizer and the gradient value, and finally the prediction model corresponding to each output information is obtained.

For example, the initial model in the embodiment of the present application includes two neural networks, namely, a threshold recurrent neural network (GRU) and a long-short term memory recurrent neural network (LSTM). The training process comprises the following steps: 1. inputting each combination parameter into the GRU; 2. calculating the updating gate, the resetting gate, the current time step memory content and the final memory content of the current time step of each layer; 3. training a network model by using an Adam optimization algorithm; 4. a set of trained GRU models is formed. 5. Inputting each combination parameter into the LSTM; 6. calculating an input gate, a forgetting gate, an output gate, the current time step memory content and the final memory content of the current time step of each layer; 7. training a network model by using an Adam optimization algorithm; 8. forming a well-trained LSTM model set; 9. and combining the GRU model set and the LSTM model set to obtain a final prediction model set.

In the process of training the threshold recurrent neural network (GRU), the threshold recurrent neural network (GRU) needs to be calculated; and calculating an update gate and a reset gate, wherein the values of the two gates are 0 to 1 and represent the filtering degree of the characteristic information.

At time step t, update gate to z _t ＝σ(w _z ·[h _t-1 ，x _t ]). Wherein x is _{_t} Is at presentVisibility information (sea fog visibility characteristic vector) input at a moment, namely the t-th component of the visibility characteristic sequence X. h is _t-1 Is the information output at the last time t-1, and sigma is the Sigmoid activation function. x is the number of _t And h _t-1 Respectively with a weight matrix w _z And (4) multiplying to realize linear transformation, adding the two parts of information, putting the information into a Sigmoid activation function, and outputting an activation value. Thus, the update gate effects control over the extent to which the state information at the previous time is brought into the state at the present time, the greater the value of the update gate, the more the state information at the previous time is brought into the state at the present time.

At time step t, reset gate to r _t ＝σ(w _r ·[h _t-1 ，x _t ]). Reset and update gates are computed in a similar manner, x _t And h _t-1 Respectively with a weight matrix w _r And (4) multiplying to realize linear transformation, adding the two parts of information, putting the information into a Sigmoid activation function, and outputting an activation value.

Then, the current memory content is calculated, and the reset gate r is calculated _t The value h of the previous hidden layer _t-1 Determining the size of the information before reservation and forgetting and the sea fog visibility characteristic vector x input by the current time node _t Respectively with a weight matrix w _p And multiplying to realize linear transformation, adding the calculation results of the two parts, and putting the result into a hyperbolic tangent activation function tanh to obtain the value of the current memory content.

And finally, calculating the final memory of the current time step. Calculating (1-Zt) and h _t-1 This part represents the information that was retained to the final memory at the previous time step. Calculating Z _t And h _t This part represents the information that the current memory content remains to the final memory. The addition of these two parts of the computation results is equal to the content of the final gated loop unit output. I.e. updating the door Z _t The larger the information is, the less the previous information is brought into the final output content, and the more the current memory content is brought into the final output content, so that the information flowing into the final output content is controlled by the update gate.

In the process of training the long-short term memory recurrent neural network (LSTM), the long-short term memory recurrent neural network (LSTM) needs to be calculated; at the current time t, calculate the input gate i _t Forgetting door f _t Output gate O _t Candidate state

Cell state C _t And a memory h _t 。

The input gate is

Wherein x is _f Sea fog visibility characteristic vector h input for the current moment _t-1 Is the information vector output at the last moment, b _i Is an offset term, W _i Is a weight parameter matrix, and sigma is a Sigmoid activation function. . x is a radical of a fluorine atom _t And h _t-1 And a weight parameter matrix W _i Multiplication, plus an offset term b _i Then, the input is put into a Sigmoid activation function to form the value of the input gate. The input threshold value ranges from 0 to 1 and determines how much of the information is stored in the current cell state.

At the current time t, the forgetting gate is

Wherein b is _f Is an offset term, W _f Is a weight parameter matrix, and sigma is a Sigmoid activation function. x is the number of _t And h _t-1 And a weight parameter matrix W _f Multiplication, plus an offset term b _f Then, the value of the forgetting gate is formed by inputting the value into a Sigmoid activation function. The forgetting threshold value ranges from 0 to 1 and selectively forgets information in the cellular state.

At the present time t, the output gate is

Wherein, b _o Is an offset term, W _o Is a weight parameter matrix, and sigma is a Sigmoid activation function. x is a radical of a fluorine atom _t And h _t-1 And weightParameter matrix W _o Multiplication, plus an offset term b _o Then, the output gate is put into a Sigmoid activation function to form a value of an output gate. The output threshold value ranges from 0 to 1, and selectively outputs information in the cell state.

At the current time t, the candidate state is

Wherein b is _c Is an offset term, W _c Is a weight parameter matrix, tanh is an activation function. x is the number of _t And h _t-1 And a weight parameter matrix W _c Multiply, add an offset term b _c After that, the tanh activation function is passed to form the value of the candidate state. The candidate state has a value ranging from-1 to 1, and represents the new knowledge summarized at the current moment and is stored in the cell state.

At the current time t, the cell state is

. Cell state C of last moment _t-1 Multiplying by the value of the forgetting gate and the current time candidate state

The sum of the values of the input gates is multiplied to form a value of the cell state, which characterizes long-term memory.

At the current time t, the memory is

. Current cell state

After the tanh activation function is passed, the current memory value is obtained by multiplying the current memory value by the output gate, and the current memory value represents the short-term memory.

In the process of training the network model, an Adam optimizer optimization model can be used, and particularly, the Adam (Adaptive motion optimization) optimizer is a parameter optimization method of Adaptive learning rate, and independent self-adaptation is designed for different parameters by calculating first-order momentum and second-order momentum of gradientAdaptive learning rate. w is the parameter to be optimized, loss function loss, learning rate lr, first-order momentum m _t And a second order momentum v _t The initialization value is 0; ε is a constant with a small value and is set to 10e ⁻⁸ 。β ₁ And beta ₂ Are constants, 0.9 and 0.999, respectively. The gradient of the loss function at the t moment relative to the current parameter can be calculated firstly by adopting m small batch samples from the training set

Wherein the loss value is a loss value between the LSTM and visibility information corresponding to the sea fog type or a loss value between the GRU and visibility information corresponding to the sea fog type; calculating the first order moment at time t

Correcting for deviations of first order momentum

(ii) a Calculating second order momentum

Correcting the deviation of the second order momentum

Then, the parameter at the time t +1 is calculated and updated

。

Continuously and iteratively updating the parameter values in the two neural networks by using an Adam optimizer, and gradually optimizing the parameter values in the model, such as a weight matrix w and a variable bias space b (b) in the LSTM _i 、b _f 、b _o 、b _c ) And performing iterative optimization until the model converges. For example, if the initial model contains two neural networks, the set of final emerging models f = (f) _a1 ，f _a1 ^’ ，f _a2 ，f _a2 ^’ ,…，f _an, ，f _an ^’ ，f _f1 ，f _f1 ^’ ，f _f2 ，f _f2 ^’ ，…，f _fn ，f _fn ^’ ，f _r1 ，f _r1 ^’ ，f _r2 ，f _r2 ^’ ，…，f _rn ，f _rn ^’ ) Total 6 × n models.

And S120, selecting a target prediction model with an initial prediction result corresponding to the actual sea fog information from the prediction model set in real time according to the actual sea fog information corresponding to the current meteorological feature data.

Because the current meteorological characteristic data is recorded historical meteorological data, actual sea fog information corresponding to the historical meteorological data, namely the sea fog type corresponding to the characteristic data of each time node in the historical meteorological data, for example, the actual sea fog information can be advection fog, frontal fog, radiation fog or no sea fog.

And calculating the accuracy or error degree of each prediction model according to the actual sea fog type and the initial prediction result of each prediction model, and selecting one or more prediction models with higher accuracy (or smaller error degree) as target prediction models to predict the actual sea fog.

In a specific application process, if only meteorological feature data of a single site is adopted, the prediction accuracy is low. In the embodiment of the application, the target adopts the meteorological characteristic data of the site and other sites around the target site to carry out comprehensive prediction (for example, meteorological characteristic data of a port and other ships around the port are adopted), so that the accuracy of sea fog prediction can be greatly improved.

For example, the accuracy score r between the predicted result of the visibility of the fog in the future of 1 month to 3 months in 2022 and the real fog information is predicted by using data of a single site ₂ Is 0.51; when a target site and data information of other sites corresponding to the target site are introduced, accurate prediction of the change rule of upstream and downstream spatial dimensions of the sea fog can be realized, so that the prediction accuracy of the visibility of the sea fog is greatly improved, and the accuracy score r between the sea fog visibility prediction accuracy and real sea fog information is increased ₂ Is 0.72.

In one embodiment, as shown in fig. 7, step S120 includes substeps S121 and S122.

S121, calculating an error value corresponding to the initial prediction result of each prediction model and the actual sea fog information according to a preset error calculation formula; and S122, screening the prediction model with the error value meeting the preset screening condition from the prediction model set to serve as a target prediction model.

The error value between the initial prediction result of each prediction model and the actual sea fog information can be calculated through an error calculation formula. For example, the Mean Absolute Error (MAE) of each prediction model may be calculated using an Error calculation formula to form a Mean Absolute Error set containing the Mean Absolute Error values of each prediction model. For example, the specific calculation formula is shown in formula (2):

（2）；

wherein, MAE is the calculated error value, M is the total number of time nodes contained in the current meteorological characteristic data, f _i An initial prediction result of the prediction model for the ith time node, the initial prediction result comprising a predicted visibility value, y _i The visibility value of the ith time node in the actual sea fog information is obtained.

Is f _i And y _i The absolute value of the difference between.

Specifically, a prediction model with the minimum error value is screened from the prediction model set to serve as a target prediction model meeting the screening condition; or screening a plurality of prediction models with smaller error values and the same quantity as the prediction models obtained from the screening conditions from the prediction model set to serve as target prediction models meeting the screening conditions.

One prediction model with the minimum error value can be screened from the prediction model set to serve as a target prediction model, and a plurality of prediction models can be obtained according to the obtaining number of the screening conditions to serve as target prediction models meeting the screening conditions. By the method, the intelligent selection of the prediction models in the prediction model set can be realized.

S130, predicting the current meteorological feature data according to the target prediction model to obtain a corresponding sea fog prediction result.

The current meteorological characteristic data can be predicted through the obtained target prediction model, and therefore an accurate sea fog prediction result is obtained.

In one embodiment, step S130 includes: acquiring the latest feature data in the current meteorological feature data as target input data; inputting the target input data into the target prediction model to obtain a model output result corresponding to the latest feature data; and integrating the output results of the models to obtain the sea fog prediction result with each time node only containing one prediction type.

And acquiring the latest feature data from the current meteorological feature data as target input data. And if the characteristic data with the time node t in the current meteorological characteristic data is obtained as target input data.

And inputting the target input data into the target prediction model so as to obtain a corresponding model output result. If target input data with the time node t is input into each prediction model, model output results of the time node t +1 and later time nodes can be obtained from each prediction model, and the model output results are results of the prediction model for predicting the sea fog type of the future time node. For example, the model output result can be a prediction result of every 5 minutes including the type of sea fog and visibility value 1-2 hours in the future.

And if the number of the target prediction models is one, directly taking the model output result as the sea fog prediction result. If the target prediction model is multiple, the obtained model output results are multiple, and the multiple model output results in the same time node can be integrated, so that an aerosol prediction result with only one prediction type in the same time node is obtained, one prediction type is also the predicted aerosol type, and the aerosol prediction result also comprises visibility values corresponding to each time node.

For example, the sea fog types corresponding to a plurality of target prediction models in the same time node may be counted, and one sea fog type with the largest number of statistics is determined as the predicted sea fog type, for example, in 5 target prediction models, the number of target prediction models corresponding to the advection fog is 3, the number of target prediction models corresponding to the frontal fog is 1, and the number of target prediction models corresponding to the radiation fog is 0.

And calculating the average value of the predicted visibility values of the target prediction models corresponding to the prediction types, and taking the average value as the visibility value of each time node in the sea fog prediction result. And if the visibility value is lower than a preset visibility threshold value, determining that the prediction type of the time node is no sea fog.

The predicted visibility value of each target prediction model may be weighted and calculated based on the error value of each target prediction model, and a specific weighted calculation formula is shown in formula (3).

（3）；

Wherein Rp is a weighted calculation value of the pth sea fog type, Q is the total number of the contained target prediction models, and MAEj is an error value of the jth target prediction model; fpj is the predicted visibility value of the jth target prediction model for the pth type of sea fog, if the model output result of the target prediction model is advection fog, the predicted visibility value of the target prediction model is the corresponding model output result when the weighted calculation value of the advection fog is calculated, and the predicted visibility value of the target prediction model is 0 when the weighted calculation values of other types of sea fog are calculated.

And after obtaining the weighted calculation value of each sea fog type, judging whether the weighted calculation value of a certain time node is lower than a preset visibility threshold value, if not, outputting the weighted calculation value as a visibility value in a sea fog prediction result, if so, determining that the prediction type is no sea fog, and simultaneously outputting the weighted calculation value as a visibility value in the sea fog prediction result.

And S140, generating risk early warning information corresponding to the sea fog prediction result according to a preset early warning classification rule.

The sea fog prediction results can be classified according to the early warning classification rules to generate corresponding risk early warning information, corresponding early warning grades are determined according to different sea fog types and different visibility values, and the risk early warning information corresponding to the corresponding early warning grades is generated.

In one embodiment, as shown in fig. 8, step S140 includes sub-steps S141 and S142.

And S141, grading the visibility value in the sea fog prediction result according to the early warning grading rule to obtain an early warning grade.

Visibility values in sea fog prediction results can be graded to obtain early warning grades, and the early warning grades are sea fog risk grades; if the national standard of sea fog early warning level can be combined, the sea fog risk level is divided into 1 level from high to low: fog with visibility less than or equal to 500 m exists in the future within 1 hour, and the visibility is 2: fog with visibility more than 500 meters and less than or equal to 1000 meters exists in the future within 1 hour, grade 3: fog with visibility more than 1000 m and less than or equal to 2000 m exists in the future within 1 hour.

And S142, generating corresponding risk early warning information according to the early warning grade and the prediction type in the sea fog prediction result.

And generating risk early warning information according to the obtained early warning level and the prediction type in the sea fog prediction result, wherein the prediction type is the type information for predicting the sea fog type. For example, the generated risk pre-warning information may be: in the future, the risk of level 1 sea fog exists in a short time, and the sea fog type is advection fog, so that the user can pay attention to prevention.

And S150, sending the risk early warning information to each terminal device.

And sending the obtained risk early warning information to each terminal device, for example, sending the risk early warning information to terminal devices on ports and ships in a wireless transmission mode. Because the predicted current meteorological characteristic data are obtained from each terminal device in real time, the risk early warning information can be quickly generated and sent based on the current meteorological characteristic data, and the timeliness of sending the risk early warning information can be ensured; meanwhile, sea fog prediction is carried out based on the current meteorological characteristic data of each terminal device, and it is ensured that the collected current meteorological characteristic data has no position deviation with the terminal devices, so that accuracy of risk early warning information sending is greatly improved.

Specifically, the risk avoidance prompting information and the risk early warning information can be generated according to the category of the terminal equipment and combined to form risk avoidance prompting early warning information which is sent to the corresponding terminal equipment, so that the early warning prompting effect of the risk early warning information is improved. Specifically, the current meteorological feature data further includes specific information such as equipment type identification and speed, the equipment type of the terminal equipment can be classified according to the equipment type identification, the speed type of the terminal equipment can be classified according to the speed, the navigation mode of the terminal equipment can be classified according to the distance from the terminal equipment to a port, and danger avoidance prompt information corresponding to the early warning level is acquired from the prompt information database according to the equipment type, the speed type and the navigation mode determined by classification. And combining the obtained emergency prompt information with the risk early warning information to generate corresponding risk avoidance prompt early warning information, and sending the generated risk avoidance prompt early warning information to corresponding terminal equipment.

If the equipment type identifier is S1, the corresponding equipment type is a large container ship, the equipment type identifier is S7, the corresponding equipment type is a civil yacht, and the equipment type identifier is R3, the corresponding equipment type is a port.

For example, if the equipment type is marked as "S1", the sailing speed is 8m/S, and the distance from the nearest port is 0.2km, it is determined that the equipment type of the terminal equipment is "large container cargo ship", the sailing speed is "low speed", the sailing mode is "berthing operation", and the determined early warning level is "level 1", then it is obtained from the prompt information database that the corresponding emergency prompt information is "please turn on the fog lamp of the ship when berthing operation is performed, and sailing is performed according to the instruction of the port berther.

In addition, the push frequency corresponding to each terminal device can be determined according to the device type identifier, the navigation speed and the distance from the nearest port, so that the push frequency can be differentiated, for example, the push frequency can be 2 minutes, 5 minutes, 10 minutes, 20 minutes, 30 minutes and the like. For example, if the type of a certain device is marked as "S1", the sailing speed is 8m/S, and the distance from the nearest port is 0.2km, it is determined that the pushing frequency of the terminal device is 10min, and the obtained danger avoidance prompt and early warning information is pushed to the terminal device every 10 minutes.

In the sea fog classification early warning method based on intelligent prediction provided by the embodiment of the invention, the current meteorological feature data from terminal equipment is received in real time and input into each prediction model of a prediction model set to obtain the initial prediction result of each prediction model; selecting a target prediction model with an initial prediction result corresponding to the actual sea fog information in real time from the prediction model set according to the actual sea fog information corresponding to the current meteorological characteristic data; predicting the current meteorological feature data according to the target prediction model to obtain a corresponding sea fog prediction result; generating risk early warning information corresponding to the sea fog prediction result according to a preset early warning classification rule; and sending the risk early warning information to each terminal device. By the method, the current meteorological characteristic data of the terminal equipment can be predicted through the prediction model set comprising the multiple prediction models to obtain the initial prediction result, the target prediction model corresponding to the initial prediction result and the actual sea fog information in the prediction model set is selected to perform sea fog prediction to obtain the sea fog prediction result, corresponding risk early warning information is generated and sent to the terminal equipment, and accurate prediction on the multi-type sea fog is greatly improved and accurate early warning information is sent out.

The embodiment of the invention also provides an intelligent prediction-based sea fog classification early warning device, which can be configured in a management server and is used for executing any embodiment of the intelligent prediction-based sea fog classification early warning method. Specifically, referring to fig. 9, fig. 9 is a schematic block diagram of an intelligent prediction based sea fog classification early warning apparatus according to an embodiment of the present invention.

As shown in fig. 9, the sea fog classification early warning apparatus 100 based on intelligent prediction includes an initial prediction result acquisition unit 110, a target prediction model construction unit 120, a sea fog prediction result acquisition unit 130, a risk early warning information generation unit 140, and a risk early warning information transmission unit 150.

An initial prediction result obtaining unit 110, configured to receive, in real time, current meteorological feature data from the terminal device, and input the current meteorological feature data into each prediction model of a prediction model set, so as to obtain an initial prediction result of each prediction model.

And the target prediction model construction unit 120 is configured to select, in real time, a target prediction model having an initial prediction result corresponding to the actual sea fog information from the prediction model set according to the actual sea fog information corresponding to the current meteorological feature data.

The sea fog prediction result obtaining unit 130 is configured to predict the current meteorological feature data according to the target prediction model to obtain a corresponding sea fog prediction result.

And a risk early warning information generating unit 140, configured to generate risk early warning information corresponding to the sea fog prediction result according to a preset early warning classification rule.

A risk early warning information sending unit 150, configured to send the risk early warning information to each terminal device.

The sea fog classification early warning device based on intelligent prediction provided by the embodiment of the invention is applied to the sea fog classification early warning method based on intelligent prediction, receives the current meteorological feature data from the terminal equipment in real time and inputs the data into each prediction model of a prediction model set to obtain the initial prediction result of each prediction model; selecting a target prediction model with an initial prediction result corresponding to the actual sea fog information in real time from the prediction model set according to the actual sea fog information corresponding to the current meteorological characteristic data; predicting the current meteorological feature data according to the target prediction model to obtain a corresponding sea fog prediction result; generating risk early warning information corresponding to the sea fog prediction result according to a preset early warning classification rule; and sending the risk early warning information to each terminal device. By the method, the current meteorological characteristic data of the terminal equipment can be predicted through the prediction model set comprising the multiple prediction models to obtain the initial prediction result, the target prediction model corresponding to the initial prediction result and the actual sea fog information in the prediction model set is selected to perform sea fog prediction to obtain the sea fog prediction result, corresponding risk early warning information is generated and sent to the terminal equipment, and accurate prediction on the multi-type sea fog is greatly improved and accurate early warning information is sent out.

The sea fog classification early warning device based on intelligent prediction can be implemented in the form of a computer program, and the computer program can be run on a computer device as shown in fig. 10.

Referring to fig. 10, fig. 10 is a schematic block diagram of a computer device according to an embodiment of the present invention. The computer equipment can be a server side used for executing the sea fog classification early warning method based on intelligent prediction so as to receive the current meteorological characteristic data uploaded by the terminal equipment to predict sea fog and obtain a sea fog prediction result and send corresponding risk early warning information.

Referring to fig. 10, the computer device 500 includes a processor 502, memory, and a network interface 505 connected by a system bus 501, where the memory may include a storage medium 503 and an internal memory 504.

The storage medium 503 may store an operating system 5031 and a computer program 5032. The computer program 5032, when executed, may cause the processor 502 to execute the sea fog classification warning method based on intelligent prediction, wherein the storage medium 503 may be a volatile storage medium or a non-volatile storage medium.

The processor 502 is used to provide computing and control capabilities that support the operation of the overall computer device 500.

The internal memory 504 provides an environment for the operation of the computer program 5032 in the storage medium 503, and when the computer program 5032 is executed by the processor 502, the processor 502 can be enabled to execute the sea fog classification warning method based on intelligent prediction.

The network interface 505 is used for network communication, such as providing transmission of data information. Those skilled in the art will appreciate that the configuration shown in fig. 10 is a block diagram of only a portion of the configuration associated with aspects of the present invention and is not intended to limit the computing device 500 to which aspects of the present invention may be applied, and that a particular computing device 500 may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.

The processor 502 is configured to run a computer program 5032 stored in the memory to implement the corresponding functions in the above-mentioned sea fog classification early warning method based on intelligent prediction.

Those skilled in the art will appreciate that the embodiment of a computer device illustrated in fig. 10 is not intended to be limiting of the specific construction of the computer device, and in other embodiments, the computer device may include more or fewer components than those shown, or some of the components may be combined, or a different arrangement of components. For example, in some embodiments, the computer device may only include a memory and a processor, and in such embodiments, the structures and functions of the memory and the processor are consistent with those of the embodiment shown in fig. 10, and are not described herein again.

It should be understood that, in the embodiment of the present invention, the Processor 502 may be a Central Processing Unit (CPU), and the Processor 502 may also be other general-purpose processors, digital Signal Processors (DSPs), application Specific Integrated Circuits (ASICs), field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components, and the like. Wherein a general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

In another embodiment of the present invention, a computer-readable storage medium is provided. The computer readable storage medium may be a volatile or non-volatile computer readable storage medium. The computer readable storage medium stores a computer program, wherein the computer program, when executed by the processor, implements the steps included in the above-mentioned sea fog classifying and early warning method based on intelligent prediction.

It can be clearly understood by those skilled in the art that, for convenience and simplicity of description, the specific working processes of the above-described devices, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again. Those of ordinary skill in the art will appreciate that the elements and algorithm steps of the examples described in connection with the embodiments disclosed herein may be embodied in electronic hardware, computer software, or combinations of both, and that the components and steps of the examples have been described in a functional general in the foregoing description for the purpose of illustrating clearly the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.

In the embodiments provided by the present invention, it should be understood that the disclosed apparatus, device and method can be implemented in other ways. For example, the above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units is only a logical division, and there may be other divisions when the actual implementation is performed, or units having the same function may be grouped into one unit, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may also be an electric, mechanical or other form of connection.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiments of the present invention.

In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit may be implemented in the form of hardware, or may also be implemented in the form of a software functional unit.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention essentially contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product stored in a computer-readable storage medium, which includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned computer-readable storage media comprise: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a magnetic disk, or an optical disk.

While the invention has been described with reference to specific embodiments, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. The sea fog classification early warning method based on intelligent prediction is characterized in that the method is applied to a management server, the management server and a plurality of terminal devices simultaneously establish network connection to transmit data information, and the method comprises the following steps:

receiving current meteorological characteristic data from the terminal equipment in real time and inputting the current meteorological characteristic data into each prediction model of a prediction model set to obtain an initial prediction result of each prediction model; the prediction model set comprises prediction models corresponding to various sea fog types;

sending the risk early warning information to each terminal device;

the method comprises the following steps that before the current meteorological characteristic data from the terminal equipment is received in real time and input into each prediction model of a prediction model set to obtain an initial prediction result of each prediction model, the method further comprises the following steps:

if the input historical meteorological information is received, dividing the historical meteorological information according to preset interval time to obtain meteorological feature data corresponding to each time node; the meteorological characteristic data comprise characteristic data corresponding to a plurality of sea fog types;

performing correlation analysis on the visibility information of various types of fog in the meteorological characteristic data and other items of characteristic data in the meteorological characteristic data according to a preset correlation analysis rule to obtain corresponding correlation analysis information;

constructing prediction models corresponding to various types of sea mists according to the correlation analysis information and the meteorological characteristic data to form the prediction model set;

the correlation analysis of the visibility information of various types of fog in the meteorological characteristic data and other item characteristic data in the meteorological characteristic data according to preset correlation analysis rules to obtain corresponding correlation analysis information comprises the following steps:

obtaining visibility information corresponding to each type of sea fog at each time node from the meteorological characteristic data;

respectively carrying out association analysis on visibility information of each type of sea fog at a plurality of time nodes and each item of characteristic data, corresponding to each type of sea fog at a plurality of time nodes, in the meteorological characteristic data to obtain corresponding association analysis information; the correlation analysis information comprises correlation coefficients between visibility information of various types of sea fog and multiple items of characteristic data corresponding to the types of sea fog;

the building of the prediction model corresponding to each type of sea fog according to the correlation analysis information and the meteorological characteristic data to form the prediction model set comprises the following steps:

constructing a parameter scheme set corresponding to each type of sea fog according to the correlation analysis information and the meteorological feature data;

inputting parameter scheme sets corresponding to various types of sea fog and corresponding visibility information into a preset initial model for training to obtain a corresponding prediction model set; the prediction model set comprises prediction models corresponding to all combination parameters; the combined parameters are the combination of each group of parameters in the parameter scheme set and the visibility information of the corresponding type.

2. The intelligent prediction based sea fog classification early warning method as claimed in claim 1, wherein before the correlation analysis of the visibility information of each type of sea fog in the meteorological characteristic data and other items of characteristic data in the meteorological characteristic data according to preset correlation analysis rules, the method further comprises:

and completing missing data in the meteorological characteristic data according to a preset preprocessing rule.

3. The intelligent prediction-based sea fog classification early warning method according to claim 1, wherein the selecting a target prediction model with an initial prediction result corresponding to the actual sea fog information from the prediction model set in real time according to the actual sea fog information corresponding to the current meteorological feature data comprises:

calculating an error value corresponding to the initial prediction result of each prediction model and the actual sea fog information according to a preset error calculation formula;

and screening the prediction model with the error value meeting the preset screening condition from the prediction model set to serve as a target prediction model.

4. The sea fog classifying and early warning method based on intelligent prediction as claimed in claim 1, wherein the generating of risk early warning information corresponding to the sea fog prediction result according to a preset early warning classification rule comprises:

grading the visibility value in the sea fog prediction result according to the early warning grading rule to obtain an early warning grade;

and generating corresponding risk early warning information according to the early warning grade and the prediction type in the sea fog prediction result.

5. The utility model provides a categorised early warning device of sea fog based on intelligent prediction which characterized in that, the device includes:

the initial prediction result acquisition unit is used for receiving current meteorological feature data from terminal equipment in real time and inputting the current meteorological feature data into each prediction model of the prediction model set to obtain an initial prediction result of each prediction model;

the target prediction model construction unit is used for selecting a target prediction model of which the initial prediction result corresponds to the actual sea fog information in real time from the prediction model set according to the actual sea fog information corresponding to the current meteorological characteristic data;

a risk early warning information sending unit, configured to send the risk early warning information to each terminal device;

before receiving the current meteorological feature data from the terminal device in real time and inputting the current meteorological feature data into each prediction model of the prediction model set to obtain an initial prediction result of each prediction model, the method further includes:

the method for performing correlation analysis on the visibility information of various types of sea fog in the meteorological characteristic data and other items of characteristic data in the meteorological characteristic data according to preset correlation analysis rules to obtain corresponding correlation analysis information comprises the following steps:

inputting parameter scheme sets corresponding to various types of sea fog and corresponding visibility information into a preset initial model for training to obtain a corresponding prediction model set; the prediction model set comprises prediction models corresponding to all combination parameters; the combination parameters are the combination of each group of parameters in the parameter scheme set and the visibility information of the corresponding type.

6. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the computer program implements the intelligent prediction based sea fog classification warning method according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program which, when executed by a processor, implements the intelligent prediction-based sea fog classification warning method according to any one of claims 1 to 4.