CN102867217A - Projection pursuit-based risk evaluation method for meteorological disasters of facility agriculture - Google Patents

Abstract

The invention discloses a method for assessing the risk of facility agricultural meteorological disasters based on projection tracking, and belongs to the technical field of meteorological disaster assessment. The present invention obtains by using the neural network to fit the historical meteorological data; counts the annual frequency of meteorological disasters of each grade, and calculates the disaster comprehensive index of the evaluation index, establishes the disaster comprehensive index grading standard of the evaluation index; constructs the historical meteorological data Using the accelerated genetic algorithm to obtain the optimal projection vector, the risk evaluation model is established based on the projection value of the meteorological data on the optimal projection vector and the disaster comprehensive index grade of the evaluation index. The present invention applies the projection pursuit method to the field of meteorological disaster risk assessment for facility agriculture, and at the same time optimizes the projection vector by combining the accelerated genetic algorithm, and uses the actual meteorological data as the input data of the evaluation model described in the present invention to obtain a risk assessment result with high precision, which is a meteorological Disaster risk assessment research provides new ideas and methods.

Description

Risk assessment method of facility agrometeorological disasters based on projection pursuit

technical field

The invention discloses a method for assessing the risk of facility agricultural meteorological disasters based on projection tracking, and belongs to the technical field of meteorological disaster assessment.

Background technique

Since the 1990s, my country's northern solar greenhouse-based facility agriculture has developed rapidly, focusing on out-of-season and off-season vegetable cultivation. The area of facility agriculture has grown from 108,000 mu in 1981 to more than 45 million mu in 2010,29 The annual growth rate has increased by more than 440 times. At the same time, the development of facility agriculture has also provided the most basic and solid guarantee for the supply of off-season vegetables in cities. The ability of solar greenhouses in northern my country to resist natural disasters is poor, and the economic benefits obtained per unit area of facilities are far behind that of foreign countries. The main reason is that the level of facilities in my country is low, and the production of facility crops is highly dependent on climatic conditions and is greatly affected by disastrous weather. Therefore, it is necessary to study the relationship between facility agriculture and meteorological conditions from multiple perspectives, and systematically evaluate the risks of facility agriculture in my country. Meteorological disaster risk assessment is the quantitative analysis and evaluation of the possibility of different intensities of meteorological disasters in a risk area in a certain period of time and the possible consequences. Meteorological disaster risk management is through risk identification, risk estimation, and risk evaluation. On the basis of optimizing and combining various risk management techniques, the risk of meteorological disasters can be effectively controlled and the consequences of losses caused by risks can be properly dealt with, so as to achieve the goal of obtaining the greatest security at the least cost. Including construction of disaster risk assessment model, establishment of meteorological disaster risk assessment system, drawing of thematic map of meteorological disaster risk, etc. Disaster risk assessment is a key link in disaster risk management. It is the scientific basis for carrying out effective disaster prevention, disaster preparation, emergency rescue and other activities, and it is also an important basis for emergency capacity building and assessment. For the increasingly severe disaster situation that the world is facing, the study of disaster risk assessment has important urgency and practical significance. The research on the evaluation and prediction of meteorological disasters in facility agriculture is in its infancy, and scholars at home and abroad have already carried out some research.

Most developed countries with facility agriculture abroad have small land areas and a single climate type. Therefore, risk assessment research often focuses on models such as social risk, economic risk, environmental risk, potential risk, and comprehensive risk. For example, Piers proposed the AWR model, and Carter Proposed the SRI model and HSE proposed the COMAH model, etc., and applied these models to carry out risk analysis, but the research mostly focused on the economic field. Scholars have used the projection pursuit model (PPE) in the flood risk assessment method of the drainage network. This method does not solve the dimensional global optimization problem of the projection vector, and the accuracy of obtaining the optimal projection direction a ^* is not high.

Genetic algorithm mainly includes operation steps such as selection, crossover and mutation. Step 1: Randomly generate N groups of uniformly distributed random variables in the value change interval of each decision variable; Step 2: Calculate the value of the objective function, and arrange them from large to small; Step 3: Calculate the order-based evaluation function (use eval(V ) indicates); Step 4: Perform selection operation to generate a new population; Step 5: Perform crossover operation on the new population generated in Step 4; Step 6: Perform mutation operation on the new population generated in Step 5; Step 7: Evolution iteration; Step 8: The variable change interval of the excellent individual produced by the first and second evolutionary iterations is used as the new initial change interval of the variable, and the algorithm enters step 1, re-runs SGA, and forms an accelerated operation until the optimal individual’s optimization criterion function value If it is less than a certain set value or the algorithm operation reaches the predetermined number of accelerations, the entire algorithm operation ends. At this time, the best individual in the current population is designated as the result of RAGA. The above 8 steps constitute the accelerated genetic algorithm (RAGA) based on real code.

At present, there are not many studies on the risk analysis of a specific agricultural disaster by using the projected pursuit method based on accelerated genetic algorithm. Most of the existing meteorological disaster risk assessment methods take field crops as the research object, and few meteorological disaster risk assessment methods take facilities as the research object. Conventional disaster assessment methods mainly include expert consultation method, group decision-making method, risk assessment matrix and other methods, but these methods have strong subjective initiative; and experience. It can be seen that in the existing meteorological disaster risk assessment methods, the human influence is strong, and the theoretical basis is lacking to a certain extent.

Contents of the invention

The technical problem to be solved by the present invention is to provide a facility agricultural meteorological disaster risk assessment method based on projection tracking in view of the above-mentioned deficiency of the background technology.

The present invention adopts following technical scheme for realizing above-mentioned purpose of the invention:

The risk assessment method for facility agrometeorological disasters based on projection pursuit includes the following steps:

Step 1, data sorting: Carry out homogeneity test on the collected historical meteorological data, and interpolate the missing data to form a complete data set, and fit the complete data set;

Step 2, establish the grading standard of facility agricultural meteorological disasters: process the data obtained by fitting in step 1, count the annual frequency of meteorological disasters of each level, calculate the comprehensive index of each meteorological disaster, and use the cluster analysis method to obtain the evaluation index Disaster comprehensive index grading standard;

Step 3, building a projection tracking evaluation model: constructing a projection function, calculating the projection value of each meteorological disaster grade evaluation index on the projection vector, and establishing a correlation model based on the projection value and the disaster comprehensive index grade of the evaluation index;

Step 4, analysis of evaluation results: take the meteorological data of the area to be evaluated as the input data of the projection pursuit evaluation model, and then combine the geographical information, according to the risk evaluation level data of the area to be evaluated according to the projection pursuit evaluation model Generate a zonal patch map.

In the method for evaluating the risk of meteorological disasters in facility agriculture based on projection pursuit, step 1 uses BP neural network to fit the complete data set to obtain the indoor daily minimum temperature.

In step 3 of the projection pursuit-based risk assessment method for facility agrometeorological disasters: the best projection vector is obtained by locally optimizing the projection vector according to the accelerated genetic algorithm.

The present invention adopts the above-mentioned technical scheme, and has the following beneficial effects:

1. Apply the projection pursuit method to the field of facility agricultural meteorological disaster risk assessment, and optimize the projection vector in conjunction with the accelerated genetic algorithm at the same time, use actual meteorological data as the input data of the evaluation model described in the present invention, and obtain the risk assessment result with high precision, which is a meteorological Disaster risk assessment research provides new ideas and methods

2. Integrating factors such as temperature, precipitation, sunshine, and wind speed that have a greater impact on facility agriculture, a meteorological disaster risk assessment model for facility agriculture was constructed, which changed the previous evaluation of relying on a single factor and comprehensively considered the influence of multiple factors.

Description of drawings

Figure 1 is a schematic diagram of the risk assessment process.

Figures 2 to 6 are the risk level maps from January to May in the embodiment.

Fig. 7 to Fig. 10 are the risk grade map of September to December in the embodiment.

Fig. 11 is a process diagram of locally optimizing the projection vector by the accelerated genetic algorithm.

Detailed ways

Below in conjunction with accompanying drawing, the technical scheme of invention is described in detail:

Taking the meteorological disaster risk zoning of solar greenhouses in the north as an example, the facility agricultural meteorological disaster risk assessment method based on accelerated genetic algorithm projection pursuit is shown in Figure 1, which specifically includes the following steps:

Step 1, first collect the meteorological data of the area. The department collected and used the meteorological data of 243 stations in 16 provinces, municipalities and autonomous regions in northern my country from 1980 to 2009. The collected meteorological data include: temperature, sunshine, precipitation, wind speed, etc. Meteorological data of evaluation index data, specific greenhouse bearing capacity data. Carry out the homogeneity test on the meteorological data, and interpolate the missing meteorological data to obtain a complete data set; collect and evaluate the disaster bearing capacity of common and representative greenhouses in the area to heavy precipitation, strong wind, etc., and obtain disasters at all levels According to the correlation between meteorological elements in the greenhouse and external meteorological conditions, a simulation model of indoor extremely low temperature is established based on BP neural network. The input parameters of the model are the highest temperature, lowest temperature, radiation, and wind speed of the previous day; refer to relevant literature , find out the low temperature and low-light disaster indicators of common facility crops tomato and cucumber; use Matlab to program and process the observed meteorological data from 1980 to 2009, and use BP neural network method to fit the complete data set to get the daily minimum temperature in the greenhouse. The combined daily minimum temperature in the greenhouse passed the significance test of 0.01.

The BP neural network uses a single hidden layer BP network to simulate the minimum temperature in the greenhouse in spring and winter. Among them, the number of neurons in the input layer is 4, the number of neurons in the hidden layer is 9, and the number of neurons in the output layer is 1. The first layer inputs the outdoor total solar radiation (Rout), maximum temperature (Tomax), minimum temperature (Tomin) and wind speed (Wout) samples, the second layer is the hidden layer, and the third layer outputs the minimum outdoor temperature (Timin) Data, the transfer function of the hidden layer adopts the S-type tangent function tansig, and the transfer function of the output layer adopts the S-type logarithmic function logsig. The parameters related to model selection are: initial learning rate η=0.1, inertia factor α=0.9, maximum number of iterations=10000, target error=0.0001.

Step 2. According to the fitted indoor daily minimum temperature, the annual frequency of meteorological disasters of each grade is counted, and the comprehensive index (RI) of each meteorological disaster is calculated, taking into account the temperature, precipitation, sunshine, wind speed and other main disaster-causing indicators. The impact correction comprehensive index (RI), and then use the cluster analysis method to obtain the grading standard of the disaster comprehensive index of the evaluation index, in which:

RI=α×DF ₁ +β×DF ₂ +γ×DF ₃ +θ×DF ₄ (1)

Among them: α, β, γ, and θ are coefficients, which are respectively 0.2, 0.3, 0.3, and 0.2, and DF ₁ , DF ₂ , DF ₃ , and DF ₄ are the frequency of disasters at each level.

The cluster analysis method divides the comprehensive index (RI) into four categories, and obtains the index ranges of the corresponding four grades, namely grade I: 0-0.25, grade II: 0.26-0.5, grade III: 0.51-0.75, grade IV: 0.76-1.0, the grading standard of the disaster comprehensive index of the corresponding evaluation indicators can be determined, the meteorological disaster grade evaluation index database can be constructed, and the evaluation index system can be formed.

Step 3. Construct the projection pursuit evaluation model based on accelerated genetic algorithm: According to the disaster comprehensive index grading standard of the evaluation index, use the projection pursuit model to calculate the projection value of each meteorological disaster grade evaluation index, and establish according to the projection value and disaster grade value A risk assessment model for facility agrometeorological disasters based on projection pursuit method;

Taking each level of disaster comprehensive index of each meteorological disaster level evaluation index as an evaluation unit, construct an index sample set {x ^* (i, j)|i=1~n, j=1~p}, n is the disaster level , p is the number of evaluation indicators, n and p are natural numbers greater than or equal to 1, x ^* (i, j) represents the jth evaluation index of the i-th grade, and the specific implementation is as follows:

(1) Normalized processing index sample set:

For the larger value of x ^* (i, j), the higher the disaster risk level, the normalization process is performed using the following expression:

x(i, j) = (x ^* (i, j) - x _min (j))/(x _max (j) - x _min (j)) (2);

When the value of x ^* (i, j) is smaller, the disaster risk level is lower, and the following expression is used for normalization processing:

x(i, j) = (x _max (j) - x ^* (i, j))/(x _max (j) - x _min (j)) (3);

Among them, x _max (j), x _min (j) are the maximum value and minimum value of the jth index respectively, and x(i, j) is the normalized evaluation index;

(2) Construct the projection indicator function:

i=1，2，…，n；Step a, synthesize the p-dimensional data {x(i,j)|j=1,2...,p} into a projection direction with a={a(1),a(2)...,a(p)} One-bit projected value z(i):

i=1,2,...,n;

Step b, classify according to the one-dimensional scatter diagram of {z(i)|i=1～n}, when comprehensively projecting index values, it is required that the scatter characteristics of the projected value z(i) should be: the local projected points should be as dense as possible, It is best to condense into several point clusters, and spread out the projection point clusters as much as possible on the whole. Therefore, the projection indicator function can be expressed as:

Q(a)＝Sz*Dz (4)

Among them: Sz is the standard deviation of the projection value z(i); Dz is the local density of the projection value z(i), namely:

$\mathrm{<span\; class="notranslate">\; Sz</span>}<span\; class="notranslate">\; =</span>\sqrt{\frac{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$

$\mathrm{<span\; class="notranslate">\; Z</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; R</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span><span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; R</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; )</span>$

Among them, E(z) is the average value of the sequence {z(i)|i=1,2,...,n}; r(i,j) represents the distance between samples, r(i,j)=|z( i)-z(j)|; u[R-r(i, j)] is a unit step function;

R is the window radius of the local density. Its selection should not only make the average number of projection points contained in the window too small to avoid too large a moving average deviation, but also not make it increase too high with the increase of n. R can be determined according to experiments, and its value range is rmax+0.5P≤R≤2P. In this study, R is 0.1Sz, r(i, j) represents the distance between samples, r(i, j)=|z (i)-z(j)|; u(t) is a unit step function, when t<0, the function value is 0, when t≥0, its function value is 1, here the value of R is 0.1 Sz;

(3) Optimize the projection index function:

The accelerated genetic algorithm (RAGA) based on real number coding is used to solve its high-dimensional global optimization problem, and the optimal projection direction a ^* is obtained. The local optimization diagram of the accelerated genetic algorithm (RAGA) based on real number coding is shown in Figure 11.

Maximize the objective function: MaxQ(a)=Sz*Dz (7)

Restrictions: $\mathrm{the\; s}.t\underset{j=1}{\overset{p}{\mathrm{\&Sigma;}}}{a}^2\left(j\right)=1---\left(8\right),$

According to the obtained optimal projection direction, the projection values corresponding to each evaluation level are calculated: the initial population size of the selected parent is n=400, the crossover probability p _c =0.8, the mutation probability p _m =0.8, the number of excellent individuals is selected Set to 16, α=0.05, and the number of accelerations is 8, the best projection directions are a ^* = (0.859,0.469,0.137,0.1527), respectively, after bringing a ^* into the phase formula, the projection value z ^* (j )=(0,0.69,1.1878,1.5177).

Establish a correlation model between the projection value and the disaster comprehensive index level of the evaluation index (that is, the risk assessment model for facility agrometeorological disasters based on projection pursuit): ${\mathrm{the\; y}}^{*}\left(i\right)=1.0229\&times;e^{0.9093z^{*}\left(i\right)}.$

Step 4. Calculate the risk level according to the evaluation model established in step 3, and generate the risk evaluation level data of each region. According to the geographical information system (GIS), the regional facility agricultural meteorological disaster risk assessment level data, and use the inverse distance weight interpolation method in ArcGIS The risk level maps shown in Figures 2 to 10 were generated. Figures 2 to 6 show the risk levels from January to May, and Figures 7 to 10 show the risk levels from September to December.

To sum up, the present invention applies the projection pursuit method to the field of meteorological disaster risk assessment in facility agriculture, and at the same time combines the accelerated genetic algorithm to optimize the projection vector, and uses the actual meteorological data as the input data of the evaluation model described in the present invention to obtain the risk assessment result The accuracy is high, which provides a new idea and method for the study of meteorological disaster risk assessment. In another method, the present invention integrates factors such as air temperature, precipitation, sunshine, and wind speed that have a greater impact on facility agriculture, and constructs a meteorological disaster risk assessment model for facility agriculture. Instead of relying on the evaluation of a single factor in the past, a variety of factors are comprehensively considered. factor influence. The above-mentioned embodiment is only an application example of the present invention. The present invention is not limited to the risk assessment of the four meteorological disasters in the northern solar greenhouse. Any embodiment that meets the gist of the present invention falls within the protection scope of the present invention.

Claims (3)

1. based on the industrialized agriculture meteorological disaster risk evaluating method of projection pursuit, it is characterized in that comprising the steps:

Step 1, data preparation: the historical weather data of collecting is carried out test of homogeneity, and carry out interpolation to lacking the survey data, form complete data acquisition, described complete data acquisition is carried out match;

Step 2, set up industrialized agriculture meteorological disaster grade scale: the data that match obtains in the treatment step 1, the frequency occurs in the year of adding up each grade meteorological disaster, calculates the aggregative index of every kind of meteorological disaster, utilizes clustering methodology to draw the disaster aggregative index grade scale of evaluation index;

Step 3 makes up PPE assessment model: make up projection function, calculate the projection value of each meteorological disaster grade evaluation index on projection vector, set up correlation model in the disaster aggregative index grade according to projection value and evaluation index;

Step 4, evaluation result is analyzed: with the weather data in zone to be evaluated input data as PPE assessment model, the Regional Assessment of Risk level data to be evaluated that generates according to PPE assessment model again, combining geographic information generates zoning color spot figure by the risk assessment grade.

2. the industrialized agriculture meteorological disaster risk evaluating method based on projection pursuit according to claim 1 is characterized in that described step 1 adopts the BP neural network that described complete data acquisition match is drawn the indoor day lowest temperature.

3. the industrialized agriculture meteorological disaster risk evaluating method based on projection pursuit according to claim 1 is characterized in that in the described step 3: obtain the best projection vector according to accelerating genetic algorithm local optimum projection vector.

Images