A design method for revisit occultation constellation based on multi-objective optimization

Abstract

A revisit occultation constellation design method based on multi-objective optimization. This invention relates to the field of satellite constellation design, and more specifically to a revisit occultation constellation design method. This method addresses the problems of diversity loss in existing occultation constellation optimization designs and the inapplicability of traditional revisit constellation design methods. The method comprises the following steps: step 1: inputting an observation area and dividing it into a grid; step 2: determining the value range of configuration parameters; step 3: calculating regression orbit pairs and recording the regression period dr; step 4: determining the optimization objective and fitness function; step 5: setting optimization algorithm parameters and randomly initializing the population; step 6: simulating the dr duration to obtain solution fitness; and step 7: iterating the population using the MOEA / D algorithm to update the non-dominated solution set. This invention is of great significance in improving constellation efficiency and optimizing observation capabilities.

Description

Technical Field

[0001] The present invention relates to the field of satellite constellation design, and in particular to a revisit occultation constellation design method. Background Art

[0002] Occultation atmospheric observation involves radio signals transmitted by a source satellite on one side of Earth, which are refracted by the Earth's atmosphere and received by an observation satellite on the other side. Inversion of the radio signals yields atmospheric observation data at the refraction point. In the context of global climate change and climate governance, designing a high-performance occultation constellation can effectively quantify greenhouse gas concentrations.

[0003] Current occultation constellation design methods primarily rely on simulation analysis, which evaluates constellation performance through simulation and analysis. However, this approach has several drawbacks. First, simulation analysis is inefficient. Due to the extensive simulation and analysis required, designing a high-performance occultation constellation consumes significant time and computing resources. Second, existing optimization design methods typically treat the problem as a single-objective optimization problem. This means only a single objective is considered, such as maximizing coverage or minimizing observation error. However, in occultation atmospheric observations, multiple objectives often exist, such as maximizing coverage, minimizing observation error, and maximizing observation stability. Simplifying the problem into a single-objective optimization problem can lead to a loss of diversity and fail to fully consider the trade-offs between multiple objectives. Furthermore, compared to traditional "satellite-to-ground" coverage observations, "satellite-to-satellite-to-ground" occultation observations involve more objects and suffer from poor observational stability. Traditional revisit constellation design methods are not well adapted to this situation and cannot guarantee stable observations at different times and locations. Therefore, it is highly meaningful to propose a revisit occultation constellation design method based on multi-objective optimization. Summary of the Invention

[0004] To address the problems of diversity loss in existing occultation constellation optimization design and the inapplicability of traditional revisit constellation design methods, the present invention provides a revisit occultation constellation design method based on multi-objective optimization, the method comprising:

[0005] A revisit occultation constellation design method based on multi-objective optimization, the method comprising:

[0006] Step 1: Collect the area to be observed and divide the area into grids to obtain a gridded regional information dataset;

[0007] Step 2: Determine the value range of the decision variable of the satellite constellation according to the rasterized regional information dataset;

[0008] Step 3: Calculate the regression orbit pair set and the corresponding regression time within the orbit altitude value range according to the value range of the decision variable of the satellite constellation;

[0009] Step 4: Determine the optimization objective and fitness function based on the regression trajectory pair set and the corresponding regression time;

[0010] Step 5: According to the optimization objective and fitness function, set the parameter values ​​of the optimized MOEA\D algorithm, and randomly generate the initial solution of the population within the value range of the decision variable;

[0011] Step 6: Based on the revisit time of the track pair and the initial solution of the population, the fitness is obtained by simulation;

[0012] Step 7: Optimize the iterative population solution based on the MOEA / D algorithm, and update the non-dominated solution set based on the fitness. The non-dominated solution set is used to design the revisited occultation constellation. If the optimization end condition is not met, return to step 6.

[0013] Furthermore, in step 1, the area to be observed is an area surrounded by four coordinate points, specifically:

[0014] X={(lon1,lat1),(lon1,lat2),(lon2,lat1),(lon2,lat2)},

[0015] (lon1,lat1) represents the longitude and latitude corresponding to the first coordinate point, (lon1,lat2) represents the longitude and latitude corresponding to the second coordinate point, (lon2,lat1) represents the longitude and latitude corresponding to the third coordinate point, and (lon2,lat2) represents the longitude and latitude corresponding to the fourth coordinate point.

[0016] Furthermore, in step 2, the value ranges of the decision variables for determining the satellite constellation are specifically as follows: the occultation constellation is set to a Walker configuration, the number of constellation satellites is determined to be a constant value; and the value ranges of the orbital inclination, orbital altitude, number of orbital planes, constellation phase, and ascending node right ascension difference are determined.

[0017] Furthermore, in step 3, the orbit height satisfies a regression period constraint, which needs to satisfy:

[0018] d e T e =k t T t =k r T r

[0019] Where, T t is the orbital period of the source satellite, T ris the orbital period of the observation satellite, T e is the Earth's rotation period, d e 、k t 、k r are all positive integers, that is, during the Earth's rotation d e At that time, the source satellite turns k t Circle, observation satellite rotates k r lock up.

[0020] Furthermore, in step 3, the method for calculating the set of regression track pairs is: setting a maximum regression day and searching for revisited track pairs within the range;

[0021] The steps of searching for revisit orbit pairs are as follows: first calculate the single satellite return orbit height set Htotal, and traverse d e With k i The selection condition is the orbit height h i Satisfy the single star coverage observation return days constraint,

[0022]

[0023] Then for each orbit height h in Htotal i , calculate the height set S of its adjacent orbits hi , the selection range is other orbitals in the Htotal set, and the regression period of the orbital pair is

[0024] d r =[d i ,d j ]≤d emax

[0025] Get, where d r is the regression period of the orbital pair in days, d i The orbit height is h i The number of days of return, d j is the orbit height h j The number of days of return, d emax The maximum number of days to return.

[0026] Furthermore, in step 4, the optimization objectives include: the number of occultation events, the uniformity of occultation events, and the revisit performance;

[0027] The fitness function includes: a fitness function of the number of occultation events, a fitness function of the uniformity of occultation events, and a fitness function of revisit performance;

[0028] The fitness function for the number of occultation events is given by:

[0029] f1=kn0 / T

[0030] Get, where k is the coefficient, n o is the sum of the number of occultation events during the simulation time, and T is the simulation time;

[0031] The fitness function for the uniformity of occultation events is given by:

[0032] f1=-σ grid / μ grid -σ lat / μ lat

[0033] Obtain, where μ grid is the mean value of the grid distribution, μ lat is the mean value of the dimension band distribution, σ grid is the standard deviation of the grid distribution, σ lat is the standard deviation of the distribution by dimension band;

[0034] Revisit the performance fitness function by:

[0035]

[0036] Get, where n ogrid is the number of grids observed during the simulation time, t cg is the average revisit time of a single grid.

[0037] Furthermore, in step 5, the weight vector of the MOEA\D algorithm is:

[0038] λ i =(λ i1 ,λ i2 ,λ i3 )

[0039] Obtained, each component of the weight vector is from the set H λ =Select from {0 / H, 1 / H…H / H}.

[0040] Furthermore, in step 5, the random generation of the population initial solution is specifically as follows: setting parameter H, correspondingly determining the population value N, setting the domain value parameter T, the number of iterations G, the mutation operator F, the real number encoding of the decision variable, randomly generating the initial solution, and rounding or rounding the index if there is a discrete value during decoding.

[0041] Furthermore, the step 6 is specifically as follows: taking the regression period of the orbit pair as the simulation time of each solution in the population, using the forward occultation simulation algorithm to perform simulation, and calculating the fitness of the solution according to step 4.

[0042] Furthermore, in step 7, the optimization iterative population solution and the update of the non-dominated solution set are specifically as follows: if not initialized, the current reference point z* and the elite population EP are initialized, individuals are selected from the neighborhood for crossover mutation to generate new individuals y, and if y does not meet the constraints, it is mutated again until the constraints are met, and the population is searched and compared in the neighborhood to update the reference point z* and the elite population EP.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] (1) The present invention realizes an occultation constellation design method based on a multi-objective optimization algorithm, which solves the problems of single optimization target, loss of diversity, and dependence on target weight in the existing occultation constellation optimization design.

[0045] (2) The present invention realizes the revisit occultation constellation design and solves the problem that the traditional coverage observation regression model is not applicable to the occultation observation regression orbit design. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 The present invention is a flow chart of a revisit occultation constellation design method based on multi-objective optimization, which is an implementation method. DETAILED DESCRIPTION

[0047] Implementation method 1: Combined Figure 1 This embodiment describes a method for designing a revisited occultation constellation based on multi-objective optimization, the method comprising:

[0048] Step 1: Collect the area to be observed and divide the area into grids to obtain a gridded regional information dataset;

[0049] Step 2: Determine the value range of the decision variable of the satellite constellation according to the rasterized regional information dataset;

[0050] Step 3: Calculate the regression orbit pair set and the corresponding regression time within the orbit altitude value range according to the value range of the decision variable of the satellite constellation;

[0051] Step 4: Determine the optimization objective and fitness function based on the regression trajectory pair set and the corresponding regression time;

[0052] Step 5: According to the optimization objective and fitness function, set the parameter values ​​of the optimized MOEA\D algorithm, and randomly generate the initial solution of the population within the value range of the decision variable;

[0053] Step 6: Based on the revisit time of the track pair and the initial solution of the population, the fitness is obtained by simulation;

[0054] Step 7: Optimize the iterative population solution based on the MOEA / D algorithm, and update the non-dominated solution set based on the fitness. The non-dominated solution set is used to design the revisited occultation constellation. If the optimization end condition is not met, return to step 6.

[0055] A multi-objective optimization algorithm (MOEA / D) is used to search and optimize satellite constellation designs. Using a gridded regional information dataset and the range of decision variables, combined with the calculation of regression orbit pairs and the evaluation of a fitness function, a set of non-dominated solutions is generated. Through continuous iteration and updating of the non-dominated solution set, an optimal satellite constellation design is ultimately found to enable revisit occultation observations within the desired observation area.

[0056] Implementation 2: This implementation further limits the multi-objective optimization-based revisit occultation constellation design method described in Implementation 1. In this implementation, in step 1, the area to be observed is an area surrounded by four coordinate points, specifically:

[0057] X={(lon1,lat1),(lon1,lat2),(lon2,lat1),(lon2,lat2)},

[0058] (lon1,lat1) represents the longitude and latitude corresponding to the first coordinate point, (lon1,lat2) represents the longitude and latitude corresponding to the second coordinate point, (lon2,lat1) represents the longitude and latitude corresponding to the third coordinate point, and (lon2,lat2) represents the longitude and latitude corresponding to the fourth coordinate point.

[0059] The scope of observation or measurement is limited by determining the boundaries of the area for subsequent data collection, analysis or other processing.

[0060] Implementation method three: This implementation method further limits the revisit occultation constellation design method based on multi-objective optimization described in implementation method one. In this implementation method, in step 2, the value range of the decision variable for determining the satellite constellation is specifically as follows: the occultation constellation is taken as a Walker configuration, and the number of constellation satellites is determined to be a constant value; and the value range of the orbital inclination, orbital altitude, number of orbital planes, constellation phase, and ascending node right ascension difference is determined.

[0061] Implementation 4: This implementation further limits the multi-objective optimization-based revisit occultation constellation design method described in Implementation 1. In this implementation, in step 3, the orbital altitude satisfies the regression period constraint, which needs to satisfy:

[0062] d e T e =k t Tt =k r T r

[0063] Where, T t is the orbital period of the source satellite, T r is the orbital period of the observation satellite, T e is the Earth's rotation period, d e 、k t 、k r are all positive integers, that is, during the Earth's rotation d e At that time, the source satellite turns k t Circle, observation satellite rotates k r lock up.

[0064] Implementation 5: This implementation further limits the revisit occultation constellation design method based on multi-objective optimization described in Implementation 1. In this implementation, in step 3, the method for calculating the set of regression orbit pairs is: setting a maximum regression day and searching for revisit orbit pairs within this range;

[0065] The steps of searching for revisit orbit pairs are as follows: first calculate the single satellite return orbit height set Htotal, and traverse d e With k i The selection condition is the orbit height h i Satisfy the single star coverage observation regression day constraint,

[0066]

[0067] Then for each orbit height h in Htotal i , calculate the height set S of its adjacent orbits hi , the selection range is other orbitals in the Htotal set, and the regression period of the orbital pair is

[0068] d r =[d i ,d j ]≤d emax

[0069] Get, where d r is the regression period of the orbital pair in days, d i The orbit height is h i The number of days of return, d j is the orbit height h j The number of days of return, d emax The maximum number of days to return.

[0070] Implementation 6: This implementation further limits the multi-objective optimization-based revisit occultation constellation design method described in Implementation 1. In step 4, the optimization objectives include: the number of occultation events, the uniformity of occultation events, and the revisit performance.

[0071] The fitness function includes: a fitness function of the number of occultation events, a fitness function of the uniformity of occultation events, and a fitness function of revisit performance;

[0072] The fitness function for the number of occultation events is given by:

[0073] f1=kn0 / T

[0074] Get, where k is the coefficient, n o is the sum of the number of occultation events during the simulation time, and T is the simulation time;

[0075] The fitness function for the uniformity of occultation events is given by:

[0076] f1=-σ grid / μ grid -σ lat / μ lat

[0077] Obtain, where μ grid is the mean value of the grid distribution, μ lat is the mean value of the dimension band distribution, σ grid is the standard deviation of the grid distribution, σ lat is the standard deviation of the distribution by dimension band;

[0078] Revisit the performance fitness function by:

[0079]

[0080] Get, where n ogrid is the number of grids observed during the simulation time, t cg is the average revisit time of a single grid.

[0081] Implementation 7: This implementation further limits the multi-objective optimization-based revisit occultation constellation design method described in Implementation 1. In this implementation, in step 5, the weight vector of the MOEA\D algorithm is obtained by:

[0082] λ i =(λ i1 ,λ i2 ,λ i3 )

[0083] Obtained, each component of the weight vector is from the set H λ=Select from {0 / H, 1 / H…H / H}.

[0084] Implementation method eight: This implementation method further limits the revisit occultation constellation design method based on multi-objective optimization described in implementation method one. In this implementation method, in step 5, the random generation of the population initial solution is specifically as follows: setting parameter H, correspondingly determining the population value N, setting the domain value parameter T, the number of iterations G, the mutation operator F, the real number encoding of the decision variable, randomly generating the initial solution, and rounding or rounding the index if there is a discrete value during decoding.

[0085] Implementation method nine: This implementation method further limits the revisit occultation constellation design method based on multi-objective optimization described in implementation method one. In this implementation method, step 6 specifically comprises taking the regression period of the orbit pair as the simulation time of each solution in the population, using the forward occultation simulation algorithm for simulation, and calculating the fitness of the solution according to step 4.

[0086] Implementation method ten: This implementation method further limits the revisit occultation constellation design method based on multi-objective optimization described in implementation method one. In this implementation method, in step 7, the optimization iterative population solution and the update of the non-dominated solution set are specifically as follows: if not initialized, the current reference point z* and the elite population EP are initialized, individuals are selected from the neighborhood for cross-mutation to generate new individuals y, and if y does not meet the constraints, it is mutated again until the constraints are met, and the population is searched and compared in the neighborhood to update, and the reference point z* and the elite population EP are updated.

[0087] Implementation eleven: The above implementation is described below with examples.

[0088] A revisit occultation constellation design method based on multi-objective optimization, the design method comprising the following steps:

[0089] Step 1: Enter the location of the area to be observed and divide it into grids;

[0090] The region position is set to X = {(-180°, 67.5°), (-180°, -67.5°), (180°, 67.5°), (180°, -67.5°)}. The region is also divided into a grid with a longitude interval of 4.5° and a latitude interval of 4.5°.

[0091] Step 2: Determine the value range of decision variables such as satellite constellation configuration parameters;

[0092] The number of source satellites is set to 12, and the number of receiving satellites is set to 3. The number of receiving satellites is small, so the number of receiving satellite orbital planes is set to 3 to ensure uniformity. The decision variable is the source constellation orbit inclination i t , source constellation orbit height h t , the number of source satellite orbital planes Pt , source satellite constellation phase F t , observe the constellation orbit inclination i r , observing constellation orbit height h r , observing satellite constellation phase F r , the right ascension difference of the ascending node Δ Ω .

[0093] The value range of the optimization variable is:

[0094]

[0095] Where Htotal and Sh are calculated by step 3, and the orbit height h satisfies:

[0096] 500km <h<1500km h∈H total US h

[0097] Step 3: Calculate the set of regression orbit pairs and the corresponding revisit time within the range of orbital altitude values;

[0098] Set the maximum number of return days d emax It is 10 days, and the regression orbit pair set and the corresponding regression time are calculated according to the formula.

[0099] Step 4: Determine the optimization objective and fitness function;

[0100] The optimization objectives are the number of occultation events, the uniformity of occultation events, and the revisit performance, where the uniformity of occultation events is specifically set as:

[0101] f1=2520n0 / T

[0102] Where, the actual meaning is the number of occultation events in the simulation of 2520 days, excluding the influence of different simulation times.

[0103] Step 5: Set the optimization algorithm parameter values ​​and randomly generate the initial solution of the population within the range of decision variables;

[0104] Assume that the parameter H is 13, the population size N is 105, the field size parameter T is 10, the iteration number G is 100, and the mutation operator F is 0.1.

[0105] Step 6: Based on the revisit time of the track pair, the fitness is obtained through simulation;

[0106] Step 7: Optimize the iterative population solution based on the MOEA / D algorithm and update the non-dominated solution set. If the optimization end condition is not met, return to step 6.

[0107] The fitness and decision variables of some non-dominated solution sets obtained by multi-objective optimization are shown in Tables 1 and 2.

[0108] The present invention proposes a revisit occultation constellation design method based on multi-objective optimization, which can be applied to occultation constellation design problems with multiple optimization objectives, especially revisit occultation constellation design based on regression orbits.

[0109] Table 1. Non-dominated solution fitness

[0110]

[0111] Table 2. Non-dominated solution decision variables

[0112]

Claims

1. A revisit occultation constellation design method based on multi-objective optimization, characterized in that: The method comprises: Step 1: Collect the area to be observed and divide the area into grids to obtain a gridded regional information dataset; Step 2: Determine the value range of the decision variable of the satellite constellation according to the rasterized regional information dataset; Step 3: Calculate the regression orbit pair set and the corresponding regression time within the orbit altitude value range according to the value range of the decision variable of the satellite constellation; Step 4: Determine the optimization objective and fitness function based on the regression trajectory pair set and the corresponding regression time; Step 5: According to the optimization objective and fitness function, set the parameter values ​​of the optimized MOEA\D algorithm, and randomly generate the initial solution of the population within the value range of the decision variable; Step 6: Based on the revisit time of the track pair and the initial solution of the population, the fitness is obtained by simulation; Step 7: Optimize the iterative population solution based on the MOEA / D algorithm, and update the non-dominated solution set based on the fitness. The non-dominated solution set is used to design the revisited occultation constellation. If the optimization end condition is not met, return to step 6. In step 3, the method for calculating the set of revisited orbit pairs is: setting a maximum number of revisited days and searching for revisited orbit pairs within the range; The steps of searching for revisit orbit pairs are: first calculate the single satellite return orbit height set Htotal , while satisfying the height constraint, traverse d e and k i The selection criteria are track height h i Satisfy the single star coverage observation regression day constraint, Again Htotal Each track height h i , calculate the set of adjacent orbital heights S hi , the selection range is Htotal For the other orbitals in the set, the return period of the orbital pair is given by Obtain, among which, d r is the regression period of the orbital pair in days, d i The orbit height is h i The number of days of return, d j is the orbit height h j The number of days of return, d emax is the longest return days; In step 4, the optimization objectives include: the number of occultation events, the uniformity of occultation events, and the revisit performance; The fitness function includes: a fitness function of the number of occultation events, a fitness function of the uniformity of occultation events, and a fitness function of revisit performance; The fitness function for the number of occultation events is given by: Obtain, among which, k is the coefficient, n o is the sum of the number of occultation events during the simulation time, T is the simulation time; The fitness function for the uniformity of occultation events is given by: Obtain, among which, μ grid is the average value of the grid distribution, μ lat is the average value of the distribution by dimension, σ grid is the standard deviation of the grid distribution, σ lat is the standard deviation of the distribution by dimension band; Revisit the performance fitness function by: Obtain, among which, n ogrid is the number of grids observed during the simulation time, t cg is the average revisit time of a single grid.

2. The method for designing a revisited occultation constellation based on multi-objective optimization according to claim 1, characterized in that: In step 1, the area to be observed is an area surrounded by four coordinate points, specifically: X={(lon1,lat1),(lon1,lat2),(lon2,lat1),(lon2,lat2)} ， (lon1,lat1) Indicates the longitude and latitude corresponding to the first coordinate point, (lon1,lat2) Indicates the longitude and latitude corresponding to the second coordinate point, (lon2,lat1) Indicates the longitude and latitude corresponding to the third point coordinate point, (lon2,lat2) Indicates the longitude and latitude corresponding to the fourth coordinate point.

3. The method for designing a revisited occultation constellation based on multi-objective optimization according to claim 1, characterized in that: In step 2, the value ranges of the decision variables for determining the satellite constellation are specifically as follows: setting the occultation constellation to a Walker configuration and determining the number of constellation satellites to be a constant; and determining the value ranges of the orbital inclination, orbital altitude, number of orbital planes, constellation phase, and ascending node right ascension difference.

4. The method for designing a revisited occultation constellation based on multi-objective optimization according to claim 1, wherein: In step 3, the orbital height satisfies the regression period constraint, which needs to satisfy: Where, T t is the orbital period of the source satellite, T r is the orbital period of the observation satellite, T e is the Earth's rotation period, d e 、 k t 、k r are all positive integers, that is, when the Earth rotates d e When the weather is right, the source satellite turns k t Circle, observe the satellite rotation k r lock up.

5. The method for designing a revisited occultation constellation based on multi-objective optimization according to claim 1, wherein: In step 5, the weight vector of the MOEA\D algorithm is: λ i =(λ i1 , λ i2 , λ i3 ) Obtained, each component of the weight vector is from the set H λ ={0 / H, 1 / H…H / H} Select from .

6. The method for designing a revisited occultation constellation based on multi-objective optimization according to claim 1, characterized in that: In step 5, the random population initial solution is specifically: setting parameters H , and the corresponding population value is determined N , set the domain value parameters T , number of iterations G , mutation operator F， The decision variables are encoded with real numbers, and the initial solution is generated randomly. If there are discrete values ​​during decoding, they are rounded or indexed.

7. The method for designing a revisited occultation constellation based on multi-objective optimization according to claim 1, wherein: Specifically, step 6 is to use the regression period of the orbit pair as the simulation time of each solution in the population, perform simulation using the forward occultation simulation algorithm, and calculate the fitness of the solution according to step 4.

8. The method for designing a revisited occultation constellation based on multi-objective optimization according to claim 1, wherein: In step 7, the optimization iterative population solution and the update of the non-dominated solution set are specifically as follows: if not initialized, initialize the current reference point z* and elite populations EP, Select individuals from the neighborhood to generate new individuals through crossover mutation y , y If the constraints are not met, mutate again until the constraints are met, search and compare in the neighborhood to update the population, and update the reference point. z* and elite populations EP .

Images