A random scenario construction method considering temporal autocorrelation and cross-correlation

Abstract

The present invention proposes a random scenario construction method that considers temporal autocorrelation and mutual correlation. The method comprises the following steps: constructing a probability distribution model f for each of q uncertain factors, using Monte Carlo random sampling to obtain N initial random scenarios for each of the q uncertain factors; calculating the N random scenarios with temporal autocorrelation corresponding to each uncertain factor; and using a particle swarm algorithm with linearly decreasing inertia weights to select the random scenario with the minimum mutual correlation error from the N random scenarios with temporal autocorrelation obtained. Finally, a random scenario that comprehensively considers temporal autocorrelation and mutual correlation and conforms to real-world conditions is constructed. By analyzing the mutual correlations between multiple uncertain factors and the temporal autocorrelation within each uncertainty factor, the method further refines uncertainty modeling, making the random scenarios generated using the construction method more consistent with real-world conditions.

Description

Technical Field

[0001] The present invention relates to the field of uncertainty technology of integrated energy systems, and specifically to a random scenario construction method that takes into account time series autocorrelation and cross-correlation. Background Art

[0002] Under the dual pressures of energy and environmental crises, the integrated energy system, with the power system as its core, breaks the existing model of separate planning, design and operation of gas, heating, cooling and other energy systems, and conducts integrated planning, design and operation optimization of the social energy system, and ultimately builds a unified social integrated energy system to improve energy utilization efficiency.

[0003] However, traditional integrated energy system planning and operation are usually based on deterministic planning scenarios, without considering the uncertainty of renewable energy such as wind and solar power, and the uncertainty of energy consumption of user loads. The results obtained do not conform to the actual scenario and may not be the optimal results during its actual operation. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention proposes a random scenario construction method that takes into account temporal autocorrelation and cross-correlation. By analyzing the cross-correlation between multiple uncertainty factors and the temporal autocorrelation within each uncertainty factor, the generated random scenario is made more consistent with the real scenario situation.

[0005] To achieve the above-mentioned purpose, the present invention designs a random scenario construction method considering time series autocorrelation and cross-correlation, which is particularly characterized by comprising the following steps:

[0006] Step 1: Based on the real data of q uncertain factors corresponding to each season of the year, calculate the reference time series autocorrelation coefficient matrix S of each q uncertain factors corresponding to each season in the real scenario. s,q,ref , and the reference cross-correlation coefficient matrix ρ between the q uncertain factors s,ref ;

[0007] Step 2: Construct a probability distribution model f for each of the q uncertain factors corresponding to each season of the year. Based on the probability distribution of the q uncertain factors in each season of the year, use Monte Carlo random sampling to obtain N initial random scenarios for each of the q uncertain factors corresponding to each season of the year.

[0008] Step 3: Calculate the time series autocorrelation coefficient matrix S of each initial random scenario for each of the q uncertain factors corresponding to each season of the year s,q , for each time series autocorrelation coefficient matrix S s,qPerform Cholesky decomposition to obtain the lower triangular matrix D, and then obtain D -1 , D -1 The temporal autocorrelation caused by random sampling is eliminated. Based on the probability distribution model f of each uncertainty factor corresponding to each season of the year, a matrix F consisting of N probability distribution sequences corresponding to each uncertainty factor is constructed. The lower triangular matrix D is multiplied by the matrix F to obtain the matrix G, which is a matrix consisting of N random scenarios corresponding to each uncertainty factor after eliminating the temporal autocorrelation caused by random sampling.

[0009] Step 4: For each reference time series autocorrelation coefficient matrix S in step 1 s,q,ref Perform Cholesky decomposition to obtain the lower triangular matrix D ref , the lower triangular matrix D ref Multiply it with matrix G to obtain the matrix G that conforms to the actual situation auto , G auto A matrix consisting of N random scenarios with temporal autocorrelation corresponding to each uncertain factor;

[0010] Step 5, for G auto The N random scenes with temporal autocorrelation in the algorithm are numbered as 1, 2, 3, ..., N, and a random scene with temporal autocorrelation is extracted from each uncertainty factor corresponding to each season of the year as a sample group to form an initial particle with D′ dimension. The dimension D′ of each initial particle is: the number of seasons in the year s×the number of uncertainty factors q; the particle swarm algorithm with linearly decreasing inertia weight is used for the N initial particles to obtain the j-th dimension velocity of the i-th initial particle, thereby calculating the mutual correlation coefficient matrix ρ between the q uncertainty factors s ;

[0011] Step 6: Calculate the mutual correlation coefficient matrix ρ between the q uncertain factors in step 5 through the fitness function s The reference cross-correlation coefficient matrix ρ between the q uncertain factors in step 1 s,ref The minimum cross-correlation error between them is then selected from the N random scenarios with temporal autocorrelation, and finally a random scenario that comprehensively considers temporal autocorrelation and cross-correlation and conforms to the actual situation is constructed.

[0012] The advantages of the present invention are:

[0013] 1. This invention considers multiple random scenarios of wind speed, light intensity, electrical load, and thermal load in the integrated energy system. Specifically, it makes uncertainty modeling more refined by considering the temporal autocorrelation of wind speed, light intensity, electrical load, and thermal load and the cross-correlation among them.

[0014] 2. Compared with the existing method for constructing random source-load scenarios, the random scenarios generated by the method proposed in this invention are more consistent with real-world scenarios, and the results obtained when they are used as data sets to input into the problem of integrated energy system planning are more accurate.

[0015] The random scenario construction method of the present invention takes into account temporal autocorrelation and cross-correlation by analyzing the cross-correlation between multiple uncertainty factors and the temporal autocorrelation within each uncertainty factor, making uncertainty modeling more refined. The random scenario generated by the construction method is more consistent with the actual scenario situation. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Schematic diagram comparing the wind speeds generated by the three random scenario construction methods with the actual historical wind speeds;

[0017] Figure 2 Schematic diagram comparing the illumination intensity generated by the three random scene construction methods with the actual historical illumination intensity;

[0018] Figure 3 Schematic diagram comparing the electric load generated by the three random scenario construction methods with the actual historical electric load;

[0019] Figure 4 Schematic diagram comparing the heat loads generated for the three random scenario construction methods with the actual historical heat loads; DETAILED DESCRIPTION

[0020] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] The integrated energy system in the following embodiments includes four uncertain factors: wind speed, light intensity, electrical load, and thermal load.

[0022] The present invention provides a random scenario construction method taking into account time series autocorrelation and cross-correlation, comprising the following steps:

[0023] Step 1: Based on the real data of q uncertain factors corresponding to each season of the year, calculate the reference time series autocorrelation coefficient matrix S of each q uncertain factors corresponding to each season in the real scenario. s,q,ref , and the reference cross-correlation coefficient matrix ρ between the q uncertain factors s,ref .

[0024] Time series autocorrelation describes the degree of correlation between values at different moments in a time series, while cross-correlation describes the degree of correlation between different time series.

[0025] The reference time series autocorrelation coefficient matrix S s,q,ref The calculation formula is

[0026]

[0027] in,

[0028] A cf (i,i+k)=A cf (i+k,i), that is, S is a symmetric matrix with all diagonals set to 1,

[0029] S represents the reference time series autocorrelation coefficient matrix S s,q,ref ,

[0030] A cf (i,i+k) represents two moment values x in a time series i and x i+k The serial correlation coefficient between

[0031] M represents the time length of a time series;

[0032] Two moment values x i and x i+k The serial correlation coefficient A cf The calculation formula for (i,i+k) is

[0033]

[0034] in,

[0035] M represents the time length of the time series.

[0036] μ represents the mean of the time series,

[0037] When k is smaller, it means the correlation is stronger.

[0038] When k = 0, A cf (i,i)=1, indicating that the serial correlation between each uncertainty factor and itself is 1.

[0039] The reference cross-correlation coefficient matrix ρ s and ρ s,ref The calculation formula is

[0040]

[0041] in,

[0042] ρ represents the reference cross-correlation coefficient matrix ρ s,ref , or the cross-correlation coefficient matrix ρ s ,

[0043] ρ(X,Y) represents the Pearson correlation coefficient between two uncertain factors X and Y.

[0044] q represents the number of uncertain factors.

[0045] In order to better describe the mutual correlation between non-normally distributed random variables, the Spearman rank correlation coefficient is introduced. The specific approach is to first sort the uncertain factors (i.e., wind speed, light intensity, electrical load, or heat load) from small to large, calculate the rank of the uncertain factors, and then calculate the Pearson correlation coefficient of the rank.

[0046] The calculation formula of the Pearson correlation coefficient ρ(X,Y) between two uncertain factors X and Y is

[0047]

[0048] in,

[0049] The closer |ρ(X,Y)| is to 1, the stronger the correlation between the two uncertain factors X and Y is.

[0050] cov(X,Y) represents the covariance between two uncertain factors X and Y.

[0051] σ(X) represents the standard deviation of the uncertainty factor X,

[0052] σ(Y) represents the standard deviation of the uncertainty factor Y.

[0053] Step 2: Construct a probability distribution model f for each of the q uncertain factors corresponding to each season of the year. Based on the probability distribution of the q uncertain factors in each season of the year, use Monte Carlo random sampling to obtain N initial random scenarios for each of the q uncertain factors corresponding to each season of the year.

[0054] Specifically, q takes the value of 4. The four uncertain factors include wind speed, light intensity, electrical load, and heat load. In addition, to make the generated random scenarios more realistic, the wind speed, light intensity, electrical load, and heat load of the year are divided into transition season, summer, and winter for processing.

[0055] The randomness of the wind speed obeys the Weibull distribution, and the probability density function of the wind speed in each season is

[0056]

[0057] in,

[0058] v(s,t) represents the actual wind speed at time t in season s.

[0059] a(s,t) represents the scale parameter at time t in season s,

[0060] b(s,t) represents the shape parameter at time t in season s,

[0061] σ(s,t) represents the standard deviation of wind speed at time t in season s,

[0062] represents the mean wind speed at time t in season s.

[0063] The randomness of the light intensity follows the beta distribution, and the probability density function of the light intensity in each season is:

[0064]

[0065] in,

[0066] g(s,t) represents the per-unit value of light intensity at time t in season s.

[0067] α(s,t) represents the shape parameter that is positive at time t in season s.

[0068] β(s,t) represents the shape parameter that is positive at time t in season s.

[0069] μ(s,t) represents the per-unit mean light intensity at time t in season s.

[0070] represents the standard deviation of light intensity at time t in season s.

[0071] The randomness of the electric load and heat load obeys the normal distribution and is considered to be within the 95% confidence interval. The probability density function of the electric load and heat load in each season is:

[0072]

[0073] That

[0074] in,

[0075] L(m,s,t) represents the load size of the mth load at time t in season s.

[0076] μ(m,s,t) represents the mean value of the mth load at time t in season s.

[0077] μ(m,s,t) represents the standard deviation of the mth load at time t in season s.

[0078] Step 3: Calculate the time series autocorrelation coefficient matrix S of each initial random scenario for each of the q uncertain factors corresponding to each season of the year s,q , for each time series autocorrelation coefficient matrix S s,q Perform Cholesky decomposition to obtain the lower triangular matrix D, and then obtain D -1 , D -1 The temporal autocorrelation caused by random sampling is eliminated; based on the probability distribution model f of each uncertain factor corresponding to each season of the year, a matrix F consisting of N probability distribution sequences corresponding to each uncertain factor is constructed; the lower triangular matrix D is multiplied by the matrix F to obtain the matrix G, which is a matrix consisting of N random scenarios corresponding to each uncertain factor after eliminating the temporal autocorrelation caused by random sampling.

[0079] The time series autocorrelation coefficient matrix S s,q The calculation formula is the reference time series autocorrelation coefficient matrix S in step 1 above. s,q,ref The calculation formula is the same.

[0080] Specifically, for each time series autocorrelation coefficient matrix S s,q The specific form of Cholesky decomposition is

[0081] S s,q =DD T

[0082] in,

[0083] S s,q Represents the autocorrelation coefficient matrix of each time series,

[0084] D means that by s,q Perform Cholesky decomposition on the lower triangular matrix.

[0085] Specifically, the calculation formula of matrix G is:

[0086] G=D -1 F

[0087] in,

[0088] D -1 represents the inverse matrix of the lower triangular matrix D,

[0089] F represents a matrix consisting of N probability distribution sequences corresponding to each uncertain factor.

[0090] Step 4: For each reference time series autocorrelation coefficient matrix S in step 1 s,q,ref Perform Cholesky decomposition to obtain the lower triangular matrix D ref , the lower triangular matrix D refMultiply it with matrix G to obtain the matrix G that conforms to the actual situation auto , G auto A matrix consisting of N random scenarios with temporal autocorrelation corresponding to each uncertain factor.

[0091] For each reference time series autocorrelation coefficient matrix S s,q,ref The specific form of Cholesky decomposition is

[0092]

[0093] in,

[0094] S s,q,ref Represents the autocorrelation coefficient matrix of each reference time series,

[0095] D ref Indicates that by S s,q,ref Perform Cholesky decomposition on the lower triangular matrix.

[0096] Specifically, the matrix G auto The calculation formula is

[0097] G auto =D ref G

[0098] in,

[0099] D ref Indicates that by S s,q,ref Perform Cholesky decomposition on the lower triangular matrix,

[0100] G represents a matrix consisting of N random scenarios corresponding to each uncertain factor after eliminating the temporal autocorrelation generated by random sampling.

[0101] Step 5, for G auto The N random scenes with temporal autocorrelation in the algorithm are numbered as 1, 2, 3, ..., N, and a random scene with temporal autocorrelation is extracted from each uncertainty factor corresponding to each season of the year as a sample group to form an initial particle with D′ dimension. The dimension D′ of each initial particle is: the number of seasons in the year s×the number of uncertainty factors q; the particle swarm algorithm with linearly decreasing inertia weight is used for the N initial particles to obtain the j-th dimension velocity of the i-th initial particle, thereby calculating the mutual correlation coefficient matrix ρ between the q uncertainty factors s .

[0102] Specifically, the velocity of the jth dimension of the i-th initial particle is calculated as follows:

[0103] v(i,j)=ω(t)·v(i,j)+c1·rand()·(pbest(i,j)-x(i,j))+c2·rand()·(gbest(i,j)-x(i,j))

[0104] in,

[0105] v(i, j) represents the velocity of the jth dimension of the i-th initial particle, 1≤i≤N,1≤j≤D′,

[0106] ω(t) represents the inertia weight at the t-th iteration,

[0107] c1 is the learning factor of the particle swarm algorithm,

[0108] c2 is the learning factor of the particle swarm algorithm,

[0109] pbest(i,j) represents the historical optimal position of the initial particle i in dimension j.

[0110] gbest(i,j) represents the global optimal position of the initial particle i in dimension j.

[0111] x(i,j) is the current position of the initial particle i in the j dimension.

[0112] The initial velocity is set to a random integer to ensure that the updated particles are valid. To balance the global search performance and local search accuracy of the particle swarm algorithm, a larger inertia weight can be set in the initial iteration. From the velocity formula of the jth dimension of the i-th initial particle, it can be seen that the particle velocity is relatively large in the initial iteration, which has good global search capabilities, but weak local search capabilities. As the number of iterations accumulates, the value of the inertia weight ω(t) decreases, and the particle velocity also decreases. At this time, the particle has good local search capabilities, but weak global search capabilities.

[0113] The calculation formula of the inertia weight ω(t) is:

[0114]

[0115] in,

[0116] ω(t) represents the inertia weight at the t-th iteration,

[0117] ω max represents the upper limit of the inertia weight change,

[0118] ω min represents the lower limit of the inertia weight change,

[0119] t max Indicates the maximum number of iterations.

[0120] Since the setting of the inertia weight ω(t) is closely related to the calculation accuracy of the particle swarm algorithm, when the inertia weight ω(t) is large, the particle swarm algorithm is conducive to jumping out of the local minimum and facilitating global search; when the inertia weight ω(t) is small, the particle swarm algorithm is conducive to accurate search in local areas and facilitates convergence. Therefore, by adopting the particle swarm algorithm based on linear decrease of inertia weight, the globality of the particle swarm algorithm search and the accuracy of local search are taken into account.

[0121] The cross-correlation coefficient matrix ρ s The calculation formula is the same as the reference cross-correlation coefficient matrix ρ in step 1 s,ref The calculation formula is the same.

[0122] Specifically, the q uncertain factors include wind speed, light intensity, electrical load, and heat load, that is, the uncertain factor q is 4. There are transition seasons, summer, and winter in a year, and the number of seasons s is 3, so the dimension D′ of each initial particle is 12.

[0123] Step 6: Calculate the mutual correlation coefficient matrix ρ between the q uncertain factors in step 5 through the fitness function s The reference cross-correlation coefficient matrix ρ between the q uncertain factors in step 1 s,ref The minimum cross-correlation error between them is then selected from the N random scenarios with temporal autocorrelation, and finally a random scenario that comprehensively considers temporal autocorrelation and cross-correlation and conforms to the actual situation is constructed.

[0124] Although the N random scenarios with temporal autocorrelation corresponding to the uncertain factors in step 4 each meet the temporal autocorrelation characteristics of the real scenario, the uncertain factors may not meet the mutual correlation characteristics of the real scenario. Therefore, the fitness function is set to the generated mutual correlation coefficient matrix ρ between wind speed, light intensity, electric load, and heat load s The reference cross-correlation coefficient matrix ρ between wind speed, light intensity, electrical load, and thermal load in real scenes s,ref The minimum cross-correlation error between .

[0125] The fitness function is expressed as

[0126]

[0127] in,

[0128] E r represents the cross-correlation error,

[0129] s max Indicates the number of seasons in a year,

[0130] q max Indicates the number of uncertain factors,

[0131] ρ h,s Represents the mutual correlation coefficient matrix between the uncertain factors of the sth season and the hth random scenario,

[0132] ρ s,ref Represents the reference cross-correlation coefficient matrix between the uncertainty factors in the sth season.

[0133] The particle swarm algorithm based on linear decrease of inertia weight is used to determine the number of random scenarios required. The random scenario with the smallest mutual correlation error can be screened out from the N random scenarios with temporal autocorrelation, and the random scenario data finally used for subsequent planning can be obtained.

[0134] In order to verify the rationality of the construction method proposed in the present invention, the Monte Carlo random scenario generation technology (MCRS), the Monte Carlo random scenario generation technology considering cross-correlation (MCMCRS), and the random scenario construction method considering temporal autocorrelation and cross-correlation (TA-MCRSGOM) of the present invention are used to generate wind speed, light intensity, electric load, and heat load. The wind speed, light intensity, electric load, and heat load generated by the three methods are compared and analyzed with the actual historical wind speed, light intensity, electric load, and heat load.

[0135] First, given the typical daily historical data of wind speed and sunlight intensity, as well as the planned data of electric load and heat load, for wind turbines, the cut-in wind speed, rated wind speed and cut-out wind speed are 2.5m / s, 10m / s and 20m / s respectively; for photovoltaics, the rated sunlight intensity is 700W / m 2 ; At the same time, it is believed that the relationship between the standard deviation and mean of wind speed and light intensity data satisfies σ(s,t)=0.25 The relationship between the standard deviation and mean of electric load and heat load satisfies 1.96σ=0.2μ. Then, 50 random scene samples are drawn by using the above three methods respectively, and the generated wind speed, light intensity, electric load and heat load are compared, as shown in the following figure. Figures 1 to 4 shown.

[0136] from Figures 1 to 4As can be seen from the figure, the random scenarios of wind, light, electricity and heat generated by the three methods are qualitatively compared. It can be found that the random scenarios obtained by the traditional MCRS method are very rough, and the generated wind, light, electricity and heat sequence changes are quite different from the real scenarios. On this basis, the MCMCRS method that considers the correlation between wind, light, electricity and heat can improve the generated random scenarios to a certain extent, but there is still a big difference with the actual typical scenarios. The method proposed in the present invention comprehensively considers the time series autocorrelation within wind, light, electricity and heat and the mutual correlation between various factors. The generated wind, light, electricity and heat sequence changes are more in line with the real scenarios. Therefore, the planning research under random scenarios using the method proposed in the present invention will be more in line with the actual situation.

[0137] The present invention analyzes uncertainty by considering the temporal autocorrelation within each uncertainty factor and the intercorrelations between them. This embodiment considers multiple random scenarios of wind speed, light intensity, electrical load, and thermal load in an integrated energy system. Specifically, by considering the temporal autocorrelations of each of these factors, as well as the intercorrelations between them, uncertainty modeling is further refined.

[0138] In addition, compared with the existing method for constructing random source-load scenarios, the random scenarios generated by the method proposed in the present invention are more consistent with real-world scenarios, and the results obtained when they are input as data sets into the problem of integrated energy system planning are more accurate.

[0139] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be considered as equivalent replacement methods and are included in the scope of protection of the present invention.

Claims

1. A random scenario construction method considering time series autocorrelation and cross-correlation, characterized in that: The steps include: Step 1: Based on the real data of q uncertain factors corresponding to each season of the year, calculate the reference time series autocorrelation coefficient matrix S of each q uncertain factors corresponding to each season in the real scenario. s,q,ref , and the reference cross-correlation coefficient matrix ρ between the q uncertain factors s,ref ; Step 2: Construct a probability distribution model f for each of the q uncertain factors corresponding to each season of the year. Based on the probability distribution of the q uncertain factors in each season of the year, use Monte Carlo random sampling to obtain N initial random scenarios for each of the q uncertain factors corresponding to each season of the year. Step 3: Calculate the time series autocorrelation coefficient matrix S of each initial random scenario for each of the q uncertain factors corresponding to each season of the year s,q , for each time series autocorrelation coefficient matrix S s,q Perform Cholesky decomposition to obtain the lower triangular matrix D, and then obtain D -1 , D -1 The temporal autocorrelation caused by random sampling is eliminated. Based on the probability distribution model f of each uncertainty factor corresponding to each season of the year, a matrix F consisting of N probability distribution sequences corresponding to each uncertainty factor is constructed. The lower triangular matrix D is multiplied by the matrix F to obtain the matrix G, which is a matrix consisting of N random scenarios corresponding to each uncertainty factor after eliminating the temporal autocorrelation caused by random sampling. Step 4: For each reference time series autocorrelation coefficient matrix S in step 1 s,q,ref Perform Cholesky decomposition to obtain the lower triangular matrix D ref , the lower triangular matrix D ref Multiply it with matrix G to obtain the matrix G that conforms to the actual situation auto , G auto A matrix consisting of N random scenarios with temporal autocorrelation corresponding to each uncertain factor; Step 5, for G auto The N random scenes with temporal autocorrelation in the algorithm are numbered as 1, 2, 3, ..., N, and a random scene with temporal autocorrelation is extracted from each uncertainty factor corresponding to each season of the year as a sample group to form an initial particle with D′ dimension. The dimension D′ of each initial particle is: the number of seasons in the year s×the number of uncertainty factors q; the particle swarm algorithm with linearly decreasing inertia weight is used for the N initial particles to obtain the j-th dimension velocity of the i-th initial particle, thereby calculating the mutual correlation coefficient matrix ρ between the q uncertainty factors s ; Step 6: Calculate the mutual correlation coefficient matrix ρ between the q uncertain factors in step 5 through the fitness function s The reference cross-correlation coefficient matrix ρ between the q uncertain factors in step 1 s,ref The minimum cross-correlation error between them is then selected from the N random scenarios with temporal autocorrelation, and finally a random scenario that comprehensively considers temporal autocorrelation and cross-correlation and conforms to the actual situation is constructed.

2. The random scenario construction method considering temporal autocorrelation and cross-correlation according to claim 1, characterized in that: The reference time series autocorrelation coefficient matrix S in step 1 s,q,ref , the time series autocorrelation coefficient matrix S in step 3 s,q The calculation formula is in, A cf (i,i+k)=A cf (i+k,i), that is, S is a symmetric matrix with all diagonals set to 1, S represents the reference time series autocorrelation coefficient matrix S s,q,ref , or time series autocorrelation coefficient matrix S s,q , A cf (i,i+k) represents two moment values x in a time series i and x i+k The serial correlation coefficient between M represents the time length of a time series; Two moment values x i and x i+k The serial correlation coefficient A cf The calculation formula for (i,i+k) is in, M represents the time length of the time series. μ represents the mean of the time series, When k is smaller, it means the correlation is stronger. When k = 0, A cf (i,i)=1, indicating that the serial correlation between each uncertainty factor and itself is 1.

3. The random scenario construction method considering temporal autocorrelation and cross-correlation according to claim 2, characterized in that: The reference cross-correlation coefficient matrix ρ in step 1 s,ref , the cross-correlation coefficient matrix ρ in step 5 s The calculation formula is in, ρ represents the reference cross-correlation coefficient matrix ρ s,ref , or the cross-correlation coefficient matrix ρ s , ρ(X,Y) represents the Pearson correlation coefficient between two uncertain factors X and Y. q represents the number of uncertain factors; The calculation formula of the Pearson correlation coefficient ρ(X,Y) between two uncertain factors X and Y is in, The closer |ρ(X,Y)| is to 1, the stronger the correlation between the two uncertain factors X and Y is. cov(X,Y) represents the covariance between two uncertain factors X and Y. σ(X) represents the standard deviation of the uncertainty factor X, σ(Y) represents the standard deviation of the uncertainty factor Y.

4. The random scenario construction method considering temporal autocorrelation and cross-correlation according to claim 1, characterized in that: In step 2, the q uncertain factors include wind speed, light intensity, electrical load, and heat load, wherein the randomness of the wind speed obeys the Weibull distribution, and the probability density function of the wind speed in each season is in, v(s,t) represents the actual wind speed at time t in season s. a(s,t) represents the scale parameter at time t in season s, b(s,t) represents the shape parameter at time t in season s, σ(s,t) represents the standard deviation of wind speed at time t in season s, represents the mean wind speed at time t in season s; The randomness of the light intensity follows the beta distribution, and the probability density function of the light intensity in each season is: in, g(s,t) represents the per-unit value of light intensity at time t in season s. α(s,t) represents the shape parameter that is positive at time t in season s. β(s,t) represents the shape parameter that is positive at time t in season s. μ(s,t) represents the per-unit mean light intensity at time t in season s. represents the standard deviation of light intensity at time t in season s; The randomness of the electric load and heat load obeys the normal distribution. The probability density function of the electric load and heat load in each season is: in, L(m,s,t) represents the load size of the mth load at time t in season s. μ(m,s,t) represents the mean value of the mth load at time t in season s. μ(m,s,t) represents the standard deviation of the mth load at time t in season s.

5. The random scenario construction method considering temporal autocorrelation and cross-correlation according to claim 1, characterized in that: In step 3, for each time series autocorrelation coefficient matrix S s,q The specific form of Cholesky decomposition is S s，q =DD T , in, S s,q Represents the autocorrelation coefficient matrix of each time series, D means that by s,q Perform Cholesky decomposition on the lower triangular matrix.

6. The random scenario construction method considering temporal autocorrelation and cross-correlation according to claim 1, characterized in that: In step 4, for each reference time series autocorrelation coefficient matrix S s,q,ref The specific form of Cholesky decomposition is in, S s,q,ref Represents the autocorrelation coefficient matrix of each reference time series, D ref Indicates that by S s,q,ref Perform Cholesky decomposition on the lower triangular matrix.

7. The random scenario construction method considering temporal autocorrelation and cross-correlation according to claim 1, characterized in that: In step 5, the velocity of the j-th dimension of the i-th initial particle is calculated by the following formula v(i,j)=ω(t)·v(i,j)+c1·rand()·(pbest(i,j)-x(i,j))+c2·rand()·(gbest(i,j)-x(i,j)) in, v(i, j) represents the velocity of the jth dimension of the i-th initial particle, 1≤i≤N,1≤j≤D′, ω(t) represents the inertia weight at the t-th iteration, c1 is the learning factor of the particle swarm algorithm, c2 is the learning factor of the particle swarm algorithm, pbest(i,j) represents the historical optimal position of the initial particle i in dimension j. gbest(i,j) represents the global optimal position of the initial particle i in dimension j. x(i,j) is the current position of the initial particle i in the j dimension.

8. The random scenario construction method considering temporal autocorrelation and cross-correlation according to claim 7, characterized in that: In step 5, the calculation formula of inertia weight ω(t) is in, ω(t) represents the inertia weight at the t-th iteration, ω max represents the upper limit of the inertia weight change, ω min represents the lower limit of the inertia weight change, t max Indicates the maximum number of iterations.

9. The random scenario construction method considering temporal autocorrelation and cross-correlation according to claim 1, characterized in that: In step 6, the fitness function is expressed as in, E r represents the cross-correlation error, s max Indicates the number of seasons in a year, q max Indicates the number of uncertain factors, ρ h,s Represents the mutual correlation coefficient matrix between the uncertain factors of the sth season and the hth random scenario, ρ s,ref Represents the reference cross-correlation coefficient matrix between the uncertainty factors in the sth season.

Images