A rough set-based method for classifying new energy and load output scenarios

Abstract

The present invention relates to a new energy and load output scene aggregation method based on rough sets. The aggregation method includes the following steps: measuring the output of photovoltaic / wind power / load at time point b every day within a day; discretizing the data and simplifying the decision table, numbering the simplified decision table in sequence, and rotating in sequence according to the number; each rotation, analyzing the data, and sequentially finding the core attribute set of the first n-1 columns to the nth column; after n rotations, converting it into an n×n matrix, adding the corresponding values ​​of the kth row and the kth column of the matrix to obtain a 1×n matrix, and judging whether it is the same as the kth scene by the corresponding values ​​of different elements in the matrix. The present invention can obtain similar photovoltaic / wind power / load output scenes through rough sets, and only one of the same scene is taken for research, thereby streamlining data and reducing the amount of calculation, and forming a typical reduced scene set that can reflect the characteristics of historical data changes within the calculation cycle.

Description

Technical field:

[0001] The present invention relates to a new energy system control method, and in particular to a new energy and load output scene aggregation method based on rough sets. Background technology:

[0002] The installed capacity of wind power and solar power generation will continue to increase, and the changes in wind / photovoltaic output and power load have certain characteristics such as randomness and volatility. At the same time, they also have certain seasonal and daily variation patterns, which leads to higher requirements for the flexible adjustment capacity of the power system and greater challenges for power quality. Scenario analysis of grid-connected power systems containing wind power / photovoltaic can provide scientific guidance for evaluating the rationality of grid planning and the feasibility of operation plans. The increase in the penetration rate of wind power and photovoltaic renewable energy generation and the diversity of loads bring uncertain risks to the operation of the power system. The purpose of scenario reduction is to analyze and optimize small samples so that planning plans and operation strategies can meet the needs of power systems containing uncertain wind power / photovoltaic power sources to the greatest extent.

[0003] The medium- and long-term planning and operation and dispatch of power systems involve the analysis and evaluation of a large number of scenarios and solutions. The output power of wind and photovoltaic power stations varies with wind energy and solar radiation, and is affected by variables such as weather, season, and region, and has strong randomness, volatility, and intermittency. The large-scale grid connection of wind power / photovoltaic power affects the real-time balance of power and threatens the security of the power grid, puts forward higher requirements for the planning and operation of the power system, and poses a severe challenge to improving the grid connection acceptance capacity of wind power / photovoltaic power generation in the power system. Considering the randomness, uncertainty and other characteristics of wind / photovoltaic output and load changes in the scenario analysis process will further increase the complexity and time of calculation. In order to take into account both the calculation efficiency and the data change characteristics, the present invention studies a new energy and load output scenario simplification method based on rough sets, draws on the idea of ​​rough sets, simplifies data, reduces the amount of calculation, and forms a typical reduction scenario set that can reflect the characteristics of historical data changes within the calculation cycle. It is of great significance to power grid planning and design, wind / photovoltaic acceptance capacity assessment, power supply optimization configuration, energy storage planning and operation and dispatch, etc., and can be widely used in various fields such as power system planning and operation. Summary of the invention:

[0004] The purpose of the present invention is to provide a method for classifying new energy and load output scenarios based on rough sets, so that only one photovoltaic / wind power / load output scenario of the same type can be studied, thereby achieving simplification. The specific technical solution is as follows:

[0005] A method for classifying new energy and load output scenarios based on rough sets includes the following steps:

[0006] Step S1: Measure the output of photovoltaic / wind power / load at time point b every day within a day. The specific process includes:

[0007] Step S1.1: Obtain the photovoltaic / wind power / load power fluctuation data within a day, collect the photovoltaic / wind power / load output at time point b every day, and obtain the photovoltaic / wind power / load power output at time point b every day within a day, that is, obtain a scenario;

[0008] Step S1.2: Convert the output data of photovoltaic / wind power / load at time point b every day within the measured duration of a day into a decision information system, which includes the domain U, the set of attributes R, and the set of attribute values ​​V. The domain U is a set of objects, i.e., time b every day, R is the photovoltaic / wind power / load scenario of day a, and V is all power data, to obtain a decision table;

[0009] Step S1.3: The domain U contains b individuals, i.e., b time periods. Each individual is an a-tuple, i.e., a photovoltaic / wind power / load scenario for a day. Each dimension of the tuple represents an attribute information of the individual. These individuals can be described by their attribute knowledge to obtain a knowledge expression system.

[0010] Step S2: discretize the data and simplify the decision table, delete the repeated rows and columns in the decision table to obtain a simplified m-row n-column decision table, number the n days and m times in the simplified m-row n-column decision table in sequence, and use one column as the nth column according to the number, and use the other columns as the first n-1 columns in the order of the numbers, and rotate them in sequence; the specific process includes:

[0011] Step S2.1: Divide the data in the decision table of step S1.2, and divide each photovoltaic / wind power / load scenario into several levels in turn. The number of levels can be the same or different, and discretize the data;

[0012] Step S2.2: Simplify the decision table, delete the repeated rows and columns in the decision table to get an m-row n-column decision table, number the n days and m times in the decision table in sequence, and the n days are numbered d1, d2, ..., d n , m The time numbers are t1, t2, ..., t m ;

[0013] Step S2.3: According to the numbers d1, d2, ..., d n The order of the columns is taken as the nth column, and the other columns are taken as the first n-1 columns in the order of their numbers, and they are rotated in turn, for a total of n times;

[0014] Step S3: Each time the rotation is performed, the first n-1 columns are columns with conditional attributes, and the nth column is a column with decision attributes. The data is analyzed to obtain the core attribute set of the first n-1 columns to the nth column in turn. The specific process includes:

[0015] Step S3.1: Let the conditional criterion set be C, the first n-1 columns be the columns with conditional attributes, the decision criterion set be D, the nth column be the column with decision attributes, classify the domain U relative to D, denoted as U / D, and classify U relative to C, denoted as U / C. If an element in the U / C set belongs to an element in U / D, then write the element into a new set, which is the positive domain of C with respect to D, denoted as

[0016] Step S3.2: Find the core attribute set. Remove the condition criterion set corresponding to the first column and record it as C1. Remove the condition criterion set corresponding to the second column and record it as C2. Remove the condition criterion set corresponding to the n-1th column and record it as C n-1 , analyze the data and solve for U / C1, U / C2, ..., U / C n-1 , determine in turn whether the photovoltaic / wind power / load scenarios corresponding to the first column to the n-1th column can be deleted relative to the photovoltaic / wind power / load scenario corresponding to the nth column. If It can be deleted if Then it cannot be deleted and written into the core attribute set;

[0017] Step S4: After each rotation, a 1×n sample matrix is ​​obtained. The matrix elements correspond to the scene numbers in sequence. After n rotations, it is converted into an n×n matrix. The corresponding values ​​of the kth row and the kth column of the matrix are added to obtain a 1×n matrix. The corresponding values ​​of different elements in the matrix are used to determine whether it is the same type as the kth scene.

[0018] As a preferred solution, step S4 includes the following steps:

[0019] Step S4.1: After each rotation, a 1×n matrix is ​​obtained, and the matrix elements correspond to the scene numbers in sequence. The scene numbers represented by the nth column with decision attributes are not assigned values, and the elements corresponding to the scene numbers in the core attribute set are assigned values ​​of 1, otherwise they are 0;

[0020] Step 4.2: After n rotations, convert it into an n×n matrix. Add the corresponding values ​​of the k-th row and the k-th column of the matrix to get a 1×n matrix. If the value of the element in the matrix is ​​2, then the scene number corresponding to the element is of the same type as the k-th scene number. If the value of the element in the matrix is ​​1, then it is uncertain whether the scene number corresponding to the element is of the same type as the k-th scene number. If the value of the element in the matrix is ​​0, then the scene number corresponding to the element is not of the same type as the k-th scene number.

[0021] Compared with the closest prior art, the excellent effects of the present invention are:

[0022] In the technical solution of the present invention, the rough set theory is used to directly analyze and reason the data, discover implicit knowledge and reveal potential laws. The rough set does not need to provide any prior knowledge other than the data set that needs to be processed, and it is highly complementary to the theory of processing other uncertainty problems. The same type of photovoltaic / wind power / load output scenarios can be obtained through the rough set, and only one of the same scenario is taken for research, thereby achieving simplification. Drawing on the idea of ​​rough sets, the data is streamlined and the amount of calculation is reduced to form a typical reduced scenario set that can reflect the characteristics of historical data changes within the calculation cycle. It is of great significance to power grid planning and design, wind / photovoltaic acceptance capacity assessment, power supply optimization configuration, energy storage planning and operation scheduling, etc., and can be widely used in various fields such as power system planning and operation. Description of the drawings:

[0023] Figure 1 It is a flow chart of the method of the present invention.

[0024] Figure 2 This is the decision expression in step S1.2 of the present invention.

[0025] FIG3( a ) is an example of a decision table in step S2.1 of the embodiment; D1 , D2 , D3 , D4 , and D5 represent five different dates, and T1 , T2 , T3 , T4 , and T5 represent five different time periods.

[0026] Fig. 3(b) is a decision table example after the data in Fig. 3(a) is divided into levels in step S2.1 of the embodiment; the meanings of D1, D2, D3, D4, D5, T1, T2, T3, T4, and T5 are the same as those in Fig. 3(a).

[0027] FIG4( a ) is a schematic diagram of decision table data of repeated rows in step S2.2 of the embodiment; the meanings of D1, D2, D3, D4, D5, T1, T2, T3, T4, and T5 are the same as those in FIG3( a ).

[0028] FIG4( b ) is a schematic diagram of decision table data after duplicate rows are deleted in step S2.2 of the embodiment; the meanings of D1, D2, D3, D4, D5, T1, T2, T3, and T4 are the same as those in FIG3( a ).

[0029] 5( a ) is a schematic diagram of the decision table data after numbering in step S2.2 of the embodiment; d1, d2, d3, d4, d5 are numbers corresponding to D1, D2, D3, D4, D5, respectively, and t1, t2, t3, t4 are numbers corresponding to T1, T2, T3, T4, respectively.

[0030] Figure 5(b) is a schematic diagram of the data of a rotation decision table in which the first column is adjusted to the last column in step S2.3 of the embodiment; the meanings of d1, d2, d3, d4, d5, t1, t2, t3, and t4 are the same as those in Figure 5(a).

[0031] Figure 5(c) is a schematic diagram of the data of a rotation decision table in which the second column is adjusted to the last column in step S2.3 of the embodiment; the meanings of d1, d2, d3, d4, d5, t1, t2, t3, and t4 are the same as those in Figure 5(a).

[0032] Figure 5(d) is a schematic diagram of the data of a rotation decision table in which the fifth column is adjusted to the last column in step S2.3 of the embodiment; the meanings of d1, d2, d3, d4, d5, t1, t2, t3, and t4 are the same as those in Figure 5(a). Specific implementation method:

[0033] Example:

[0034] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0035] The present invention provides a method for simplifying new energy and load output scenarios based on rough sets. Figure 1 The implementation process of the method described in this embodiment is shown, including the following steps:

[0036] Step S1: Data conversion. Taking the photovoltaic / wind power / load output scenario as an example, the output of photovoltaic / wind power / load at time b every day within a day is measured, and a knowledge expression system is made for the original data. The specific process is as follows:

[0037] Step S1.1: Obtain the photovoltaic / wind power / load power fluctuation data within a day, sample the photovoltaic / wind power / load fluctuation power value every Δt period every day, that is, collect the photovoltaic / wind power / load output at time point b every day, and obtain the photovoltaic / wind power / load power output at time point b every day within a day, that is, obtain a scenario;

[0038] Step S1.2: Convert the output data of photovoltaic / wind power / load at time b every day within the measured duration of a day into a decision information system, which includes the domain U, the set of attributes R, and the set of attribute values ​​V. The domain U is a set of objects, i.e., the output at time b every day, R is the photovoltaic / wind power / load scenario on day a, and V is all power data. The decision table is obtained. The decision expression intent is as follows: Figure 2 As shown;

[0039] Step S1.3: The domain U contains b individuals, i.e., b times. Each individual is an a-tuple, i.e., a photovoltaic / wind power / load scenario on a day. Each dimension of the tuple represents an attribute information of the individual. These individuals can be described by their attribute knowledge to obtain a knowledge expression system. For example, a time can be represented by its photovoltaic / wind power / load scenario from day 1 to day a.

[0040] Step S2: discretize the data and simplify the decision table. Delete the repeated rows and columns in the decision table to obtain a simplified m-row n-column decision table. Number the n days and m time periods in the decision table in sequence. Use one column as the nth column according to the number, and use the other columns as the first n-1 columns in the order of the numbers. Rotate them in sequence for a total of n times. The specific process is as follows:

[0041] Step S2.1: Divide the data in the decision table of step S1.2, and divide each photovoltaic / wind power / load scenario into several levels in turn. The number of levels can be the same or different, and discretize the data. For example, the decision table is shown in Figure 3(a). D1 days are divided into two levels, 5 to 15 is 0, 15 to 25 is 1, D2 days are divided into three levels, 20 to 30 is 0, 30 to 40 is 1, 40 to 50 is 2, D3 days are divided into two levels, 10 to 15 is 0, 15 to 20 is 1, D4 days are divided into three levels, 5 to 10 is 0, 10 to 15 is 1, 15 to 20 is 2, and D5 days are divided into two levels, 10 to 20 is 0, 20 to 30 is 1, as shown in Figure 3(b);

[0042] Step 2.2: Simplify the decision table, delete the repeated rows and columns in the decision table to get an m-row n-column decision table, and number the n days and m times in the decision table in sequence. The n days are numbered d1, d2, ..., d n , m The time numbers are t1, t2, ..., t m ,For example, as shown in Figure 4(a), if rows T1 and T5 are repeated, then row T5 is deleted, and the simplification of the decision table is completed, as shown in Figure 4(b), and numbering is performed, as shown in Figure 5(a);

[0043] Step 2.3: Press numbers d1, d2, ..., d n The order of d1, d2, d3, d4, and d5 is used as the nth column, and the other columns are used as the first n-1 columns in the order of their numbers, and they are rotated in turn, for a total of n times. For example, as shown in Figures 5(b), (c), and (d), the order of d1, d2, d3, d4, and d5 is used as the 5th column, and the other columns are used as the first 4 columns in the order of their numbers, and they are rotated 5 times in total;

[0044] Step S3: Each time the rotation is performed, the first n-1 columns are columns with conditional attributes, and the nth column is a column with decision attributes. The data is analyzed to obtain the core attribute set of the first n-1 columns to the nth column in turn. The specific process is as follows:

[0045] Step S3.1: Let the conditional criterion set be C, the first n-1 columns be the columns with conditional attributes, the decision criterion set be D, the nth column be the column with decision attributes, classify the domain U relative to D, denoted as U / D, and classify U relative to C, denoted as U / C. If an element in the U / C set belongs to an element in U / D, then write the element into a new set, which is the positive domain of C with respect to D, denoted as Taking d1 as the last column in Figure 5(b) as an example, U / D = {{t1, t2, t4}, {t3}}, U / C = {t1, t2, t3, t4},

[0046] Step S3.2: Find the core attribute set. Remove the condition criterion set corresponding to the first column and record it as C1. Remove the condition criterion set corresponding to the second column and record it as C2. Remove the condition criterion set corresponding to the n-1th column and record it as C n-1 , analyze the data and solve for U / C1, U / C2, ..., U / C n-1 , determine in turn whether the photovoltaic / wind power / load scenarios corresponding to the first column to the n-1th column can be deleted relative to the photovoltaic / wind power / load scenario corresponding to the nth column. If It can be deleted, indicating that the similarity between this scene and the scene corresponding to the nth column is very small, and it is not written into the core attribute set. It cannot be deleted and written into the core attribute set. Taking d1 as the last column in Figure 5(b) as an example, U / C1 = {t1, t2, {t3, t4}}, Therefore, d2 photovoltaic / wind power / load scenarios are similar to d1 and cannot be deleted. They are written into the core attribute set. Similarly, U / C2 = {t1, t2, t3, t4}, Therefore, d3 photovoltaic / wind power / load scenarios are not similar to d1 and can be deleted without writing into the core attribute set. U / C3 = {t1, t2, t3, t4}, Therefore, the photovoltaic / wind power / load conditions of d4 are not similar to those of d1 and can be deleted without writing into the core attribute set. U / C4 = {t1, t2, t3, t4}, Therefore, d5 photovoltaic / wind power / load scenarios are not similar to d1 and can be deleted without being written into the core attribute set. The core attribute set is {d2};

[0047] Step S4: After each rotation, a 1×n matrix is ​​obtained. The matrix elements correspond to the scene numbers in sequence. The scene numbers represented by the nth column with decision attributes are not assigned values. The corresponding elements of the scene numbers in the core attribute set are assigned values ​​of 1, otherwise they are 0. After n rotations, it is converted into an n×n matrix. The corresponding values ​​of the kth row and the kth column of the matrix are added to obtain a 1×n matrix. The corresponding values ​​of different elements in the matrix are used to determine whether they are the same as the kth scene. The specific process is as follows:

[0048] Step S4.1: After each rotation, a 1×n matrix is ​​obtained. The matrix elements correspond to the scene numbers in sequence. The scene numbers represented by the nth column with decision attributes are not assigned values. The corresponding elements of the scene numbers in the core attribute set are assigned values ​​of 1, otherwise they are 0. For example, in Figure 5(b), d1 is the last column, and the core attribute set is {d2}, then the corresponding matrix is ​​[-1 0 0 0];

[0049] Step S4.2: After n rotations, it is converted into an n×n matrix. The corresponding values ​​of the kth row and the kth column of the matrix are added to obtain a 1×n matrix. If the value of the element in the matrix is ​​2, the scene number corresponding to the element is of the same type as the kth scene number. If the value of the element in the matrix is ​​1, it is uncertain whether the scene number corresponding to the element is of the same type as the kth scene number. If the value of the element in the matrix is ​​0, the scene number corresponding to the element is not of the same type as the kth scene number. Taking the rotation of Figure 5(a), Figure 5(b), Figure 5(c), and Figure 5(d) as an example, d1 is used as the last column, and the corresponding matrix is ​​[-1 0 0 0], d2 is used as the last column, and the corresponding matrix is ​​[1- 0 0 0], d3 is used as the last column, and the corresponding matrix is ​​[0 0-0 0], d4 is used as the last column, and the corresponding matrix is ​​[0 0 0-0], and d5 is used as the last column, and the corresponding matrix is ​​[0 0 0 0-]. The obtained n×n matrix is The sum of the first row and the first column is [- 2 0 0 0], the sum of the second row and the second column is [2 - 0 0 0], the sum of the third row and the third column is [0 0 - 0 0], the sum of the fourth row and the fourth column is [0 0 0 - 0], and the sum of the fifth row and the fifth column is [0 0 0 0 -]. Then scene d1 is of the same type as scene d2, and the other scenes are of different types. If scenarios need to be considered, only one of scene d1 and scene d2 can be taken for study to simplify the scenarios. In this example, there are fewer scenes. When there are more scenes, there will be more of the same type, thus achieving simplification.

[0050] Finally, it should be noted that the described embodiments are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

Claims

1. A rough set-based new energy and load output scenario aggregation method, characterized in that: The steps include: Step S1: Measure the output of photovoltaic / wind power / load at time point b every day within a day. The specific process includes: Step S1.1: Obtain the photovoltaic / wind power / load power fluctuation data within a day, collect the photovoltaic / wind power / load output at time point b every day, and obtain the photovoltaic / wind power / load power output at time point b every day within a day, that is, obtain a scenario; Step S1.2: Convert the output data of photovoltaic / wind power / load at time point b every day within the measured duration of a day into a decision information system, which includes the domain U, the set of attributes R, and the set of attribute values ​​V. The domain U is a set of objects, i.e., time b every day, R is the photovoltaic / wind power / load scenario of day a, and V is all power data, to obtain a decision table; Step S1.3: The domain U contains b individuals, i.e., b time periods. Each individual is an a-tuple, i.e., a photovoltaic / wind power / load scenario on a day. Each dimension of the tuple represents an attribute information of the individual. These individuals can be described by their attribute knowledge to obtain a knowledge expression system. Step S2: discretize the data and simplify the decision table, delete the repeated rows and columns in the decision table to obtain a simplified m-row n-column decision table, number the n days and m times in the simplified m-row n-column decision table in sequence, and use one column as the nth column according to the number, and use the other columns as the first n-1 columns in the order of the numbers, and rotate them in sequence; the specific process includes: Step S2.1: Divide the data in the decision table of step S1.2, and divide each photovoltaic / wind power / load scenario into several levels in turn. The number of levels can be the same or different, and discretize the data; Step S2.2: Simplify the decision table, delete the repeated rows and columns in the decision table to get an m-row n-column decision table, number the n days and m times in the decision table in sequence, and the n days are numbered d1, d2, ..., d n , the time numbers are t1, t2, ..., t m ; Step S2.3: According to the numbers d1, d2, ..., d n The order of the columns is taken as the nth column, and the other columns are taken as the first n-1 columns in the order of their numbers, and they are rotated in turn, for a total of n times; Step S3: Each time the rotation is performed, the first n-1 columns are columns with conditional attributes, and the nth column is a column with decision attributes. The data is analyzed to obtain the core attribute set of the first n-1 columns to the nth column in turn. The specific process includes: Step S3.1: Let the conditional criterion set be C, the first n-1 columns be the columns with conditional attributes, the decision criterion set be D, the nth column be the column with decision attributes, classify the domain U relative to D, denoted as U / D, and classify U relative to C, denoted as U / C. If an element in the U / C set belongs to an element in U / D, then write the element into a new set, which is the positive domain of C with respect to D, denoted as Step S3.2: Find the core attribute set. Remove the condition criterion set corresponding to the first column and record it as C1. Remove the condition criterion set corresponding to the second column and record it as C2. Remove the condition criterion set corresponding to the n-1th column and record it as C n-1 , analyze the data and solve for U / C1, U / C2, ..., U / C n-1 , determine in turn whether the photovoltaic / wind power / load scenarios corresponding to the first column to the n-1th column can be deleted relative to the photovoltaic / wind power / load scenario corresponding to the nth column. If It can be deleted if Then it cannot be deleted and written into the core attribute set; Step S4: After each rotation, a 1×n sample matrix is ​​obtained, and the matrix elements correspond to the scene numbers in sequence. After n rotations, it is converted into an n×n matrix. The corresponding values ​​of the kth row and the kth column of the matrix are added to obtain a 1×n matrix. The corresponding values ​​of different elements in the matrix are used to determine whether it is the same as the kth scene. Step S4 includes the following steps: Step S4.1: After each rotation, a 1×n matrix is ​​obtained, and the matrix elements correspond to the scene numbers in sequence. The scene numbers represented by the nth column with decision attributes are not assigned values, and the elements corresponding to the scene numbers in the core attribute set are assigned values ​​of 1, otherwise they are 0; Step S4.2: After n rotations, convert it into an n×n matrix, add the corresponding values ​​of the k-th row and the k-th column of the matrix to obtain a 1×n matrix. If the value of the element in the matrix is ​​2, then the scene number corresponding to the element is of the same type as the k-th scene number. If the value of the element in the matrix is ​​1, then it is uncertain whether the scene number corresponding to the element is of the same type as the k-th scene number. If the value of the element in the matrix is ​​0, then the scene number corresponding to the element is not of the same type as the k-th scene number.

Images