Power system day-ahead dispatching method and system based on big data filling and clustering

Abstract

The application provides a power system day-ahead scheduling method and system based on big data filling and clustering, wherein the method comprises the following steps: obtaining power system historical load data; processing the power system historical load data by using K-means clustering to obtain a load and wind and light scene set; filling the load and wind and light data according to the load and wind and light scene set generated by clustering; constructing a wind and light power system day-ahead scheduling model; and solving the day-ahead scheduling model according to the filled load and wind and light data and a particle swarm algorithm to obtain an action scheme of each adjustable device. The application considers the comparison of the scene of missing data in the horizontal and vertical directions, appropriately fills the missing values, improves the recognition effect of the model, improves the utilization rate of a large amount of load data, can foresee the time change of load characteristics, perfects the rationality of the day-ahead scheduling strategy, and the scheduling result meets the development demand of a new energy power grid.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power system dispatching strategy, and particularly relates to a power system day-ahead dispatching method and system based on big data filling and clustering. BACKGROUND

[0002] With the continuous improvement of power grid automation and the wide application of power grid data intelligent collection system, the power grid data collection system is becoming mature. However, there are problems of lack of frequency and precision in the process of collecting power grid data, which leads to the existence of missing values in the data, so as to interfere with the process of data analysis and affect the final recognition effect of the model. Therefore, how to effectively fill in the missing values of power grid data has gradually become a difficult problem to be solved.

[0003] The existing day-ahead dispatching strategy has some defects under the condition of incomplete historical data. Firstly, the comparison between the horizontal and vertical scenes of missing data is not considered, so that the missing values cannot be filled in properly, which interferes with the recognition effect of the model. Secondly, a large amount of load data is not effectively utilized, the time variation of load characteristics cannot be predicted, the rationality of day-ahead dispatching strategy needs to be improved, and the dispatching result cannot meet the development needs of power grid containing new energy. SUMMARY

[0004] In view of the deficiencies in the prior art, the application provides a power system day-ahead dispatching method and system based on big data filling and clustering.

[0005] In a first aspect, the application provides a power system day-ahead dispatching method based on big data filling and clustering, comprising:

[0006] obtaining historical load data of a power system;

[0007] processing the historical load data of the power system by K-means clustering to obtain a scene set of load and wind and light;

[0008] filling in the load and wind and light data according to the scene set of load and wind and light generated by clustering;

[0009] constructing a day-ahead dispatching model of a power system containing wind and light;

[0010] solving the day-ahead dispatching model according to the filled-in load and wind and light data and a particle swarm algorithm to obtain an action scheme of each adjustable device.

[0011] Further, the processing of the historical load data of the power system by K-means clustering to obtain the scene set of load and wind and light comprises:

[0012] S101, obtaining a clustering number N and a maximum iteration number M;

[0013] S102, selecting N objects as initial clustering centers in the load and wind power output data;

[0014] S103, calculating the Euclidean distance between the load and wind power output data and the N initial clustering centers;

[0015] S104, classifying the load and wind power output data according to the Euclidean distance;

[0016] S105, calculating the average of each class of objects and updating the initial clustering centers;

[0017] S106, constructing a squared error criterion function according to the load and wind power output data and the updated clustering centers;

[0018] S107, judging whether the squared error criterion function converges;

[0019] S108, if it converges, confirming the end of clustering;

[0020] S109, if it does not converge, judging whether the iteration number is greater than M;

[0021] S110, if it is, confirming the end of clustering;

[0022] S111, if it is not, returning to the operation of step S103;

[0023] S112, evaluating the clustering quality by using a silhouette coefficient, wherein the greater the silhouette coefficient, the higher the clustering quality;

[0024] The expression of the silhouette coefficient is:

[0025]

[0026] Wherein, S is the silhouette coefficient; n is the total number of load and wind power output data; s a is the silhouette coefficient of the a-th load and wind power output data; x a is the average distance between the a-th load and wind power output data in the x-th class of objects and the remaining load and wind power output data in the x-th class of objects, as the cohesion degree in the class; y a is the average distance between the x a th class of objects and the load and wind power output data in all the remaining classes except the x-th class of objects.

[0027] Further, the scenario set of load and wind generated according to the clustering fills in the load and wind data, comprising:

[0028] According to the fluctuation cross-correlation analysis algorithm, the fluctuation cross-correlation coefficient of the known attribute power system historical load data and the missing attribute power system historical load data at the same time is calculated;

[0029] determine whether the fluctuation cross-correlation coefficient exceeds a comparison threshold value;

[0030] If yes, keep the known attribute power system historical load data;

[0031] Calculate the combined weight value of the known attribute and the missing attribute respectively;

[0032] Perform scenario analysis on the hour containing the missing attribute, and select H similar scenario hours in the historical load data of the power system; the hour containing the missing attribute is regarded as the missing hour; the selected H similar scenario hours are regarded as the H similar hours;

[0033] Obtain the time period of the missing attribute power system historical load data in the missing date;

[0034] For the same time period of each similar date, calculate the similarity between the known attribute power system historical load data of the missing hour and the known attribute power system historical load data of each similar hour by dynamic time warping distance;

[0035] According to the combined weight value of the known attribute and the missing attribute, the comprehensive similarity of the missing attribute of each similar scenario is calculated using the following formula:

[0036]

[0037] In the formula: C h is the comprehensive similarity of the missing attribute in the hth similar scenario, h = 1, 2, …, H; S (m,h) is the similarity of the mth known attribute in the hth similar hour and the missing hour, m = 1, 2, …, M, M is the number of known attributes of the retained power system historical load data; w m is the combined weight value of the known attribute and the missing attribute;

[0038] Obtain the scenario whose comprehensive similarity of the missing attribute meets the threshold range;

[0039] Extract the missing attribute power system historical load data at the scenario time from the cth scenario as longitudinal filling data;

[0040] Obtain the power system historical load data at the cth scenario time by linear fitting of the curve as horizontal filling data;

[0041] If no, discard the known attribute power system historical load data.

[0042] Further, the construction of the wind and solar power system day-ahead scheduling model comprises:

[0043] The objective function of the wind and solar power system day-ahead scheduling model is:

[0044]

[0045] where P loss is the active power loss of the power system; U i and U j are the voltage magnitudes of node i and node j, respectively; N l is the number of branches of the power system; G k is the conductance of branch k; θ ij is the voltage phase angle difference between node i and node j;

[0046] constructing constraint conditions, the constraint conditions including system power flow constraints, node voltage constraints, and wind and solar generator operation constraints;

[0047] where the condition of the system power flow constraints is:

[0048]

[0049] where P Li is the active power injected at node i; Q Li is the reactive power injected at node i; P DGi is the active power injected at node i by the wind and solar generators; Q DGi is the reactive power injected at node i by the wind and solar generators; δ ij is the phase angle difference between node i and node j; G ij is the real part of the admittance between node i and node j on branch k; B ij is the imaginary part of the admittance between node i and node j on branch k;

[0050] the condition of the node voltage constraints is:

[0051] U imin U i U imax , i = 1, 2, …, node;

[0052] where node is the number of nodes of the power system; U imin is the minimum value of the voltage magnitude of node i; U imax is the maximum value of the voltage magnitude of node i;

[0053] the condition of the wind and solar generator operation constraints is:

[0054] Q DGimin Q DGi Q DGimax ;

[0055] where Q DGimin is the minimum value of the reactive power injected at node i by the wind and solar generators; Q DGimaxThe maximum reactive power injected into node i for wind and solar power generation;

[0056] Construct a wind and solar power output model, which includes a doubly-fed asynchronous wind turbine power output model and a photovoltaic power generation array power output model;

[0057] The output model of the doubly-fed asynchronous wind turbine includes:

[0058] Active power output P of doubly fed asynchronous wind turbine DFIG The relationship between wind speed v and wind speed v is:

[0059]

[0060] Among them, P e This refers to the rated power of the doubly-fed asynchronous wind turbine; v i For the cut-in wind speed; v e Rated wind speed; v o To cut off the wind speed; k1 and k2 are parameters of the wind turbine power generation system;

[0061] The relationship between the active power output and reactive power output of a doubly-fed asynchronous wind turbine is as follows:

[0062]

[0063] Among them, P DFIG Q represents the active power of the doubly-fed asynchronous wind turbine. DFIG U represents the reactive power of the doubly-fed asynchronous wind turbine; s represents the slip rate; U s Stator-side voltage; I s X represents the stator winding current. s For stator leakage reactance; X m For the magnetizing reactance; I r This refers to the rotor-side converter current.

[0064] The expression for the output model of a photovoltaic power generation array is:

[0065]

[0066] Where η is the photoelectric conversion efficiency; A is the effective illumination area; |Q PV | max This represents the maximum reactive power regulation capability of the photovoltaic power generation array; S PV P represents the capacity of the grid-connected inverter for the photovoltaic power generation array. PV E represents the active power of the photovoltaic array; E represents the radiation received per unit area of ​​the photovoltaic array.

[0067] Furthermore, the step of solving the day-ahead scheduling model based on the supplemented load and wind / solar data and the particle swarm optimization algorithm to obtain the action plan for each adjustable device includes:

[0068] S501, initializing particle population according to the filled power system historical load data, including random position and speed, setting gama parameter and learning factor;

[0069] S502, taking the day-ahead scheduling model objective function of the wind-solar power system as the fitness function, and bringing the initial particle into the fitness function for simulation calculation under the constraint condition and the wind-solar output model;

[0070] S503, updating the individual optimal value of each particle and the global optimal value;

[0071] S504, updating the position and speed of each particle;

[0072] S505, judging whether the iteration number is reached;

[0073] S506, if yes, outputting the global optimal value as the action scheme of the adjustable device;

[0074] S507, if no, returning to the operation of S502.

[0075] In a second aspect, the present application provides a power system day-ahead scheduling system based on big data filling and clustering, characterized in that it comprises:

[0076] a data acquisition module for acquiring power system historical load data;

[0077] a data clustering module for processing the power system historical load data by K-means clustering to obtain a load and wind-solar scenario set;

[0078] a data filling module for filling the load and wind-solar data according to the load and wind-solar scenario set generated by clustering;

[0079] a construction module for constructing a wind-solar power system day-ahead scheduling model;

[0080] a calculation module for solving the day-ahead scheduling model according to the filled load and wind-solar data and the particle swarm algorithm to obtain the action scheme of each adjustable device.

[0081] Further, the data clustering module comprises:

[0082] a first acquisition unit for acquiring the clustering number N and the maximum iteration number M;

[0083] a first selection unit for selecting N objects as initial clustering centers in the load and wind-solar output data;

[0084] a first calculation unit for calculating the Euclidean distance between the load and wind-solar output data and the N initial clustering centers;

[0085] a categorizing unit configured to categorize the load and wind power output data according to the Euclidean distance;

[0086] a second calculating unit configured to calculate the average value of each category of objects and update the initial clustering center;

[0087] a first constructing unit configured to construct a square error criterion function according to the load and wind power output data and the updated clustering center;

[0088] a first judging unit configured to judge whether the square error criterion function converges;

[0089] a first confirming unit configured to confirm the clustering to end if the first judging unit determines that the square error criterion function converges;

[0090] a second judging unit configured to judge whether the iteration number is greater than M if the first judging unit determines that the square error criterion function does not converge;

[0091] a second confirming unit configured to confirm the clustering to end if the second judging unit determines that the iteration number is greater than M;

[0092] a third confirming unit configured to confirm to return to execute the operation of the first calculating unit if the second judging unit determines that the iteration number is not greater than M;

[0093] a quality evaluating unit configured to evaluate the clustering quality by using a silhouette coefficient, wherein the greater the silhouette coefficient is, the higher the clustering quality is; and an expression of the silhouette coefficient is:

[0094]

[0095] wherein S is the silhouette coefficient; n is the total number of the load and wind power output data; s a is the silhouette coefficient of the a-th load and wind power output data; x a is the average distance between the a-th load and wind power output data in the x-th category of objects and the remaining load and wind power output data in the x-th category of objects, so as to be the cohesion degree in the category; y a is the average distance between the x a -th category of objects and the load and wind power output data in all the remaining categories except the x-th category of objects.

[0096] Further, the data filling module comprises:

[0097] a third calculating unit configured to calculate the fluctuation cross-correlation coefficient of the known attribute power system historical load data and the missing attribute power system historical load data at the same time according to a fluctuation cross-correlation analysis algorithm;

[0098] a third judging unit configured to judge whether the fluctuation cross-correlation coefficient exceeds a comparison threshold.

[0099] the fourth confirming unit is configured to confirm to reserve the known attribute power system historical load data when the third judging module determines that the fluctuation cross-correlation coefficient exceeds the comparison threshold value;

[0100] the fourth calculating unit is configured to calculate the combined weight values of the known attributes and the missing attributes respectively;

[0101] the second selecting unit is configured to perform scene analysis on the hour with the missing attributes, and select H similar scene hours in the historical load data of the power system; take the hour with the missing attributes as the missing hour; and take the selected H similar scene hours as the H similar hours;

[0102] the second acquiring unit is configured to acquire the time period of the missing attribute power system historical load data in the missing date;

[0103] the fifth calculating unit is configured to calculate the similarity between the known attribute power system historical load data of the missing hour and the known attribute power system historical load data of each similar hour in the same time period of each similar date by using the dynamic time warping distance;

[0104] the sixth calculating unit is configured to calculate the comprehensive similarity of the missing attribute of each similar scene according to the combined weight values of the known attributes and the missing attributes by using the following formula:

[0105]

[0106] In the formula, C h is the comprehensive similarity of the missing attribute in the hth similar scene, h = 1, 2, …, H; S (m,h) is the similarity of the mth known attribute between the hth similar hour and the missing hour, m = 1, 2, …, M, and M is the number of the known attributes of the reserved power system historical load data; w m is the combined weight value of the known attributes and the missing attributes;

[0107] the third acquiring unit is configured to acquire the scene whose comprehensive similarity of the missing attribute meets the threshold range;

[0108] the extracting unit is configured to extract the missing attribute power system historical load data at the scene time from the cth scene as the longitudinal filling data;

[0109] the fourth acquiring unit is configured to acquire the power system historical load data at the cth scene time as the horizontal filling data by using linear fitting of the curve;

[0110] the fifth confirming unit is configured to confirm to discard the known attribute power system historical load data when the third judging module determines that the fluctuation cross-correlation coefficient does not exceed the comparison threshold value.

[0111] Further, the construction module comprises:

[0112] a second construction unit for constructing a wind-solar power system day-ahead scheduling model objective function:

[0113]

[0114] wherein P loss is the active power loss of the power system; U i and U j are the voltage amplitudes of nodes i and j respectively; N l is the number of branches of the power system; G k is the conductance of branch k; θ ij is the voltage phase angle difference between nodes i and j;

[0115] a third construction unit for constructing a constraint condition; the constraint condition comprises a system power flow constraint, a node voltage constraint and a wind-solar unit operation constraint;

[0116] wherein the condition of the system power flow constraint is:

[0117]

[0118] wherein P Li is the active power injected at node i; Q Li is the reactive power injected at node i; P DGi is the active power injected at node i by the wind-solar power generation; Q DGi is the reactive power injected at node i by the wind-solar power generation; δ ij is the phase angle difference between nodes i and j; G ij is the real part of the admittance of branch k between nodes i and j; B ij is the imaginary part of the admittance of branch k between nodes i and j;

[0119] the condition of the node voltage constraint is:

[0120] U imin U i U imax , i = 1, 2, …, node;

[0121] wherein node is the number of nodes of the power system; U imin is the minimum value of the voltage amplitude of node i; U imax is the maximum value of the voltage amplitude of node i;

[0122] the condition of the wind-solar unit operation constraint is:

[0123] Q DGimin Q DGiQ DGimax ;

[0124] wherein, Q DGimin is the minimum value of the reactive power injected by the wind-solar power generation at the node i; Q DGimax is the maximum value of the reactive power injected by the wind-solar power generation at the node i;

[0125] a fourth construction unit, configured to construct a wind-solar output model; the wind-solar output model comprises a doubly-fed asynchronous wind turbine output model and a photovoltaic array output model;

[0126] wherein, the doubly-fed asynchronous wind turbine output model comprises:

[0127] a relationship between the active power P DFIG of the doubly-fed asynchronous wind turbine and the wind speed v:

[0128]

[0129] wherein, P e is the rated power of the doubly-fed asynchronous wind turbine; v i is the cut-in wind speed; v e is the rated wind speed; v o is the cut-out wind speed; k1 and k2 are both wind turbine generation system parameters;

[0130] a relationship between the active power and the reactive power of the doubly-fed asynchronous wind turbine:

[0131]

[0132] wherein, P DFIG is the active power of the doubly-fed asynchronous wind turbine; Q DFIG is the reactive power of the doubly-fed asynchronous wind turbine; s is the slip; U s is the stator-side voltage; I s is the stator winding current; X s is the stator leakage reactance; X m is the excitation reactance; I r is the rotor-side converter current;

[0133] the expression of the photovoltaic array output model is:

[0134]

[0135] wherein, η is the photoelectric conversion efficiency; A is the effective light area; |Q PV | max is the maximum reactive power regulation capability of the photovoltaic array; S PV is the grid-connected inverter capacity of the photovoltaic array; P PV is the active power of the photovoltaic array; E is the radiation received per unit area of the photovoltaic array.

[0136] Further, the calculation module comprises:

[0137] An initialization unit is configured to initialize a particle population according to the filled power system historical load data, including random position and speed, set gama parameter and learning factor;

[0138] A seventh calculation unit is configured to take the wind and light power system day-ahead scheduling model objective function as the fitness function, and bring the initial particle into the fitness function for simulation calculation under the constraint condition and the wind and light output model;

[0139] A first updating unit is configured to update the individual optimal value and the global optimal value of each particle;

[0140] A second updating unit is configured to update the position and speed of each particle;

[0141] A fourth judging unit is configured to judge whether the iteration number is reached;

[0142] A sixth confirming unit is configured to confirm the global optimal value as the action scheme of the adjustable device in the case that the fourth judging unit determines that the iteration number is reached;

[0143] A seventh confirming unit is configured to confirm the operation performed by the seventh calculation unit in the case that the fourth judging unit determines that the iteration number is not reached.

[0144] The present application provides a kind of based on big data filling and clustering's power system day-ahead scheduling method and system, wherein method includes obtaining power system historical load data;With K-means clustering, power system historical load data is handled, and the scene set of load and wind and light is obtained;According to the scene set of load and wind and light generated by clustering, load and wind and light data are filled;Wind and light power system day-ahead scheduling model is constructed;According to the filled load and wind and light data and particle swarm optimization, day-ahead scheduling model is solved, and the action scheme of each adjustable device is obtained.The present application considers the comparison of the scene where missing data is located horizontally and longitudinally, appropriately fills in missing value, improves the identification effect of model;Improve the utilization rate of a large amount of load data, can foresee the time variation of load characteristic, perfect day-ahead scheduling strategy rationality, and scheduling result meets the development demand of new energy power grid. BRIEF DESCRIPTION OF DRAWINGS

[0145] In order to more clearly illustrate the technical scheme of the present application, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0146] Figure 1A flow chart of a power system day-ahead scheduling method based on big data filling and clustering is provided for the embodiment of the present application.

[0147] Figure 2 A missing value filling method based on historical data assisted scene analysis is provided for the embodiment of the present application.

[0148] Figure 3 An improved IEEE14 node schematic diagram is provided for the embodiment of the present application.

[0149] Figure 4 A regulation strategy diagram of each adjustable device per hour is provided for the embodiment of the present application

[0150] Figure 5 A structure diagram of a power system day-ahead scheduling system based on big data filling and clustering is provided for the embodiment of the present application. DETAILED DESCRIPTION

[0151] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0152] In an embodiment, as shown in Figure 1 The embodiment of the present application provides a power system day-ahead scheduling method based on big data filling and clustering, characterized in that it comprises the following steps:

[0153] S1, obtaining historical load data of a power system.

[0154] S2, processing the historical load data of the power system by using K-means clustering to obtain a scene set of load and wind and light output.

[0155] Exemplarily, this step comprises:

[0156] S101, obtaining a clustering number N and a maximum iteration number M.

[0157] S102, selecting N objects in the load and wind and light output data as initial clustering centers.

[0158] S103, calculating the Euclidean distances between the load and wind and light output data and the N initial clustering centers.

[0159] S104, classifying the load and wind and light output data according to the Euclidean distances.

[0160] S105, calculating the average values of each class of objects and updating the initial clustering centers.

[0161] S106, construct the squared error criterion function based on load and wind and solar power output data and the updated cluster centers.

[0162] S107, determine whether the squared error criterion function converges.

[0163] S108. If convergence is achieved, the clustering process is considered complete.

[0164] S109. If convergence fails, determine if the number of iterations is greater than M.

[0165] S110, if yes, then confirm the end of clustering.

[0166] S111, if not, return to step S103.

[0167] S112 uses the silhouette coefficient to evaluate cluster quality, where a larger silhouette coefficient indicates higher cluster quality.

[0168] The expression for the contour coefficient is:

[0169]

[0170] Where S is the profile coefficient; n is the total number of load and wind / solar output data; s a x is the profile coefficient for the a-th load and wind / solar output data; a Let y be the average distance between the a-th load and wind / solar output data in the x-th object and the remaining load and wind / solar output data in the x-th object, and use it as the cohesion within the class; a For x a The minimum value of the average distance to the load and solar power output data of all remaining classes except for class x.

[0171] S3, fills in the load and scenery data based on the scene sets of load and scenery generated by clustering.

[0172] For example, such as Figure 2 As shown, this step includes:

[0173] The fluctuation cross-correlation analysis algorithm is used to calculate the fluctuation cross-correlation coefficient between historical load data of power systems with known attributes and historical load data of power systems with missing attributes at the same time.

[0174] Determine whether the cross-correlation coefficient of fluctuations exceeds the comparison threshold.

[0175] If so, retain the historical load data of the power system with known attributes.

[0176] Calculate the combined weight values ​​for known attributes and missing attributes respectively.

[0177] A scene analysis is performed on the hour containing the missing attribute, and H similar hours are selected from the historical load data of the power system; the hour containing the missing attribute is taken as the missing hour; and the H similar hours are taken as the H similar hours.

[0178] The time period in the missing date of the missing attribute power system historical load data is obtained.

[0179] For the same time period of each similar date, the similarity between the known attribute power system historical load data of the missing hour and the known attribute power system historical load data of each similar hour is calculated by using the dynamic time warping distance.

[0180] According to the combined weight value of the known attribute and the missing attribute, the comprehensive similarity of the missing attribute of each similar scene is calculated by using the following formula:

[0181]

[0182] In the formula, C h is the comprehensive similarity of the missing attribute in the hth similar scene, h = 1, 2, …, H; S (m,h) is the similarity of the mth known attribute in the hth similar hour and the missing hour, m = 1, 2, …, M, M is the number of known attributes of the power system historical load data; w m is the combined weight value of the known attribute and the missing attribute.

[0183] The scene whose comprehensive similarity of the missing attribute meets the threshold range is obtained.

[0184] The missing attribute power system historical load data at the scene time of the cth scene is extracted as the longitudinal filling data.

[0185] The linear fitting of the curve is used to obtain the power system historical load data at the time of the cth scene as the horizontal filling data.

[0186] If no, the known attribute power system historical load data is discarded.

[0187] S4, a day-ahead scheduling model of the wind-solar power system is constructed.

[0188] Exemplarily, the objective function of the day-ahead scheduling model of the wind-solar power system is:

[0189]

[0190] In the formula, P loss is the active power loss of the power system; U i and U j are the voltage amplitudes of node i and node j, respectively; N lis the number of branches of the power system; G k is the conductance of branch k; θ ij is the phase angle difference between node i and node j.

[0191] The constraint conditions include system flow constraints, node voltage constraints, and wind and solar unit operation constraints.

[0192] The conditions of the system flow constraints are:

[0193]

[0194] wherein, P Li is the active power injected at node i; Q Li is the reactive power injected at node i; P DGi is the active power injected at node i by wind and solar power generation; Q DGi is the reactive power injected at node i by wind and solar power generation; δ ij is the phase angle difference between node i and node j; G ij is the real part of admittance between node i and node j on branch k; B ij is the imaginary part of admittance between node i and node j on branch k.

[0195] The conditions of the node voltage constraints are:

[0196] U imin U i U imax , i = 1, 2, …, node.

[0197] wherein, node is the number of nodes of the power system; U imin is the minimum value of the voltage amplitude of node i; U imax is the maximum value of the voltage amplitude of node i.

[0198] The conditions of the wind and solar unit operation constraints are:

[0199] Q DGimin Q DGi Q DGimax .

[0200] wherein, Q DGimin is the minimum value of the reactive power injected at node i by wind and solar power generation; Q DGimax is the maximum value of the reactive power injected at node i by wind and solar power generation.

[0201] The wind and solar output model is constructed, and the wind and solar output model includes a doubly-fed asynchronous wind turbine output model and a photovoltaic array output model.

[0202] The doubly-fed asynchronous wind turbine output model includes:

[0203] Active power of doubly-fed induction wind turbine P DFIG The relationship between P and wind speed v is:

[0204]

[0205] Where, P e is the rated power of doubly-fed induction wind turbine; v i is the cut-in wind speed; v e is the rated wind speed; v o is the cut-out wind speed; k1 and k2 are parameters of wind turbine generation system.

[0206] The relationship between active power and reactive power of doubly-fed induction wind turbine is:

[0207]

[0208] Where, P DFIG is the active power of doubly-fed induction wind turbine; Q DFIG is the reactive power of doubly-fed induction wind turbine; s is the slip; U s is the stator side voltage; I s is the stator winding current; X s is the stator leakage reactance; X m is the excitation reactance; I r is the rotor side converter current.

[0209] The expression of output model of photovoltaic generation array is:

[0210]

[0211] Where, η is the photoelectric conversion efficiency; A is the effective light area; |Q PV | max is the maximum reactive power regulation capability of photovoltaic generation array; S PV is the capacity of grid-connected inverter of photovoltaic generation array; P PV is the active power of photovoltaic generation array; E is.

[0212] S5, according to the filled load and wind and light data and particle swarm algorithm, the day-ahead scheduling model is solved, and the action scheme of each adjustable device is obtained.

[0213] Exemplarily, the step includes:

[0214] S501, initializing particle population according to the filled historical load data of power system, including random position and speed, setting gama parameter and learning factor.

[0215] S502, taking the objective function of day-ahead scheduling model of wind and light containing power system as fitness function, under the constraint condition and wind and light output model, the initial particle is brought into fitness function for simulation calculation.

[0216] S503, update the individual optimal value and the global optimal value for each particle.

[0217] S504 updates the position and velocity of each particle.

[0218] S505 determines whether the number of iterations has been reached.

[0219] S506, if so, outputs the globally optimal value as the action plan for the adjustable device.

[0220] S507, if not, return to the operation performed in S502.

[0221] The embodiments of this invention employ improved IEEE 14-bus power system simulation analysis, such as... Figure 3 As shown, a large wind farm is connected at node 9, and a centralized photovoltaic power station is connected at node 5. Ten switchable capacitor banks, each with a reactive power capacity of 200 kVar, are connected at nodes 2 and 10 respectively. It is assumed that the transformer between branch 56 and branch 49 is a 17-tap transformer with adjustable taps under load. The power system voltage benchmark is selected as 220 kV, and the capacity benchmark is 100 MVA. The wind turbine cut-in wind speed, rated wind speed, and cut-out wind speed are 4 m / s, 12 m / s, and 25 m / s respectively. The rated capacity is 8 MW, and the effective solar irradiance area of ​​the photovoltaic power station is 30,000 m². 2 The conversion efficiency is 0.9, and the rated capacity is 8MW. The allowable range of node voltage is ±6% of the rated voltage. The particle swarm optimization algorithm has a population size of 50, 150 iterations, a gama parameter of 0.95, and learning factors c1=c2=0.5.

[0222] Using historical data from a typical power grid in a certain city as a filler database, and assuming that there is a data gap on a certain typical day, this data is used as the filler object. After filling the data on this day using the strategy of this application, a portion containing the filler data is extracted and placed into the improved IEEE 14-node system, and its day-ahead scheduling is optimized.

[0223] The day-ahead scheduling results obtained based on the strategy in this application are as follows: Figure 4 As shown in Table 1, the network loss of the calculation system is as follows.

[0224] Table 1 System Network Loss

[0225] Number Comparison type Network loss (kW) Error comparison 1 Run a dispatch cycle with actual data 341.62 3.75％ 2 Run a dispatch cycle after day-ahead scheduling 329.28 0％ 3 Run a dispatch cycle after filling in missing data with the strategy used in this patent 327.53 0.532％

[0226] In summary, the application proposes that according to power grid historical data, a method of big data K-means clustering can generate a representative load and wind and light output scene set; a missing value filling method based on historical data auxiliary scene analysis is proposed, which improves the filling accuracy of missing values of power system day-ahead scheduling historical data; a day-ahead scheduling model of a wind and light power system is established; a particle swarm algorithm is used to solve the device action with the lowest network loss as the target, to meet the development needs of the new energy power grid.

[0227] The application considers the comparison of the missing data in the horizontal and vertical directions of the scene, appropriately fills the missing values, improves the recognition effect of the model, improves the utilization rate of a large amount of load data, can predict the time variation of the load characteristics, and perfects the rationality of the day-ahead scheduling strategy.

[0228] Based on the same inventive concept, the embodiments of the application also provide a power system day-ahead scheduling system based on big data filling and clustering, since the principle of solving problems of the system is similar to the power system day-ahead scheduling method based on big data filling and clustering, therefore the implementation of the system can be referred to the implementation of the power system day-ahead scheduling method based on big data filling and clustering, and the repeated parts will not be described again.

[0229] In another embodiment, the power system day-ahead scheduling system based on big data filling and clustering provided by the embodiments of the application, as shown in Figure 5 , comprises:

[0230] The data acquisition module 10 is used for acquiring power system historical load data.

[0231] The data clustering module 20 is used for processing the power system historical load data by using K-means clustering to obtain a load and wind and light scene set.

[0232] The data filling module 30 is used for filling the load and wind and light data according to the load and wind and light scene set generated by clustering.

[0233] The construction module 40 is used for constructing a day-ahead scheduling model of a wind and light power system.

[0234] The calculation module 50 is used for solving the day-ahead scheduling model according to the filled load and wind and light data and a particle swarm algorithm to obtain an action scheme of each adjustable device.

[0235] Exemplarily, the data clustering module comprises:

[0236] The first acquisition unit is used for acquiring the number of clusters N and the maximum number of iterations M.

[0237] The first selection unit is used for selecting N objects as initial cluster centers in the load and wind and light output data.

[0238] The first computing unit is configured to calculate the Euclidean distance between the load and wind power output data and N initial clustering centers.

[0239] The clustering unit is configured to cluster the load and wind power output data according to the Euclidean distance.

[0240] The second computing unit is configured to calculate the average value of each type of object and update the initial clustering center.

[0241] The first constructing unit is configured to construct a square error criterion function according to the load and wind power output data and the updated clustering center.

[0242] The first judging unit is configured to judge whether the square error criterion function converges.

[0243] The first confirming unit is configured to confirm the clustering to end when the first judging unit determines that the square error criterion function converges.

[0244] The second judging unit is configured to judge whether the iteration number is greater than M when the first judging unit determines that the square error criterion function does not converge.

[0245] The second confirming unit is configured to confirm the clustering to end when the second judging unit determines that the iteration number is greater than M.

[0246] The third confirming unit is configured to confirm to return to execute the operation of the first computing unit when the second judging unit determines that the iteration number of the square error criterion function is not greater than M.

[0247] The quality evaluating unit is configured to evaluate the clustering quality by using a silhouette coefficient.

[0248] The expression of the silhouette coefficient is:

[0249]

[0250] wherein S is the silhouette coefficient; n is the total number of the load and wind power output data; s a is the silhouette coefficient of the a th load and wind power output data; x is the distance between the a th load and wind power output data in the x th type of object and the remaining load and wind power output data in the x th type of object, which is taken as the cohesion degree in the class; y is the minimum value of the average distance between the x th type of object and the load and wind power output data in all the remaining classes except the x th type of object. a a a a

[0251] Exemplarily, the data filling module comprises:

[0252] ​​​​The third calculation unit is configured to calculate the fluctuation cross-correlation coefficient of the known attribute power system historical load data and the missing attribute power system historical load data at the same time according to a fluctuation cross-correlation analysis algorithm.

[0253] The third judgment unit is configured to judge whether the fluctuation cross-correlation coefficient exceeds a comparison threshold.

[0254] The fourth confirmation unit is configured to confirm to retain the known attribute power system historical load data when the third judgment module determines that the fluctuation cross-correlation coefficient exceeds the comparison threshold.

[0255] The fourth calculation unit is configured to calculate the combined weight value of the known attribute and the missing attribute respectively.

[0256] The second selection unit is configured to perform scene analysis on the hour with the missing attribute, and select H similar scene hours in the historical load data of the power system; take the hour with the missing attribute as a missing hour; and take the selected H similar scene hours as H similar hours.

[0257] The second acquisition unit is configured to acquire the time period of the missing attribute power system historical load data in the missing date.

[0258] The fifth calculation unit is configured to calculate the similarity of the known attribute power system historical load data of the missing hour and the known attribute power system historical load data of each similar hour in each similar time period of each similar date by using a dynamic time warping distance.

[0259] The sixth calculation unit is configured to calculate the comprehensive similarity of the missing attribute of each similar scene according to the combined weight value of the known attribute and the missing attribute by using the following formula:

[0260]

[0261] In the formula, C h is the comprehensive similarity of the missing attribute in the hth similar scene, h=1, 2, …, H; S (m,h) is the similarity of the mth known attribute in the hth similar hour and the missing hour, m=1, 2, …, M, M is the number of known attributes of the retained power system historical load data; w m is the combined weight value of the known attribute and the missing attribute.

[0262] The third acquisition unit is configured to acquire the scene whose comprehensive similarity of the missing attribute meets a threshold range.

[0263] The extraction unit is configured to extract the missing attribute power system historical load data at the scene from the cth scene as longitudinal filling data.

[0264] The fourth acquisition unit is configured to acquire the power system historical load data in the cth scenario as the horizontal filling data by using linear fitting of the curve.

[0265] The fifth confirmation unit is configured to confirm to discard the known attribute power system historical load data in a case where the third judging module determines that the fluctuation cross-correlation coefficient does not exceed the comparison threshold.

[0266] Exemplarily, the construction module comprises:

[0267] The second construction unit is configured to construct a wind-solar power system day-ahead scheduling model objective function:

[0268]

[0269] wherein, P loss is the active power loss of the power system; U i and U j are the voltage amplitudes of the node i and the node j respectively; N l is the number of branches of the power system; G k is the conductance of the branch k; and θ ij is the voltage phase angle difference between the node i and the node j.

[0270] The third construction unit is configured to construct a constraint condition; the constraint condition comprises a system power flow constraint, a node voltage constraint and a wind-solar unit operation constraint.

[0271] The condition of the system power flow constraint is:

[0272]

[0273] wherein, P Li is the active power injected by the node i; Q Li is the reactive power injected by the node i; P DGi is the active power injected by the wind-solar power generation into the node i; Q DGi is the reactive power injected by the wind-solar power generation into the node i; δ ij is the phase angle difference between the node i and the node j; G ij is the real part of the admittance between the node i and the node j on the branch k; and B ij is the imaginary part of the admittance between the node i and the node j on the branch k.

[0274] The condition of the node voltage constraint is:

[0275] U imin U i U imax , i = 1, 2, …, node.

[0276] wherein, node is the number of nodes of the power system; U iminU is the minimum voltage amplitude at node i; imax This represents the maximum voltage amplitude at node i.

[0277] The operating constraints for wind and solar turbine units are as follows:

[0278] Q DGimin Q DGi Q DGimax .

[0279] Among them, Q DGimin The minimum reactive power injected into node i for wind and solar power generation; Q DGimax The maximum reactive power injected into node i for wind and solar power generation.

[0280] The fourth building unit is used to build a wind and solar power output model; the wind and solar power output model includes a doubly-fed asynchronous wind turbine power output model and a photovoltaic power generation array power output model;

[0281] The output model of the doubly-fed asynchronous wind turbine includes:

[0282] Active power output P of doubly fed asynchronous wind turbine DFIG The relationship between wind speed v and wind speed v is:

[0283]

[0284] Among them, P e This refers to the rated power of the doubly-fed asynchronous wind turbine; v i For the cut-in wind speed; v e Rated wind speed; v o The cut-off wind speed; k1 and k2 are parameters of the wind turbine power generation system.

[0285] The relationship between the active power output and reactive power output of a doubly-fed asynchronous wind turbine is as follows:

[0286]

[0287] Among them, P DFIG Q represents the active power of the doubly-fed asynchronous wind turbine. DFIG U represents the reactive power of the doubly-fed asynchronous wind turbine; s represents the slip rate; U s Stator-side voltage; I s X represents the stator winding current. s For stator leakage reactance; X m For the magnetizing reactance; I r This refers to the rotor-side converter current.

[0288] The expression for the output model of a photovoltaic power generation array is:

[0289]

[0290] Wherein, η is photoelectric conversion efficiency; A is effective light area; |Q PV | max is the maximum reactive power regulation capability of the photovoltaic power generation array; S PV is the grid-connected inverter capacity of the photovoltaic power generation array; P PV is the active power of the photovoltaic power generation array; E is.

[0291] Exemplarily, the calculation module comprises:

[0292] The initialization unit is configured to initialize the particle population according to the filled power system historical load data, including random position and speed, setting the gama parameter and the learning factor.

[0293] The seventh calculation unit is configured to take the day-ahead scheduling model objective function of the wind-solar power system as the fitness function, and bring the initial particle into the fitness function for simulation calculation under the constraint condition and the wind-solar output model.

[0294] The first updating unit is configured to update the individual optimal value and the global optimal value of each particle.

[0295] The second updating unit is configured to update the position and speed of each particle.

[0296] The fourth judging unit is configured to judge whether the iteration number is reached.

[0297] The sixth confirming unit is configured to confirm the global optimal value as the action scheme of the adjustable device when the fourth judging unit determines that the iteration number is reached.

[0298] The seventh confirming unit is configured to confirm the operation performed by the seventh calculation unit when the fourth judging unit determines that the iteration number is not reached.

[0299] The more specific working process of each module described above can refer to the corresponding content disclosed in the foregoing embodiments, and will not be described here.

[0300] In another embodiment, the present application provides a computer device comprising a processor and a memory; wherein the processor implements the steps of the power system day-ahead scheduling method based on big data filling and clustering described above when executing the computer program saved in the memory.

[0301] The more specific process of the method described above can refer to the corresponding content disclosed in the foregoing embodiments, and will not be described here.

[0302] In another embodiment, the present application provides a computer readable storage medium for storing a computer program; the computer program is executed by the processor to implement the steps of the power system day-ahead scheduling method based on big data filling and clustering described above.

[0303] More specific processes of the above method can refer to the corresponding content disclosed in the foregoing embodiments, which will not be repeated here.

[0304] The various embodiments are described in a progressive manner in the specification, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be mutually referred to. For the system, device and storage medium disclosed by the embodiments, since it corresponds to the method disclosed by the embodiments, the description is relatively simple, and the relevant part can be referred to the method part.

[0305] Those skilled in the art can clearly understand that the technical solutions in the embodiments of the present application can be implemented by means of software and necessary general hardware platforms. Based on such understanding, the technical solutions in the embodiments of the present application can be embodied in the form of a software product, which can be stored in a storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in the various embodiments or some parts of the embodiments.

[0306] The above detailed description of the present application is combined with the specific implementation and exemplary examples, but these descriptions cannot be understood as limitations of the present application. Those skilled in the art understand that the technical solutions and their implementation can be variously replaced, modified or improved without departing from the spirit and scope of the present application, which all fall within the scope of the present application. The scope of protection of the present application is subject to the appended claims.

Claims

1. A day-ahead dispatching method for power systems based on big data completion and clustering, characterized in that, include: Obtain historical load data of the power system; K-means clustering was used to process historical load data of the power system to obtain scene sets of load and wind and solar power. The load and scenery data are filled in based on the scene sets of load and scenery generated by clustering; Construct a day-ahead dispatch model for wind and solar power systems; The day-ahead scheduling model is solved using the supplemented load and wind / solar data and the particle swarm optimization algorithm to obtain the action plan for each adjustable device. The process of filling in the load and wind / sunlight data based on the scene sets generated by clustering includes: The fluctuation cross-correlation analysis algorithm is used to calculate the fluctuation cross-correlation coefficient between historical load data of power systems with known attributes and historical load data of power systems with missing attributes at the same time. Determine whether the cross-correlation coefficient of fluctuations exceeds the comparison threshold; If so, retain the historical load data of the power system with known attributes; Calculate the combined weight values ​​for known attributes and missing attributes separately; Scenario analysis is performed on hours with missing attributes, and H hours with similar scenarios are selected from the historical load data of the power system; hours with missing attributes are taken as missing hours; and the hours of the selected H similar scenarios are taken as H similar hours. Retrieve the time period for missing attributes in historical load data of the power system within the missing date; For the same time period on each similar date, the similarity between the historical load data of the power system with known attributes for each missing hour and the historical load data of the power system with known attributes for each similar hour is calculated by using the dynamic time curvature distance. Based on the combined weight values ​​of known and missing attributes, the comprehensive similarity of missing attributes for each similar scene is calculated using the following formula: In the formula: C h S represents the comprehensive similarity of the missing attribute in the h-th similar scene, where h = 1, 2, ..., H; (m,h) Let w be the similarity between the m-th known attribute and the h-th similar hour and the missing hour, where m = 1, 2, ..., M, and M is the number of known attributes in the retained historical load data of the power system; m The combined weight value of the known and missing attributes; Scenarios where the overall similarity of missing attributes meets the threshold range; Extract the missing power system load data for the cth scenario as longitudinal fill data; The historical load data of the power system in the c-th scenario is obtained by linear fitting of the curve and used as the horizontal filling data; If not, discard historical load data of the power system with known attributes.

2. The day-ahead dispatching method for power systems according to claim 1, characterized in that, The process of using K-means clustering to process historical load data of the power system yields a set of load and wind / solar scene data, including: S101, obtain the number of clusters N and the maximum number of iterations M; S102, select N objects as initial cluster centers from the load and wind and solar power output data; S103, calculate the Euclidean distance between load and wind / solar power output data and N initial cluster centers; S104, classifies load and wind and solar power output data according to Euclidean distance; S105, calculate the average value of each type of object and update the initial cluster centers; S106, construct the squared error criterion function based on load and wind and solar power output data and the updated cluster centers; S107, Determine whether the squared error criterion function converges; S108. If convergence is achieved, the clustering process is considered complete. S109, if convergence fails, determine if the number of iterations is greater than M; S110, if yes, then confirm the end of clustering; S111, if not, return to step S103; S112, using the silhouette coefficient to evaluate clustering quality; The expression for the contour coefficient is: Where S is the profile coefficient; n is the total number of load and wind / solar output data; s a x is the profile coefficient for the a-th load and wind / solar output data; a Let y be the average distance between the a-th load and wind / solar output data in the x-th object and the remaining load and wind / solar output data in the x-th object, and use it as the cohesion within the class; a For x a The minimum value of the average distance to the load and solar power output data of all remaining classes except for class x.

3. The day-ahead dispatching method for a power system according to claim 1, characterized in that, The construction of the day-ahead dispatch model for wind and solar power systems includes: The objective function for constructing a day-ahead scheduling model for a wind and solar power system is: Among them, P loss For active power losses in the power system; U i and U j These represent the voltage magnitudes at nodes i and j, respectively; N l G represents the number of branches in the power system. k Let θ be the conductance of branch k; ij The voltage phase angle difference between node i and node j; Construct constraints, including system power flow constraints, node voltage constraints, and wind and solar turbine operation constraints; The conditions for system power flow constraints are as follows: Among them, P Li Active power injected into node i; Q Li The reactive power injected into node i; P DGi Inject active power into node i for wind and solar power generation; Q DGi Inject reactive power into node i for wind and solar power generation; δ ij G represents the phase angle difference between node i and node j; ij B is the real part of the admittance on branch k between node i and node j; ij Let be the imaginary part of the admittance on branch k between node i and node j; The conditions for node voltage constraints are: U imin ≤U i ≤U imax ,i=1,2,…,node; Where node represents the number of nodes in the power system; U imin U is the minimum voltage amplitude at node i; imax The maximum value of the voltage amplitude at node i; The operating constraints for wind and solar turbine units are as follows: Q DGimin ≤Q DGi ≤Q DGimax ; Among them, Q DGimin The minimum reactive power injected into node i for wind and solar power generation; Q DGimax The maximum reactive power injected into node i for wind and solar power generation; Construct a wind and solar power output model, which includes a doubly-fed asynchronous wind turbine power output model and a photovoltaic power generation array power output model; The output model of the doubly-fed asynchronous wind turbine includes: Active power output P of doubly fed asynchronous wind turbine DFIG The relationship between wind speed v and wind speed v is: Among them, P e This refers to the rated power of the doubly-fed asynchronous wind turbine; v i For the cut-in wind speed; v e Rated wind speed; v o To cut off the wind speed; k1 and k2 are parameters of the wind turbine power generation system; The relationship between the active power output and reactive power output of a doubly-fed asynchronous wind turbine is as follows: Among them, P DFIG Q represents the active power of the doubly-fed asynchronous wind turbine. DFIG U represents the reactive power of the doubly-fed asynchronous wind turbine; s represents the slip rate; U s Stator-side voltage; I s X represents the stator winding current. s For stator leakage reactance; X m For the magnetizing reactance; I r This refers to the rotor-side converter current. The expression for the output model of a photovoltaic power generation array is: Where η is the photoelectric conversion efficiency; A is the effective illumination area; |Q PV | max This represents the maximum reactive power regulation capability of the photovoltaic power generation array; S PV P represents the capacity of the grid-connected inverter for the photovoltaic power generation array. PV E represents the active power of the photovoltaic array; E represents the radiation received per unit area of ​​the photovoltaic array.

4. The day-ahead dispatching method for a power system according to claim 3, characterized in that, The step involves solving the day-ahead scheduling model using the supplemented load and wind / solar data and the particle swarm optimization algorithm to obtain the action plan for each adjustable device, including: S501, initialize the particle population based on the supplemented historical load data of the power system, including random positions and velocities, and set the gama parameters and learning factors; S502 uses the objective function of the day-ahead dispatch model of the wind and solar power system as the fitness function. Under the constraints and the wind and solar power output model, the initial particles are substituted into the fitness function for simulation calculation. S503, update the individual optimal value and the global optimal value for each particle; S504 updates the position and velocity of each particle; S505, determine whether the iteration count has been reached; S506, if so, output the globally optimal value as the action plan for the adjustable device; S507, if not, return to the operation performed in S502.

5. A day-ahead dispatching system for power systems based on big data completion and clustering, characterized in that, include: The data acquisition module is used to acquire historical load data of the power system; The data clustering module is used to process historical load data of the power system using K-means clustering to obtain load and wind / solar scene sets; The data imputation module is used to impute load and landscape data based on the scene sets of load and landscape generated by clustering. The building module is used to construct a day-ahead scheduling model for wind and solar power systems; The calculation module is used to solve the day-ahead scheduling model based on the supplemented load and wind and solar data and the particle swarm algorithm to obtain the action plan of each adjustable device. The data imputation module includes: The third calculation unit is used to calculate the fluctuation cross-correlation coefficient between the historical load data of power systems with known attributes and the historical load data of power systems with missing attributes at the same time, based on the fluctuation cross-correlation analysis algorithm. The third judgment unit is used to determine whether the cross-correlation coefficient of fluctuations exceeds the comparison threshold. The fourth confirmation unit is used to confirm the retention of historical load data of the power system with known attributes when the third judgment module determines that the fluctuation cross-correlation coefficient exceeds the comparison threshold. The fourth calculation unit is used to calculate the combined weight values ​​of known attributes and missing attributes, respectively. The second selection unit is used to perform scenario analysis on hours with missing attributes and select H hours with similar scenarios from the historical load data of the power system; the hours with missing attributes are taken as missing hours; and the hours with the selected H similar scenarios are taken as H similar hours. The second acquisition unit is used to acquire the time period of the missing attribute power system historical load data in the missing date; The fifth calculation unit is used to calculate the similarity between the historical load data of the power system with known attributes for each missing hour and the historical load data of the power system with known attributes for each similar hour for the same time period on each similar date, by using the dynamic time curvature distance. The sixth calculation unit is used to calculate the comprehensive similarity of the missing attributes for each similar scene using the following formula, based on the combined weight values ​​of known and missing attributes: In the formula: C h S represents the comprehensive similarity of the missing attribute in the h-th similar scene, where h = 1, 2, ..., H; (m,h) Let w be the similarity between the m-th known attribute and the h-th similar hour and the missing hour, where m = 1, 2, ..., M, and M is the number of known attributes in the retained historical load data of the power system; m The combined weight value of the known and missing attributes; The third acquisition unit is used to acquire scenarios where the overall similarity of the missing attributes meets the threshold range; The extraction unit is used to extract the missing attribute of the power system historical load data at the time of the cth scenario as vertical fill data. The fourth acquisition unit is used to acquire the historical load data of the power system in the c-th scenario by linear fitting of the curve as horizontal fill data; The fifth confirmation unit is used to confirm the discarding of historical load data of the power system with known attributes if the fluctuation cross-correlation coefficient is determined by the third judgment module to be no more than the comparison threshold.

6. The day-ahead dispatching system for power systems according to claim 5, characterized in that, The data clustering module includes: The first acquisition unit is used to acquire the number of clusters N and the maximum number of iterations M; The first selection unit is used to select N objects from the load and wind and solar power output data as initial cluster centers; The first calculation unit is used to calculate the Euclidean distance between the load and wind and solar power output data and the N initial cluster centers; The classification unit is used to classify load and wind and solar power output data according to Euclidean distance; The second calculation unit is used to calculate the average value of each type of object and update the initial cluster centers; The first building unit is used to construct the squared error criterion function based on load and wind and solar power output data and the updated cluster centers; The first judgment unit is used to determine whether the squared error criterion function has converged; The first confirmation unit is used to confirm the end of clustering if the squared error criterion function is determined to be converged by the first judgment unit. The second judgment unit is used to determine whether the number of iterations is greater than M if the square error criterion function is not converged as determined by the first judgment unit. The second confirmation unit is used to confirm the end of clustering when the second judgment unit determines that the number of iterations is greater than M. The third confirmation unit is used to confirm the return to execute the operation of the first calculation unit if the second judgment unit determines that the number of iterations is not greater than M. Quality evaluation unit, used to evaluate cluster quality using silhouette coefficient; The expression for the contour coefficient is: Where S is the profile coefficient; n is the total number of load and wind / solar output data; s a x is the profile coefficient for the a-th load and wind / solar output data; a Let y be the average distance between the a-th load and wind / solar output data in the x-th object and the remaining load and wind / solar output data in the x-th object, and use it as the cohesion within the class; a For x a The minimum value of the average distance to the load and solar power output data of all remaining classes except for class x.

7. The day-ahead dispatching system for power systems according to claim 6, characterized in that, The building module includes: The second building unit is used to construct the objective function of the day-ahead scheduling model for wind and solar power systems: Among them, P loss For active power losses in the power system; U i and U j These represent the voltage magnitudes at nodes i and j, respectively; N l G represents the number of branches in the power system. k Let θ be the conductance of branch k; ij The voltage phase angle difference between node i and node j; The third building unit is used to build constraints; the constraints include system power flow constraints, node voltage constraints, and wind and solar turbine operation constraints. The conditions for system power flow constraints are as follows: Among them, P Li Active power injected into node i; Q Li The reactive power injected into node i; P DGi Inject active power into node i for wind and solar power generation; Q DGi Inject reactive power into node i for wind and solar power generation; δ ij G represents the phase angle difference between node i and node j; ij B is the real part of the admittance on branch k between node i and node j; ij Let be the imaginary part of the admittance on branch k between node i and node j; The conditions for node voltage constraints are: U imin ≤U i ≤U imax ,i=1,2,…,node; Where node represents the number of nodes in the power system; U imin U is the minimum voltage amplitude at node i; imax The maximum value of the voltage amplitude at node i; The operating constraints for wind and solar turbine units are as follows: Q DGimin ≤Q DGi ≤Q DGimax ; Among them, Q DGimin The minimum reactive power injected into node i for wind and solar power generation; Q DGimax The maximum reactive power injected into node i for wind and solar power generation; The fourth building unit is used to build a wind and solar power output model; the wind and solar power output model includes a doubly-fed asynchronous wind turbine power output model and a photovoltaic power generation array power output model; The output model of the doubly-fed asynchronous wind turbine includes: Active power output P of doubly fed asynchronous wind turbine DFIG The relationship between wind speed v and wind speed v is: Among them, P e This refers to the rated power of the doubly-fed asynchronous wind turbine; v i For the cut-in wind speed; v e Rated wind speed; v o To cut off the wind speed; k1 and k2 are parameters of the wind turbine power generation system; The relationship between the active power output and reactive power output of a doubly-fed asynchronous wind turbine is as follows: Among them, P DFIG Q represents the active power of the doubly-fed asynchronous wind turbine. DFIG U represents the reactive power of the doubly-fed asynchronous wind turbine; s represents the slip rate; U s Stator-side voltage; I s X represents the stator winding current. s For stator leakage reactance; X m For the magnetizing reactance; I r This refers to the rotor-side converter current. The expression for the output model of a photovoltaic power generation array is: Where η is the photoelectric conversion efficiency; A is the effective illumination area; |Q PV | max This represents the maximum reactive power regulation capability of the photovoltaic power generation array; S PV P represents the capacity of the grid-connected inverter for the photovoltaic power generation array. PV E represents the active power of the photovoltaic array; E represents the radiation received per unit area of ​​the photovoltaic array.

8. The day-ahead dispatching system for power systems according to claim 7, characterized in that, The computing module includes: The initialization unit is used to initialize the particle population based on the supplemented historical load data of the power system, including random positions and velocities, setting gama parameters and learning factors; The seventh calculation unit is used to simulate and calculate the initial particles by substituting the objective function of the day-ahead dispatch model of the wind and solar power system as the fitness function, under the constraints and the wind and solar power output model. The first update unit is used to update the individual optimal value and the global optimal value for each particle; The second update unit is used to update the position and velocity of each particle; The fourth judgment unit is used to determine whether the number of iterations has been reached; The sixth confirmation unit is used to confirm the output of the globally optimal value as the action plan of the adjustable device when the fourth judgment unit determines that the number of iterations has been reached. The seventh confirmation unit is used to confirm the operation to be performed by the seventh calculation unit if the fourth judgment unit determines that the number of iterations has not been reached.

Images