Method for quickly solving panoramic time sequence operation simulation problem of power system

Abstract

The invention relates to a method for rapidly solving a panoramic time sequence operation simulation problem of a power system. The method comprises the following steps: S1, constructing a unit commitment model considering intra-day power balance; s2, performing hour-by-hour operation simulation on the intra-day unit commitment model to realize accurate description of annual time sequence fluctuation characteristics; s3, determining the optimal clustering number of the time sequence of each month of the K-means clustering; s4, performing K-means clustering dimension reduction on the natural time sequence to obtain a target scene day set; s5, performing time sequence reconstruction on the target clustering scene day according to the original time sequence to form a reconstruction time sequence; and S6, solving the unit combination model of the whole year operation by using piecewise linearization of the operation cost to obtain a panoramic power system operation simulation result. According to the method, the time sequence dimension can be remarkably reduced, the model solving complexity is reduced, the model solving efficiency is remarkably improved, high precision is kept, higher precision and accuracy are shown, the simulation result is closer to the actual operation condition, and a more reliable basis is provided.

Description

Technical Field

[0001] The present invention relates to the technical field of power system operation simulation optimization, and more specifically, to a method for quickly solving a power system panoramic time sequence operation simulation problem. Background Art

[0002] The proportion of renewable energy in new power systems is gradually increasing. Since renewable energy exhibits strong uncertainty at different time scales, the significance of simulating the year-round operation of power systems is gradually becoming apparent.

[0003] The uncertainty of renewable energy in various time spans, as well as the imbalance of energy supply and demand between sources and loads at different time scales, significantly increase the complexity of the optimization planning model, and thus increase the challenges in the calculation and decision-making process. Most of the existing time series operation simulation methods are based on typical days. The power system operation optimization based on this can only ensure the feasibility of the situation on typical days, and cannot guarantee the balance of power energy in the whole year. In fact, as the proportion of renewable energy increases, the typical day is also constantly changing. In addition, the traditional 8760-hour time series operation simulation method needs to add the timing constraints of the whole year, which will result in excessive calculation and low calculation efficiency.

[0004] Therefore, the existing panoramic time-series operation simulation of the power system cannot achieve supply and demand balance on a long time scale, and the solution method cannot take into account both speed and accuracy, which makes the power balance of the power system face the risk of imbalance and cannot provide a reliable basis for operation and dispatching decisions, which is not conducive to improving the efficiency and utilization benefits of the entire energy system. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a method for quickly solving the panoramic timing operation simulation problem of an electric power system, which can solve the technical problem that the existing electric power system timing operation simulation has a small time scale and high solution complexity.

[0006] The technical solution adopted by the present invention to solve the technical problem is: constructing a method for quickly solving the panoramic timing operation simulation problem of the power system, comprising the following steps:

[0007] S1: Construct a unit combination model considering intraday power balance;

[0008] S2: Expand the daily unit combination model to the whole year to simulate hourly operation, and achieve an accurate description of the annual time series fluctuation characteristics;

[0009] S3: Use the DBI indicator to determine the optimal number of clusters for the K-means clustering time series for each month;

[0010] S4: Based on the calculated optimal number of clusters, K-means clustering is used to reduce the dimension of the natural time series to obtain the target scene day set;

[0011] S5: Reconstruct the typical days obtained by clustering and couple them together, and reconstruct the time series of the target clustering scene days according to the original time series to form a reconstructed time series;

[0012] S6: Based on the reconstructed time series, the power system timing simulation is carried out, and the unit combination model for the whole year operation is solved by piecewise linearization of the operating cost to obtain the panoramic power system operation simulation results.

[0013] According to the above scheme, in step S1, the mathematical expression of the unit combination model is as follows:

[0014]

[0015]

[0016] Wherein, formulas (1)-(3) represent the combined operating costs of the units, including operating costs and load shedding penalty costs, F Gi is the minimum operating cost of the started unit, I i,t is the startup state of unit i at time period t, F punish is the penalty coefficient for the necessary load shedding of the system, and Costs for unit startup and shutdown;

[0017] Equations (4) and (5) represent the power balance constraints of the power system at the system level and the line level. Gi,t ,P w,t is the output of the i-th conventional coal-fired power unit and wind turbine unit at time t, P Bj,t is the load value of load node j at time t, They are the node-branch, node-thermal power generator, node-wind power generator and node-load matrix element of line l, is the active power flow of the branch, is the active power output of the thermal power generating set and wind power generating set connected to line l, is the active power of the load connected to line l;

[0018] Formula (6) represents the line capacity constraint, G(b), W(b), L(b) are the number of thermal power units, wind power units and loads at node b, F l max is the normal allowable maximum value of the power flow of line l;

[0019] Formula (7) represents the climbing constraint of thermal power unit, is the up and down ramp rate of the thermal power unit in the startup state, It is the climbing constraint of the thermal power unit when it is in the changing state of startup and shutdown;

[0020] Formula (8) represents the upper and lower limits of thermal power unit output, λ i min is the minimum output rate of thermal power unit i;

[0021] Formula (9) represents the DC power flow transmission network model, θ nm is the difference between the phase angle of the starting node n and the ending node m of the branch l power flow, which linearizes the nonlinear power flow model under some reasonable assumptions and links the node phase angle with the line power flow;

[0022] Formulas (10) and (11) represent the minimum start and stop time constraints of thermal power units. The initial required fixed duration of the startup and shutdown period meets and T i on ,T i off Minimum continuous startup and shutdown time.

[0023] According to the above scheme, in step S3, the optimal number of clusters for the K-means clustering time series of each month is determined month by month using the DBI index. The specific method is: K-means clustering is performed on the time series of each month, and the K value is set to 2-7 for each month, and clustering is performed in turn, and the DBI index of each clustering result is calculated according to the clustering results. The K value when the DBI index number is the largest is the optimal K value, and the clustering result under the optimal K value is selected as the basis for subsequent analysis.

[0024] According to the above scheme, the K-means clustering includes the following steps:

[0025] S301, setting the K value to 2-7 in sequence to divide the data set into K clusters;

[0026] S302, randomly selecting K data points from the data set as initial cluster centers, i.e., centroids;

[0027] S303, calculating the distance from each data point in the data set to each centroid, and assigning the data points to corresponding clusters based on the shortest distance principle;

[0028] S304, when each cluster contains a certain number of data points, the cluster is updated, and a new centroid is determined by calculating the mean of each cluster;

[0029] S305, repeat the above process of allocating and updating the centroid until the iteration termination condition is met, that is, the clustering result remains stable and no longer changes.

[0030] According to the above scheme, the specific method of determining the K value using the DBI indicator is:

[0031] For the dataset to be classified P = {x1, x2, ..., x n}, there are n data samples, each sample is m-dimensional, assuming that several data samples are divided into K categories, with sample c i Represents class K i The center point, N(K i ) represents class K i The number of samples in d(x i ,x j ) represents the distance between samples, and the distance between samples is the Euclidean distance Class Radius is the average distance from all data in the class to the class center, and the median point between classes It is the point where the line connecting the centers of different classes is equally divided according to the ratio of the class radius;

[0032] The DBI indicator is the ratio of the intra-class data density to the inter-class data density, and the expression is:

[0033]

[0034] CD(i) in formula (12) is the class center density, which is expressed as the number of in-class data whose distance from the class center is less than the class radius, and represents the in-class data density, which is expressed as:

[0035] CD(i)=N(K' i = {x|d(x,c i ),,r i}) (13)

[0036] BD(i) in formula (12) is the class edge density, which is the number of data between any two classes whose distance to the middle point between classes is less than the mean of the class radius, indicating the inter-class dispersion, and the expression is:

[0037]

[0038] The optimal number of clusters for K-means clustering based on the DBI index corresponds to formula (12), the larger the better, that is:

[0039] K best =arg max{DBI(i),i=1,2,…,K} (15).

[0040] According to the above scheme, in step S5, the typical days obtained by clustering are reconstructed and coupled together, and the target clustering scene days are reconstructed in time series according to the original time series to form a reconstructed time series. The specific method is: based on the obtained target scene day, the typical days obtained by clustering are reconstructed according to the time series relationship of the original natural day. The principle of reconstruction coupling follows the time series logical relationship of the original natural day to form a new reconstructed time series.

[0041] According to the above scheme, in step S6, the power system timing simulation is performed based on the reconstructed time series, and the unit combination model for the whole year operation is solved by piecewise linearization of the operating costs to obtain the specific method of obtaining the panoramic power system operation simulation result: the power system panoramic timing simulation is performed on the reconstructed time series, while considering the timing coupling of the intraday system constraints and the daytime operation constraints, the operating cost of each generator is piecewise linearized, and the unit combination model for the whole year operation of the reconstructed time series is solved.

[0042] The method for quickly solving the panoramic sequential operation simulation problem of the power system of the present invention has the following beneficial effects:

[0043] 1. The present invention accurately captures the seasonal fluctuation characteristics of renewable energy generation and power load in different months by dynamically adjusting the number of clusters, and introduces the Davies-Bouldin Index (DBI) as an evaluation index to effectively measure the comprehensive effect of the internal compactness of clusters and the separation between clusters, so as to determine the optimal number of clusters month by month; it not only reflects the random fluctuations in the operation of the power system, but also deeply considers the impact of seasonal changes in natural resources on the supply and demand balance of the power system, providing a scientific basis for the flexible scheduling and planning of the power system;

[0044] 2. The present invention provides a clustering dimension reduction and reconstruction technology for time series, which realizes the feature extraction and modeling of time series data on multiple time scales, clusters the original time series data into several typical days, extracts the fluctuation characteristics of typical days of each month, and reconstructs the time series using these characteristic points, thereby significantly reducing the computational complexity of subsequent optimization problems while retaining key information, which not only improves the processing efficiency of the model, but also enhances its ability to characterize the fluctuation characteristics of different seasons;

[0045] 3. The present invention provides a multi-time scale panoramic time series operation simulation model, which comprehensively simulates the time series operation characteristics of the renewable energy power system throughout the year, fully integrates the results of time series clustering analysis, and uses the typical day and time series characteristics obtained after clustering to achieve fast and accurate simulation of the power system at different time scales. It can provide strong support for the long-term planning and daily scheduling of the power system while ensuring the calculation accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0047] Figure 1 It is a flow chart of a method for quickly solving a panoramic sequential operation simulation problem of a power system according to the present invention;

[0048] Figure 2 It is a topological diagram of a 6-node system of the present invention;

[0049] Figure 3 It is the optimal cluster number diagram for each month using K-means clustering of the present invention;

[0050] Figure 4 It is a comparison diagram of the power generation of thermal power units hour by hour in a typical day of the present invention;

[0051] Figure 5 It is a comparison chart of the power generation of thermal power units month by month in the scenario of the present invention. DETAILED DESCRIPTION

[0052] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.

[0053] like Figure 2 As shown in the figure, a method for quickly solving the panoramic timing operation simulation problem of the power system is tested in a 6-node system. 6-node system topology diagram The system contains 6 key grid nodes, which are interconnected by 13 transmission lines. In terms of generator configuration, the test case considers 4 generators, of which 3 are thermal power units and 1 is a wind power unit. Three cases are set to illustrate the effectiveness:

[0054] Case 1: Run the simulation hour by hour based on the original 8760 time series. This is a conventional benchmark solution, but the computational complexity is high and the solution time is too long.

[0055] Case 2: Cluster the whole year’s time series through K-means, reconstruct the time series of typical days based on the clustering results, and perform operation simulation based on the reconstructed time series;

[0056] The third case: This chapter proposes an improved K-means method, which performs monthly clustering through the clustering effectiveness index DBI, and generates a new panoramic time series by clustering dimension reduction and reconstruction of the full-year time series to support the efficient operation simulation of the panoramic time series;

[0057] The selection of traditional typical days is based on the annual fluctuation time series. The target scenario day time series is selected by K-means clustering algorithm. This method is not good at describing the differences in random volatility between months. Due to the influence of renewable energy volatility in different seasons, the clustering results should be different between different months. This paper determines the optimal K value of K-means clustering by DBI coefficient month by month, and reflects the seasonal fluctuation characteristics of different months by the difference in the number of clusters.

[0058] To simplify the calculation, some reasonable assumptions are made when substituting into the system for simulation:

[0059] ① Modeling of wind power, a typical renewable energy source, is carried out through hourly time series modeling;

[0060] ② Select typical scenarios by hour and month, and model the typical scenarios as typical days;

[0061] ③ Assume that in the original natural day corresponding to the same typical day, the operating status of the power system, including the start and stop status of the generator set, load power, etc., are the same during the same time period of the day.

[0062] In this 6-node system, the improved K-means method proposed in the third case performs monthly clustering through the clustering effectiveness index DBI, and generates a new panoramic time series by clustering dimension reduction and reconstruction of the annual time series to support the efficient operation simulation method of the panoramic time series, such as Figure 1 As shown, the following steps are included:

[0063] S1: Construct a unit combination model considering intraday power balance;

[0064] S2: The daily unit combination model is extended to 8760 hours throughout the year to perform hourly operation simulation to achieve an accurate description of the annual time series fluctuation characteristics. The mathematical expression of the unit combination problem is as follows:

[0065]

[0066] Wherein, formulas (1)-(3) represent the combined operating costs of the units, including operating costs and load shedding penalty costs, F Gi is the minimum operating cost of the started unit, I i,t is the startup state of unit i at time period t, F punish is the penalty coefficient for the necessary load shedding of the system, and Costs for unit startup and shutdown;

[0067] Equations (4) and (5) represent the power balance constraints of the power system at the system level and the line level. Gi,t ,Pw,t is the output of the i-th conventional coal-fired power unit and wind turbine unit at time t, P Bj,t is the load value of load node j at time t, They are the node-branch, node-thermal power generator, node-wind power generator and node-load matrix element of line l, is the active power flow of the branch, is the active power output of the thermal power generating set and wind power generating set connected to line l, is the active power of the load connected to line l;

[0068] Formula (6) represents the line capacity constraint, G(b), W(b), L(b) are the number of thermal power units, wind power units and loads at node b, F l max is the normal allowable maximum value of the power flow of line l;

[0069] Formula (7) represents the climbing constraint of thermal power unit, is the up and down ramp rate of the thermal power unit in the startup state, It is the climbing constraint of the thermal power unit when it is in the changing state of startup and shutdown;

[0070] Formula (8) represents the upper and lower limits of thermal power unit output, λ i min is the minimum output rate of thermal power unit i;

[0071] Formula (9) represents the DC power flow transmission network model, θ nm is the difference between the phase angle of the starting node n and the ending node m of the branch l power flow, which linearizes the nonlinear power flow model under some reasonable assumptions and links the node phase angle with the line power flow;

[0072] Formulas (10) and (11) represent the minimum start and stop time constraints of thermal power units. The initial required fixed duration of the startup and shutdown period meets and T i on ,T i off Minimum continuous startup and shutdown time.

[0073] S3: Use the DBI index to determine the optimal number of clusters for the K-means clustering of each month's time series on a monthly basis; the specific method is: perform K-means clustering on each month's time series on a monthly basis, set the K value to 2-7 for each month, cluster them in turn, and calculate the DBI index for each clustering result based on the clustering results. The K value when the DBI index number is the largest is the optimal K value, and select the clustering result under the optimal K value as the basis for subsequent analysis.

[0074] Among them, K-means clustering includes the following steps:

[0075] S301, setting the K value to 2-7 in sequence to divide the data set into K clusters;

[0076] S302, randomly selecting K data points from the data set time series as initial clustering centers, i.e., centroids;

[0077] S303, calculating the distance from each data point in the data set to each centroid, and assigning the data points to corresponding clusters based on the shortest distance principle;

[0078] S304, when each cluster contains a certain number of data points, the cluster is updated, and a new centroid is determined by calculating the mean of each cluster;

[0079] S305, repeating the above process of allocating and updating the centroid until the iteration termination condition is met, that is, the clustering result remains stable and no longer changes;

[0080] Among them, the specific method of using the DBI index to determine the optimal number of clusters for the K-means clustering time series of each month is:

[0081] For the dataset to be classified P = {x1, x2, ..., x n Here, we use the wind power data obtained by a meteorological bureau in one year as the time series. There are 365 data samples in total, and the sample dimension is 1. Assume that several data samples are divided into K categories, and use sample c i Represents class K i The center point, N(K i ) represents class K i The number of samples in d(x i ,x j ) represents the distance between samples, and the distance between samples is the Euclidean distance Class Radius is the average distance from all data in the class to the class center, and the median point between classes It is the point where the line connecting the centers of different classes is equally divided according to the ratio of the class radius;

[0082] The DBI indicator is the ratio of the intra-class data density to the inter-class data density, and the expression is:

[0083]

[0084] CD(i) in formula (12) is the class center density, which is expressed as the number of in-class data whose distance from the class center is less than the class radius, and represents the in-class data density, which is expressed as:

[0085] CD(i)=N(K' i = {x|d(x,c i ),,r i}) (13)

[0086] BD(i) in formula (12) is the class edge density, which is the number of data between any two classes whose distance to the middle point between classes is less than the mean of the class radius, indicating the inter-class dispersion, and the expression is:

[0087]

[0088] The optimal number of clusters for K-means clustering based on the DBI index corresponds to formula (12), the larger the better, that is:

[0089] K best =arg max{DBI(i),i=1,2,…,K} (15)

[0090] S4: Based on the calculated optimal number of clusters, K-means clustering is used to reduce the dimension of the natural time series to obtain the target scene day set;

[0091] Each month can be clustered into several typical days. The operating status of the power system in these typical days is consistent. The daily fluctuation of renewable energy output is reflected by the output of wind turbines in the day. The fluctuation characteristics of the monthly power renewable energy of the system can be characterized by combination, thus achieving the characterization of characteristics at different time scales.

[0092] S5: Reconstruct the typical days obtained by clustering and couple them together, and reconstruct the time series of the target clustering scene days according to the original time series to form a reconstructed time series;

[0093] Based on the obtained target scenario day, the typical day obtained by clustering is reconstructed according to the temporal relationship of the original natural day. The principle of reconstruction coupling follows the temporal logic relationship of the original natural day, ensuring that the reconstructed time series can truly reflect the actual operation status of the system and form a new reconstructed time series.

[0094] S6: Based on the reconstructed time series, the power system timing simulation is performed, and the unit combination model for the whole year operation is solved by piecewise linearization of the operating cost to obtain the panoramic power system operation simulation results;

[0095] The reconstructed time series is simulated for the entire power system, and the timing coupling between intraday system constraints and daytime operation constraints is considered. The operating cost of each generator is piecewise linearized, and the unit combination model for the whole year operation is solved for the reconstructed time series.

[0096] According to calculations, from the comparison of various schemes in Table 1 below from the perspectives of unit power generation, power generation cost, number of optimization variables and solution time, it can be seen that when comparing the total power generation of thermal power units in a whole year with the power generation on a specific typical day, different schemes all show low calculation errors in the long time scale (i.e. annual power generation). Compared with Situation 1, Situation 2 has only a small relative error difference of 1.0% in the calculation of the annual power generation of thermal power units; however, when focusing on a more detailed intraday operation scenario, the error of Scheme ② is more significant, with an error rate of up to 30.4%, while Scheme 2 has a relatively small relative error difference of 1.0% in the calculation of the annual power generation of thermal power units. Case 3 shows relatively small errors in both long and short time scales (i.e., daily output), with error rates as low as 2.3% and 0.4%, respectively, indicating that Case 3 has high accuracy in predicting and calculating the power generation of thermal power units; in terms of the cost per kilowatt-hour, although Case 2 has achieved a lower solution error in terms of the cost per kilowatt-hour, it shows obvious deficiencies in the accuracy of other calculation results compared with Case 3; in terms of the number of optimization variables and solution time, Case 3 has a significant reduction compared with the baseline solution, and has also been greatly improved compared with Case 2. Comprehensively considered, Case 3 not only performs well in improving the efficiency of solving optimization problems, but also maintains a certain degree of solution accuracy. This result fully verifies the effectiveness of the operation simulation solution method provided by the present invention.

[0097] Table 1 Comparison of simulation results of various schemes

[0098]

[0099]

[0100] in, Figure 3 The optimal number of clusters for each month is shown. These numbers of clusters vary with the change of months, which actually reflects the random fluctuation characteristics of time series data in these months. Figure 4 It shows the comparison of power generation of thermal power units on a typical day. Figure 5 The comparison of the power generation of thermal power units in each month of the scenario year is demonstrated, which further verifies that the proposed method for quickly solving the panoramic time-series operation simulation problem of the power system shows higher precision and accuracy than other situations when simulating the fluctuation characteristics of power and electricity at multiple time scales. The simulation results are closer to the actual operating conditions, providing a more reliable basis for power system planning, scheduling and operation management.

[0101] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are within the protection of the present invention.

Claims

1. A method for quickly solving a panoramic sequential operation simulation problem of a power system, characterized in that: The following steps are involved: S1: Construct a unit combination model considering intraday power balance; S2: Expand the daily unit combination model to the whole year to simulate hourly operation, and achieve an accurate description of the annual time series fluctuation characteristics; S3: Use the DBI indicator to determine the optimal number of clusters for the K-means clustering time series for each month; S4: Based on the calculated optimal number of clusters, K-means clustering is used to reduce the dimension of the natural time series to obtain the target scene day set; S5: Reconstruct the typical days obtained by clustering and couple them together, and reconstruct the time series of the target clustering scene days according to the original time series to form a reconstructed time series; S6: Based on the reconstructed time series, the power system timing simulation is carried out, and the unit combination model for the whole year operation is solved by piecewise linearization of the operating cost to obtain the panoramic power system operation simulation results.

2. The method for quickly solving the panoramic sequential operation simulation problem of the power system according to claim 1 is characterized in that: In step S1, the mathematical expression of the unit combination model is as follows: Wherein, formulas (1)-(3) represent the combined operating costs of the units, including operating costs and load shedding penalty costs, F Gi is the minimum operating cost of the started unit, I i,t is the startup state of unit i at time period t, F punish is the penalty coefficient for the necessary load shedding of the system, and Costs for unit startup and shutdown; Equations (4) and (5) represent the power balance constraints of the power system at the system level and the line level. Gi,t ,P w,t is the output of the i-th conventional coal-fired power unit and wind turbine unit at time t, P Bj,t is the load value of load node j at time t, They are the node-branch, node-thermal power generator, node-wind power generator and node-load matrix element of line l, is the active power flow of the branch, is the active power output of the thermal power generating set and wind power generating set connected to line l, is the active power of the load connected to line l; Formula (6) represents the line capacity constraint, G(b), W(b), L(b) are the number of thermal power units, wind power units and loads at node b, F l max is the normal allowable maximum value of the power flow of line l; Formula (7) represents the climbing constraint of thermal power unit, is the up and down ramp rate of the thermal power unit in the startup state, It is the climbing constraint of the thermal power unit when it is in the changing state of startup and shutdown; Formula (8) represents the upper and lower limits of thermal power unit output. is the minimum output rate of thermal power unit i; Formula (9) represents the DC power flow transmission network model, θ nm is the difference between the phase angle of the starting node n and the ending node m of the branch l power flow, which linearizes the nonlinear power flow model under some reasonable assumptions and links the node phase angle with the line power flow; Formulas (10) and (11) represent the minimum start and stop time constraints of thermal power units. The initial required fixed duration of the startup and shutdown period meets and T i on ,T i off Minimum continuous startup and shutdown time.

3. The method for quickly solving the panoramic sequential operation simulation problem of the power system according to claim 1 is characterized in that: In step S3, the optimal number of clusters for the K-means clustering time series of each month is determined month by month using the DBI index. The specific method is: K-means clustering is performed on the time series of each month month by month, and the DBI index of each clustering result is calculated based on the clustering results. The K value when the DBI index number is the largest is the optimal K value, and the clustering result under the optimal K value is selected as the basis for subsequent analysis.

4. The method for quickly solving the panoramic sequential operation simulation problem of the power system according to claim 3 is characterized in that: The K-means clustering includes the following steps: S301, setting the K value to 2-7 in sequence to divide the data set into K clusters; S302, randomly selecting K data points from the data set as initial cluster centers, i.e., centroids; S303, calculating the distance from each data point in the data set to each centroid, and assigning the data points to corresponding clusters based on the shortest distance principle; S304, when each cluster contains a certain number of data points, the cluster is updated, and a new centroid is determined by calculating the mean of each cluster; S305, repeat the above process of allocating and updating the centroid until the iteration termination condition is met, that is, the clustering result remains stable and no longer changes.

5. The method for quickly solving the panoramic sequential operation simulation problem of the power system according to claim 3 is characterized in that: The specific method of determining the K value using the DBI indicator is: For the dataset to be classified P = {x1, x2, ..., x n }, there are n data samples, each sample is m-dimensional, assuming that several data samples are divided into K categories, with sample c i Represents class K i The center point, N(K i ) represents class K i The number of samples in d(x i ,x j ) represents the distance between samples, and the distance between samples is the Euclidean distance Class Radius is the average distance from all data in the class to the class center, and the median point between classes It is the point where the line connecting the centers of different classes is equally divided according to the ratio of the class radius; The DBI indicator is the ratio of the intra-class data density to the inter-class data density, and the expression is: CD(i) in formula (12) is the class center density, which is expressed as the number of in-class data whose distance from the class center is less than the class radius, and represents the in-class data density, which is expressed as: CD(i)=N(K′ i ={x∣d(x,c i ),,r i }) (13) BD(i) in formula (12) is the class edge density, which is the number of data between any two classes whose distance to the middle point between classes is less than the mean of the class radius, indicating the inter-class dispersion, and the expression is: The optimal number of clusters for K-means clustering based on the DBI index corresponds to formula (12), the larger the better, that is: K best =arg max{DBI(i),i=1,2,…,K} (15)。 6. The method for quickly solving the panoramic sequential operation simulation problem of the power system according to claim 1 is characterized in that: In step S5, the typical days obtained by clustering are reconstructed and coupled together, and the target clustering scene days are reconstructed in time series according to the original time series to form a reconstructed time series. The specific method is: based on the obtained target scene day, the typical days obtained by clustering are reconstructed according to the time series relationship of the original natural day, and the principle of reconstruction coupling follows the time series logical relationship of the original natural day to form a new reconstructed time series.

7. The method for quickly solving the panoramic sequential operation simulation problem of the power system according to claim 1 is characterized in that: In the step S6, the power system timing simulation is performed based on the reconstructed time series, and the unit combination model for the whole year operation is solved by piecewise linearization of the operating costs to obtain the specific method of obtaining the panoramic power system operation simulation result: the power system panoramic timing simulation is performed on the reconstructed time series, while considering the timing coupling of the intraday system constraints and the daytime operation constraints, the operating cost of each generator is piecewise linearized, and the unit combination model for the whole year operation is solved for the reconstructed time series.

Images