Power supply and demand simulation method and system based on panoramic time series construction strategy

Abstract

The present invention relates to the technical field of power systems, and particularly to a power supply and demand simulation method and system based on a panoramic time series construction strategy, including: obtaining annual power supply and demand related data in several dimensions of the power system, determining the monthly optimal clustering number by using the silhouette coefficient method based on the original time series data hour by hour throughout the year; obtaining each monthly clustering center based on the monthly optimal clustering number, coupling each monthly clustering center with the original time series data to obtain an annual panoramic time series, and constructing a medium- and long-term time series production simulation model based on the annual panoramic time series; obtaining the power supply and demand time series simulation results according to the annual power supply and demand related data in each dimension and in combination with the preset medium- and long-term time series production simulation model. The present invention obtains the annual panoramic time series by adaptive clustering on a monthly basis, improves the solution speed, and thus helps to achieve efficient and accurate simulation of the power supply and demand time series.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and particularly to a power supply and demand simulation method and system based on a panoramic time series construction strategy. Background Art

[0002] A relatively high penetration rate of renewable energy increases the imbalance between power supply and demand in the long-term time scale of the power system. Therefore, it is necessary to carry out the operation planning of the power system. Among them, the time-series production simulation is an important means for the operation planning of the power system, which can perform the power balance calculation on a long-term and fine hourly time scale to simulate the power and electricity balance process. However, the medium- and long-term time-series production simulation contains a large number of integer variables and complex system constraints, making its solution time span long, problem scale large, and calculation difficult, resulting in limited application scope. Therefore, efficiently solving the production simulation model is the key to ensuring the long-term reliable operation and system planning of the power system.

[0003] In recent years, a large number of solution technologies for improving the calculation efficiency of time-series production simulation have been proposed and played an important role in practical applications. The existing method of selecting typical days is the most common. However, the existing technology mainly focuses on the time series clustering on the annual time scale when selecting typical representative time periods, and does not fully consider the data fluctuation characteristics of local time periods, resulting in low efficiency of power supply and demand time-series simulation. Summary of the Invention

[0004] In order to solve the above technical problem of low efficiency of the existing power supply and demand time-series simulation, the purpose of the present invention is to provide a power supply and demand simulation method and system based on a panoramic time series construction strategy, and the specific technical solutions adopted are as follows:

[0005] An embodiment of the present invention provides a power supply and demand simulation method based on a panoramic time series construction strategy, and the method includes the following steps:

[0006] Obtain the annual power supply and demand related data of several dimensions of the power system; wherein, the annual power supply and demand related data is load time series data, wind and light time series data, traditional unit data, hydroelectric unit data, pumped storage unit data, battery energy storage system data or wind power and photovoltaic unit data;

[0007] Based on the original hourly time series data of the whole year, use the silhouette coefficient method to determine the monthly optimal clustering number, and based on the monthly optimal clustering number, use the k-means++ clustering method to obtain the clustering center of each month; wherein, the original time series data includes load time series data and wind and light time series data;

[0008] Couple each of the monthly clustering centers with the original time series data to obtain an annual panoramic time series, and construct a medium- and long-term time series production simulation model based on the annual panoramic time series;

[0009] According to the annual power supply and demand related data of each dimension, and in combination with the preset medium- and long-term time series production simulation model, obtain the time series simulation results of power supply and demand.

[0010] Further, based on the original time series data of each hour of the whole year, using the silhouette coefficient method to determine the optimal monthly clustering number, including:

[0011] Label each data in the original time series data of each hour of the whole year with the month and date to which it belongs;

[0012] Based on the labeled original time series, determine the natural clustering number and proportional coefficient of each month; wherein, the natural clustering number is the test clustering number that maximizes the silhouette coefficient;

[0013] For any month, calculate the product of the natural clustering number and the proportional coefficient of this month, round down the product, and add a preset parameter to the rounded-down value as the optimal monthly clustering number of this month;

[0014] Obtain the optimal monthly clustering number of each month, and then obtain the set of optimal monthly clustering numbers corresponding to the original time series.

[0015] Further, the method for determining the natural clustering number of each month based on the labeled original time series includes:

[0016] The natural clustering number of the mth month The calculation formula is:

[0017] ; where, represents the function of finding the maximum of the independent variable, represents the test clustering number, represents the number of days in the mth month, d represents the dth day, represents the input data on the dth day of the mth month, that is, the time series data of 24 hours, represents when the test clustering number is the silhouette coefficient SC of the monthly input data ;

[0018] Further, the steps for obtaining the silhouette coefficient include:

[0019] For the input data of any day, on the premise that the test clustering number is , first calculate the average distance between the input data of this day and other input data within its cluster, denoted as ; Then calculate the shortest average distance between the input data of this day and the input data within other clusters, denoted as ; Finally, determine the silhouette coefficient based on the average distance and the shortest average distance corresponding to the input data of this day.

[0020] Further, when the number of test clusters is , the formula for calculating the silhouette coefficient SC of the monthly input data is:

[0021] .

[0022] Further, determine the proportionality coefficient for each month based on the labeled original time series, including:

[0023] The proportionality coefficient for the m-th month is calculated as: ; In the formula, represents the compactness of the m-th month, represents the minimum compactness, and M represents the set of month serial numbers;

[0024] The compactness of the m-th month is calculated as:

[0025] ;

[0026] In the formula, represents the natural number of clusters in the m-th month, represents the data set of the m-th month belonging to the i-th cluster center, represents the number of input data in the data set of the m-th month belonging to the i-th cluster center, represents the input data of the d-th day in the m-th month, represents the input data corresponding to the i-th cluster center, represents the distance between two input data.

[0027] Further, based on the monthly optimal number of clusters, use the k-means++ clustering method to obtain each monthly cluster center, including:

[0028] For the monthly optimal number of clusters in any month, randomly select a data point from the monthly data set of this month as the to-be-determined cluster center;

[0029] Calculate the distance between each to-be-determined cluster center and each data point in the monthly data set of this month, determine the next to-be-determined cluster center, and continuously determine and repeat the next to-be-determined cluster center until the optimal to-be-determined cluster center is selected as the monthly cluster center of this month.

[0030] Further, the coupling of each monthly clustering center with the original time series data to obtain the annual panoramic time series includes:

[0031] Taking the set composed of each monthly clustering center as the typical day set, replacing the data of each day in the original time series data with the belonging typical day, and obtaining the annual panoramic time series through splicing.

[0032] Further, the medium- and long-term time series production simulation model includes a traditional thermal power unit model, an energy storage device model, a pumped storage model, a wind power and photovoltaic model, a run-of-river hydropower model, and a network model.

[0033] Another embodiment of the present invention provides a power supply and demand simulation system based on a panoramic time series construction strategy, including:

[0034] A data acquisition module, configured to acquire annual power supply and demand related data of several dimensions of the power system; wherein, the annual power supply and demand related data is load time series data, wind and light time series data, traditional unit data, hydropower unit data, pumped storage unit data, battery energy storage system data, or wind power and photovoltaic unit data;

[0035] A model construction module, configured to determine the monthly optimal clustering number by using the silhouette coefficient method based on the original time series data of the whole year hour by hour, and obtain each monthly clustering center by using the k-means++ clustering method based on the monthly optimal clustering number; wherein, the original time series data is load time series data and wind and light time series data;

[0036] Coupling each monthly clustering center with the original time series data to obtain the annual panoramic time series, and constructing a medium- and long-term time series production simulation model based on the annual panoramic time series;

[0037] A time series simulation module, configured to obtain a power supply and demand time series simulation result according to the annual power supply and demand related data of each dimension and in combination with the preset medium- and long-term time series production simulation model.

[0038] The present invention has the following beneficial effects:

[0039] The present invention provides a power supply and demand simulation method and system based on a panoramic time series construction strategy. The method completes the power supply and demand time series simulation process through a pre-constructed medium- and long-term time series production simulation model and combines annual power supply and demand-related data in different dimensions. It adaptively clusters the time series data monthly, taking into account both the intra-day fluctuations and monthly fluctuations of wind, light, and load data, which helps to improve the efficiency of power supply and demand time series simulation, that is, to accelerate the solution speed. In the process of constructing the medium- and long-term time series production simulation model, based on the original hourly time series data for the whole year, this process fully considers the seasonal fluctuation characteristics between months on the time scale and is not prone to losing its intra-day fluctuation characteristics, ensuring the accuracy of the simulation; based on the silhouette coefficient method, the optimal monthly clustering number is determined, and based on the optimal monthly clustering number and the k-means++ clustering method, the monthly clustering centers are obtained. This process can effectively avoid the problem of unstable clustering results caused by randomly selecting the initial clustering centers. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0041] Figure 1 It is a flowchart of a power supply and demand simulation method based on a panoramic time series construction strategy according to an embodiment of the present invention;

[0042] Figure 2 It is a step flowchart of step S21 in the embodiment of the present invention;

[0043] Figure 3 It is an example diagram of the silhouette coefficient under different test clustering numbers in January in the embodiment of the present invention;

[0044] Figure 4 It is a schematic diagram of the load clustering clusters in January in the embodiment of the present invention;

[0045] Figure 5 It is a schematic diagram of the wind power data clustering clusters in January in the embodiment of the present invention;

[0046] Figure 6 It is a schematic diagram of the photovoltaic data clustering clusters in January in the embodiment of the present invention;

[0047] Figure 7 It is a schematic diagram of the load clustering center in January in the embodiment of the present invention;

[0048] Figure 8Schematic diagram of the clustering center of wind power data in January in the embodiments of the present invention;

[0049] Figure 9 Schematic diagram of the clustering center of photovoltaic data in January in the embodiments of the present invention;

[0050] Figure 10 Schematic diagram of the structure of the medium - and long - term time - series production simulation model in the embodiments of the present invention;

[0051] Figure 11 Flowchart of the execution of a power supply - demand simulation system based on a panoramic time - series construction strategy according to an embodiment of the present invention. Detailed implementation manners

[0052] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following, in combination with the accompanying drawings and preferred embodiments, details the specific implementation manners, structures, features and their effects of the technical solutions proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0053] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0054] In order to fully consider the seasonal fluctuation characteristics of the load and new energy among months, while not losing their intra - day fluctuation characteristics, a method for constructing an annual panoramic time - series with an adaptive number of clusters is proposed. Specifically, an embodiment of the present invention provides a power supply - demand simulation method based on a panoramic time - series construction strategy, as Figure 1 shown, including the following steps:

[0055] S1. Obtain the annual power supply - demand related data of several dimensions of the power system.

[0056] In this embodiment, the annual power supply - demand related data is load time - series data, wind - solar time - series data, traditional unit data, hydro - power unit data, pumped - storage unit data, battery energy storage system data or wind - power and photovoltaic unit data. Of course, the implementer can also use other - dimension annual power supply - demand related data.

[0057] S2. Based on the original hourly time - series data for the whole year, use the silhouette coefficient method to determine the optimal number of clusters for each month. Based on the optimal number of clusters for each month, use the k - means++ clustering method to obtain the clustering center for each month.

[0058] Here, the "whole year" in the hourly original time series data for the whole year means data covering a full year, that is, a complete cycle from the beginning to the end of the year. "Hourly" means the data is recorded hourly, usually to capture the power supply and demand situation every hour of the day, that is, the 8736-hour original time series data for the year. The original time series data refers to unprocessed, untreated or smoothed data, which maintains the original time order and numerical values. For the original time series data, the original time series data includes load time series data and wind-solar time series data, that is, the hourly original time series data for the whole year includes renewable energy output (wind power, photovoltaic) and load demand curves. Renewable energy output refers to the electrical output from renewable energy sources (such as wind energy, solar energy, etc.). This part of the data records the power generation fluctuations of these energy sources, while the load demand curve represents the change trend of power demand and shows the changes in electricity load at different time points (such as every hour).

[0059] Since the existing selection methods for typical representative periods mainly focus on time series clustering on the annual time scale, they do not fully consider the seasonal electricity consumption fluctuation characteristics between months, and the selection of the number of clusters is relatively random, which results in limited efficiency of time series generation simulation. Therefore, based on the hourly original time series data for the whole year in this embodiment, the silhouette coefficient method is used to adaptively determine the optimal number of monthly clusters, and then the monthly cluster center is determined as the typical representative data for subsequent construction of the medium- and long-term time series production simulation model.

[0060] The above step S2 can be implemented through steps S21 to S22 (not shown in the figure):

[0061] S21, based on the hourly original time series data for the whole year, use the silhouette coefficient method to determine the optimal number of monthly clusters.

[0062] The above step S21 can be implemented through Figure 2 the steps S211 to S214 shown:[[]]END]]

[0063] S211, label each data in the hourly original time series for the whole year with the month and date to which it belongs.

[0064] In this embodiment, labeling each data in the original time series with the month and date to which it belongs is to distinguish the input data on the same day in different months, so as to facilitate subsequent clustering processing of monthly input data.

[0065] S212, based on the labeled original time series, determine the natural number of clusters and the proportionality coefficient for each month.

[0066] Here, the natural number of clusters is the test number of clusters that maximizes the silhouette coefficient, and the proportionality coefficient is the ratio of the compactness of each month to the minimum compactness.

[0067] First, determine the natural number of clusters for each month based on the labeled original time series.

[0068] As an example, the natural number of clusters for the m-th month can be calculated by the formula:

[0069] ; where represents the function to find the maximum independent variable, represents the number of test clusters, represents the number of days in the m-th month, d represents the d-th day, represents the input data for the d-th day of the m-th month, that is, the time series data for 24 hours, represents when the number of test clusters is , the silhouette coefficient SC of the monthly input data .

[0070] In the formula for calculating the natural number of clusters, the number of test clusters is greater than or equal to 5, and the input data for the m-th month is clustered into clusters (sets), which can be denoted as . Implementers can set the value range of the number of test clusters according to specific actual situations, without specific limitations.

[0071] As a specific implementation, the steps to obtain the silhouette coefficient may include:

[0072] For the input data of the d-th day, on the premise that the number of test clusters is , assuming that is in the cluster , first calculate the average distance between the input data of this day and other input data within its cluster, denoted as ; then calculate the shortest average distance between the input data of this day and the input data in other clusters, denoted as ; finally, determine the silhouette coefficient according to the average distance and the shortest average distance corresponding to the input data of this day. The specific formula is as follows:

[0073] ;

[0074] ; ;

[0075] ;

[0076] where represents the number of data in the cluster , Denote the input data on the d-th day (the cluster it belongs to is ) and the average distance from all data in other clusters except the cluster it belongs to , for example, the average distance from the second cluster is , Denote the j-th input data in the i-th cluster in the m-th month, and min represents the minimum value function.

[0077] Obviously, when the number of test clusters is , the calculation formula of the silhouette coefficient SC of the monthly input data can be that the silhouette coefficient is also expressed as:

[0078] ;

[0079] Secondly, determine the proportionality coefficient for each month based on the labeled original time series.

[0080] In this embodiment, the proportionality coefficient can be defined as the ratio of the compactness of each month to the minimum compactness. Therefore, it is necessary to calculate the compactness of each month. For a certain cluster, a greater compactness means a greater intra-cluster distance, that is, the data is more dispersed.

[0081] The proportionality coefficient for the m-th month can be calculated as:

[0082] ; where represents the compactness of the m-th month, represents the minimum compactness, and M represents the set of month numbers.

[0083] The compactness of the m-th month can be calculated as:

[0084] ;

[0085] where represents the number of natural clusters in the m-th month, represents the set of data belonging to the i-th cluster center in the m-th month, represents the number of input data in the set of data belonging to the i-th cluster center in the m-th month, represents the input data on the d-th day of the m-th month, represents the input data corresponding to the i-th cluster center, represents the distance between two input data.

[0086] S213. For any month, calculate the product of the natural clustering number and the proportionality coefficient of that month, round down the product, and add the preset parameter to the rounded-down value to obtain the monthly optimal clustering number of that month.

[0087] As an example, the calculation formula for the monthly optimal clustering number of the m-th month can be:

[0088] ; where, represents the monthly optimal clustering number of the m-th month, represents rounding down, represents the proportionality coefficient of the m-th month, represents the natural clustering number of the m-th month, and 1 represents the preset parameter.

[0089] S214. Obtain the monthly optimal clustering number for each month, and then obtain the set of monthly optimal clustering numbers corresponding to the original time series.

[0090] In this embodiment, after obtaining the monthly optimal clustering number for each month, the value range of m is from 1 to 12, that is , , N represents the set of monthly optimal clustering numbers, that is, the set of monthly optimal clustering numbers corresponding to the original time series is obtained. The example results of the monthly optimal clustering numbers are shown in Table 1:

[0091] Table 1

[0092]

[0093] S22. Based on the monthly optimal clustering number, use the k-means++ clustering method to obtain each monthly clustering center.

[0094] In this embodiment, for the monthly optimal clustering number of any month, randomly select a data point from the monthly dataset of that month as the to-be-determined clustering center; calculate the distance between each to-be-determined clustering center and each data point in the monthly dataset of that month. The greater the distance, the greater the probability of being selected. Determine the next to-be-determined clustering center, and continuously determine and repeat the next to-be-determined clustering center until the optimal to-be-determined clustering center is selected as the monthly clustering center of that month.

[0095] It should be noted that obtaining each monthly clustering center can effectively avoid the problem of unstable clustering results caused by randomly selecting the initial clustering center.

[0096] Specifically, the steps for the k-means++ algorithm to implement obtaining each monthly clustering center include:

[0097] The first step, input month m = 1, monthly dataset , optimal clustering number And a variable k representing the current number of clusters is k = 1.

[0098] Step 2, randomly select a sample from the monthly dataset as the initial clustering center .

[0099] Step 3, calculate the shortest distance between each sample in the monthly dataset and the existing clustering centers. The calculation formula for the shortest distance can be: .

[0100] Step 4, calculate the probability that each sample is selected as the next clustering center. The calculation formula for the probability can be: , and then select the next clustering center according to the roulette wheel method, k = k + 1.

[0101] Step 5, repeat Step 3 to Step 4 until the target number of clusters k is equal to the monthly optimal number of clusters obtained in Step S21, that is .

[0102] Step 6, m = m + 1, jump to Step 2 until m is greater than 12.

[0103] So far, this embodiment has obtained monthly clustering clusters for 12 months and monthly clustering centers .

[0104] S3, couple each monthly clustering center with the original time series data to obtain an annual panoramic time series, and construct a medium - and long - term time - series production simulation model based on the annual panoramic time series.

[0105] The above - mentioned Step S3 can be implemented through the following Steps S31 to S32 (not shown in the figure):

[0106] S31, couple each monthly clustering center with the original time series data to obtain an annual panoramic time series.

[0107] In this embodiment, the set composed of each monthly clustering center is used as the typical day set, and the daily data in the original time series data is replaced by the corresponding typical day, and the annual panoramic time series is obtained through splicing.

[0108] Specifically, the input data for each month is clustered into several typical days , these typical days contain the daily fluctuation characteristics of wind, light, and load data while maintaining the monthly fluctuation characteristics. In the medium- and long-term time-series production simulation model, considering the coupling of adjacent days for operating constraints such as unit ramp rate limits, start-stop constraints, and energy storage state of charge, the daily data in the original time series is replaced by the corresponding typical day, and the annual panoramic time series is obtained by splicing.

[0109] The expression of the original time series can be:

[0110] ;

[0111] ;

[0112] In the expression of the original time series, the clustering center of the monthly clustering cluster is .

[0113] The expression of the annual panoramic time series can be:

[0114] ;

[0115] In the expression of the annual panoramic time series, i, j, and k are determined by the clustering cluster to which the corresponding original sequence belongs.

[0116] An example of the clustering cluster to which some daily data belongs is shown in Table 2:

[0117] Table 2

[0118]

[0119] In a specific embodiment, an example diagram of the silhouette coefficient under different test clustering numbers in January is as Figure 3 shown. Taking January as an example, the silhouette coefficient under different test clustering numbers is obtained, and the optimal clustering number is selected as 6.

[0120] As Figures 4 to 6 shown, taking January as an example, the clustering clusters of the load, wind power, and photovoltaic time series data are shown in the case where the clustering number is set to the optimal clustering number 6.

[0121] As Figures 7 to 9 shown, taking January as an example, the clustering centers of the load, wind power, and photovoltaic time series data are shown in the case where the clustering number is set to the optimal clustering number 6.

[0122] S32. Based on the annual panoramic time series, a medium- and long-term time-series production simulation model is constructed.

[0123] In this embodiment, the medium- and long-term time-series production simulation model includes a traditional thermal power unit model, an energy storage device model, a pumped-storage model, a wind power and photovoltaic model, a run-of-river hydropower model, and a network model. Refer to Figure 10 , the structural schematic diagram of the medium- and long-term time-series production simulation model is as shown in Figure 10 .

[0124] The expression of the traditional thermal power unit model can be:

[0125]

[0126] In the expression of the traditional thermal power unit model, is a 0-1 variable representing the unit's on-off state, is a 0-1 variable representing the unit's start-up operation, is a continuous variable representing the unit's output level, is a linear or quadratic function representing the unit's power generation cost, is the ramping power during the unit's operation, is the ramping power after the unit's start-stop, represents the unit's minimum technical output, represents the unit's maximum output, represents the unit's reserve, t represents the time, and T represents the total operation time, represents the unit's minimum shutdown time, represents the state of the unit at time t, refers to the constraint subject to, represents the start-stop operation of the unit at the i-th time.

[0127] It should be noted that the traditional thermal power unit model requires the unit's output to be within the maximum and minimum output ranges, satisfying the start-up operation, minimum start-stop time constraint, and ramping constraint.

[0128] The expression of the energy storage device model can be:

[0129]

[0130] In the expression of the energy storage device model, represents the operation cost of the energy storage device, is a 0-1 variable representing the operation state of the energy storage, is a continuous variable representing the energy storage output, is a continuous variable representing the energy storage input, represents the charging efficiency, represents the power generation efficiency, represents the state of charge of the energy storage at time t, t represents the energy storage charge-discharge time step, taking 1h.

[0131] It should be noted that the energy storage device model requires that the energy storage input and output be within the maximum and minimum output ranges, and the state of charge (SOC) meets the specified requirements.

[0132] The expression of the pumped-storage model can be:

[0133]

[0134] In the expression of the pumped-storage model, represents the operating cost of the pumped-storage unit, represents a binary variable indicating that the pumped-storage is operating in the pumping state, represents a binary variable indicating that the pumped-storage is operating in the generating state, represents a binary variable indicating the start of the pumped-storage entering the pumping state, represents a binary variable indicating the start of the pumped-storage entering the generating state, represents the pumping power of the pumped-storage, represents the generating power of the pumped-storage, represents a continuous variable of the upper reservoir capacity of the pumped-storage, represents a continuous variable of the lower reservoir capacity of the pumped-storage; represents the pumping efficiency of the pumped-storage, represents the generating efficiency of the pumped-storage.

[0135] The expression of the wind power and photovoltaic model can be:

[0136]

[0137]

[0138] In the expression of the wind power and photovoltaic model, represents the operating cost of the photovoltaic unit, represents the operating cost of the wind turbine unit, represents a continuous variable of the actual output of the photovoltaic, represents a continuous variable of the curtailed electricity of the photovoltaic, represents the installed capacity of the photovoltaic, represents the per-unit curve of the photovoltaic output; represents a continuous variable of the actual output of the wind power, represents a continuous variable of the curtailed electricity of the wind power, represents the installed capacity of the wind power, represents the per-unit curve of the wind power output.

[0139] The expression of the run-of-river hydropower model can be:

[0140]

[0141] In the expression of the runoff hydropower model, represents the operating cost of runoff hydropower, is a continuous variable representing the actual output of runoff hydropower, is a continuous variable representing the abandoned electricity of runoff hydropower, represents the installed capacity of runoff hydropower, represents the per-unit curve of the output of runoff hydropower.

[0142] The expression of the network model can be:

[0143]

[0144] In the expression of the network model, represents the loss-of-load penalty cost, is a continuous variable representing the total power of the tie line in region n, is a continuous variable representing the total input power in region n, represents the load in region n, is a continuous variable representing the load shedding in region n.

[0145] It should be noted that the network model requires the input and output of energy storage to be within the maximum and minimum output ranges.

[0146] The expression of the objective function can be:

[0147]

[0148] It is worth noting that the medium- and long-term chronological production simulation model is constructed with the goal of economic optimality, and economic optimality includes the lowest operating cost of each unit and the lowest penalty cost.

[0149] In a specific embodiment, the comparison results of the solution times of the medium- and long-term chronological production simulation are shown in Table 3:

[0150] Table 3

[0151]

[0152] As can be seen from Table 3, the medium- and long-term chronological production simulation model constructed based on the annual panoramic time series is much less in terms of solution time than the chronological production simulation model based on the original time series while ensuring the solution accuracy, verifying the effectiveness of the method provided by the present invention.

[0153] So far, the medium- and long-term chronological production simulation model has been obtained in this embodiment.

[0154] S4. According to the annual power supply and demand related data of each dimension and in combination with the preset medium- and long-term chronological production simulation model, obtain the time series simulation results of power supply and demand.

[0155] In this embodiment, the annual power supply and demand related data in all dimensions are input into a preset medium- and long-term time series production simulation model, so as to complete the simulation of the power supply and demand time series, and then obtain the power supply and demand time series simulation result.

[0156] An embodiment of the present invention further provides a power supply and demand simulation system based on a panoramic time series construction strategy. The execution flowchart of the system is as Figure 11 shown and includes:

[0157] A data acquisition module, configured to acquire annual power supply and demand related data in several dimensions of the power system; wherein, the annual power supply and demand related data is load time series data, wind and light time series data, traditional unit data, hydroelectric unit data, pumped storage unit data, battery energy storage system data or wind power and photovoltaic unit data;

[0158] A model construction module, configured to determine the monthly optimal clustering number by using the silhouette coefficient method based on the original time series data of each hour of the whole year, and obtain each monthly clustering center by using the k-means++ clustering method based on the monthly optimal clustering number; wherein, the original time series data is load time series data and wind and light time series data;

[0159] Couple each monthly clustering center with the original time series data to obtain an annual panoramic time series, and construct a medium- and long-term time series production simulation model based on the annual panoramic time series;

[0160] A time series simulation module, configured to obtain a power supply and demand time series simulation result according to the annual power supply and demand related data in each dimension and in combination with the preset medium- and long-term time series production simulation model.

[0161] In summary, the present invention adaptively clusters time series data monthly, which can take into account the intra-day and monthly fluctuations of wind, light and load data while accelerating the solution speed, and helps to improve the efficiency of power supply and demand time series simulation. The improved silhouette coefficient method is used to adaptively obtain the optimal clustering number of each month, and the k-means++ clustering method is used to cluster the monthly time series data to obtain each monthly clustering center, effectively avoiding the problem of unstable clustering results caused by randomly selecting the initial center, fully considering the data fluctuation characteristics of local time periods, improving the efficiency of power supply and demand time series simulation, and finally coupling the clustered typical days together according to the original time series to obtain an annual panoramic time series.

[0162] The above-described embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A power supply and demand simulation method based on a panoramic time series construction strategy, characterized in that: The following steps are involved: Obtaining annual power supply and demand related data of several dimensions of the power system; wherein the annual power supply and demand related data is load time series data, wind and solar time series data, traditional unit data, hydropower unit data, pumped storage unit data, battery energy storage system data or wind power photovoltaic unit data; Based on the original hourly time series data for the whole year, the silhouette coefficient method is used to determine the optimal number of monthly clusters, and based on the monthly optimal number of clusters, the k-means++ clustering method is used to obtain each monthly cluster center; wherein the original time series data includes load time series data and wind and solar time series data; Coupling each monthly cluster center with the original time series data to obtain an annual panoramic time series, and building a medium- and long-term time series production simulation model based on the annual panoramic time series; According to the annual power supply and demand related data of each dimension, and in combination with the preset medium- and long-term time series production simulation model, the power supply and demand time series simulation results are obtained; The method of determining the optimal number of monthly clusters based on the original hourly time series data for the whole year using the silhouette coefficient method includes: Label each data in the original hourly time series for the whole year with its month and day; Determine the number of natural clusters and the proportional coefficient for each month based on the annotated original time series; wherein the number of natural clusters is the number of test clusters that maximizes the silhouette coefficient; For any month, the product of the natural cluster number and the proportional coefficient of the month is calculated, the product is rounded down, and the rounded value plus the preset parameter is used as the monthly optimal cluster number of the month; Get the monthly optimal clustering number for each month, and then obtain the set of monthly optimal clustering numbers corresponding to the original time series; The method of determining the number of natural clusters for each month based on the labeled original time series includes: Number of natural clusters in month m The calculation formula is: ; In the formula, It means to find the maximum function of the independent variable. represents the number of test clusters, represents the day of the mth month, d represents the dth day, represents the input data of the dth day of the mth month, that is, 24 hours of time series data, Indicates that the number of test clusters is Monthly input data The silhouette coefficient SC; When the number of test clusters is Monthly input data The calculation formula of the silhouette coefficient SC is: ; In the formula, represents the average distance between the input data of the dth day of the mth month and other input data in its cluster. Indicates the shortest average distance between the input data on the dth day of the mth month and the input data in other clusters.

2. The power supply and demand simulation method based on the panoramic time series construction strategy according to claim 1 is characterized in that: Determine the proportion coefficient for each month based on the labeled original time series, including: Proportional coefficient for the mth month The calculation formula is: ; In the formula, Indicates the compactness of the mth month, Indicates the minimum compactness, M represents the set of month numbers; The compactness of the mth month The calculation formula is: ； In the formula, represents the number of natural clusters in the mth month, represents the data set belonging to the i-th cluster center in the m-th month, represents the number of input data in the data set belonging to the i-th cluster center in the m-th month, represents the input data of the dth day of the mth month, Represents the input data corresponding to the i-th cluster center, Represents the distance between two input data.

3. The power supply and demand simulation method based on the panoramic time series construction strategy according to claim 1 is characterized in that: The method of obtaining each monthly cluster center based on the monthly optimal cluster number by using the k-means++ clustering method includes: For the monthly optimal number of clusters in any month, a data point is randomly selected from the monthly data set of that month as the cluster center to be determined; Calculate the distance between each pending cluster center and each data point in the monthly data set of that month, determine the next pending cluster center, and continue to determine and repeat the next pending cluster center until the optimal pending cluster center is selected as the monthly cluster center of that month.

4. The power supply and demand simulation method based on the panoramic time series construction strategy according to claim 1 is characterized in that: The coupling of each monthly cluster center with the original time series data to obtain an annual panoramic time series includes: The set of cluster centers of each month is taken as the typical day set, and the daily data in the original time series data is replaced by the corresponding typical day, and the annual panoramic time series is obtained by splicing.

5. The power supply and demand simulation method based on the panoramic time series construction strategy according to claim 1 is characterized in that: The medium- and long-term time-series production simulation model includes a traditional thermal power unit model, an energy storage device model, a pumped storage model, a wind power photovoltaic model, a run-of-river hydropower model and a network model.

6. A power supply and demand simulation system based on a panoramic time series construction strategy, the system being used to implement the steps of the power supply and demand simulation method based on a panoramic time series construction strategy as described in any one of claims 1 to 5, characterized in that: include: A data acquisition module is used to acquire annual power supply and demand related data of several dimensions of the power system; wherein the annual power supply and demand related data is load time series data, wind and solar time series data, traditional unit data, hydropower unit data, pumped storage unit data, battery energy storage system data or wind power photovoltaic unit data; A model building module is used to determine the monthly optimal number of clusters based on the original hourly time series data throughout the year by using the silhouette coefficient method, and based on the monthly optimal number of clusters, obtain each monthly cluster center by using the k-means++ clustering method; wherein the original time series data are load time series data and wind and solar time series data; Coupling each monthly cluster center with the original time series data to obtain an annual panoramic time series, and building a medium- and long-term time series production simulation model based on the annual panoramic time series; The time series simulation module is used to obtain the power supply and demand time series simulation results based on the annual power supply and demand related data of each dimension and in combination with the preset medium- and long-term time series production simulation model.

Images